JP5057964B2 - Video signal processing device, video signal processing method, and video signal display device - Google Patents

Description

  The present invention relates to a video signal processing apparatus for removing a noise in a video signal and converting a scanning signal (interlace) signal into a scanning signal to a progressive scanning signal to obtain a progressive signal in which the noise of the video signal is reduced, The present invention relates to a video signal processing method and a video signal display device to which the video signal processing device or method is applied.

  In recent years, with the increase in the size of display screens of television receivers and the increase in image quality of video signals, when video signals are processed and displayed in high image quality, unnecessary components included in the input video signal, that is, noise components In addition, high signal reliability is required in video signal processing. In order to reduce the noise component contained in this video signal and obtain a high-quality video, there is a noise removal (also called noise reduction, abbreviated as NR) processing, for example, one frame or several frames before using a frame memory. A number of three-dimensional noise removal processes have been proposed in which a noise component having no correlation between frames (frame correlation) in the time axis is obtained. There is also a proposal for a two-dimensional noise removal process that removes noise in a frame of an image by using a one-dimensional filter or a two-dimensional filter process without using a frame memory.

  On the other hand, when displaying an interlaced television signal with high image quality, a process of converting the interlaced signal into a progressive signal (referred to as IP conversion) is performed. In IP conversion, in general, when a video is a still image, interpolation (inter-field interpolation) is performed by fitting pixels of corresponding scanning lines in temporally adjacent fields, and the video is a moving image. In this case, it is known to perform motion adaptive IP conversion in which an interpolation signal is generated by pixels on adjacent scanning lines in the same field (intra-field interpolation). Further, it is detected from the movement of the video signal whether the video signal to be processed is a video signal subjected to telecine conversion (for example, a 2-3 pulldown signal and a 2-2 pulldown signal, referred to as a telecine video signal) (telecine detection). For a telecine video signal, the progressive signal before being telecine-converted by the adjacent field before or after each field converted from the same frame is restored and converted to a progressive video signal. IP conversion may be performed. Hereinafter, IP conversion for a telecine video signal is referred to as telecine interpolation. These IP conversions require a signal of one frame or several frames before using a frame memory, and are three-dimensional IP conversions.

  Conventionally, in the three-dimensional noise removal processing, the output signal after noise removal is delayed by one frame, and a certain coefficient (cyclic coefficient) is used by using a difference between frames of the signal one frame before and the current frame signal. A cyclic noise removal process (called frame cyclic noise removal process) that removes noise components by adding / subtracting the product of the current frame signal to / from the current frame signal, or by delaying the current frame signal by multiple frames. There is a non-cyclic type noise removal process (referred to as a non-cyclic type noise removal process) in which a noise component is removed by multiplying the above signal by a coefficient and filtering using a signal of a plurality of frames.

  In the case of non-cyclic noise removal, the calculation is performed using the frame correlation between the current frame and the frame obtained by the frame delay. However, in the case of frame cyclic noise removal, the video signal after noise removal is frame-delayed. Therefore, the calculation is performed not only between the number of frames to be used but also over many frames. In the non-cyclic type noise removal process, it is necessary to increase the number of frames of the frame delay signal that can be used in order to improve the noise removal effect. Therefore, in order to improve the effect, the circuit scale in the video signal processing apparatus increases. Actually, since the number of frame delay signals that can be used is limited, the noise removal effect is limited. Cyclic noise removal has a larger noise removal effect than non-cyclic noise removal, but in the case of a video signal with a lot of movement, it is due to frame correlation between limited frames. There is an advantage that tailing and afterimage can be reduced.

  In such a three-dimensional noise removal process, noise components having no frame correlation on the time axis can be removed, and a noise removal effect can be obtained. On the other hand, tails, afterimages, and outline blurring occur in a portion with motion. There is a case. Therefore, conventionally, the above coefficients and filter coefficients are calculated using motion detection information obtained from differences between frames and the like. Pulling and afterimages are reduced (see, for example, Patent Document 1).

  In addition, the two-dimensional noise removal processing that removes noise within a frame by using a one-dimensional filter or two-dimensional filter processing without using a frame-delayed signal as in the three-dimensional noise removal processing is performed by reducing a signal below a certain threshold to zero. There are known a method using coring and a method of applying a low-pass filter process in the horizontal and vertical directions. Noise can be reduced without causing tailing and afterimages in the moving part as in the three-dimensional noise removal processing, but due to coring and filtering, contour blurring, contour degradation, and the like occur.

  Therefore, in the conventional noise removal device, the frame cyclic noise removal device and the non-cyclic noise removal device by two-dimensional processing are combined to control noise removal based on the motion detection signal and the video signal level. In consideration of video motion and video signal level, there is no afterimage where there is motion, and noise that corresponds to the motion and level of the video signal is reduced by focusing on low level signals when stationary. There is a proposal of an apparatus for performing removal (see, for example, Patent Document 2).

On the other hand, in a video signal processing apparatus for obtaining a progressive signal with reduced noise in the video signal, a progressive signal is obtained by performing a three-dimensional IP conversion process on the video signal from which noise has been removed by the noise removing apparatus. Can be considered. In this case, it is necessary to obtain a frame (or field) delayed signal for performing three-dimensional IP conversion after performing the noise removal processing. In addition, it is conceivable to apply the above-described noise removing device to the progressive signal after the three-dimensional IP conversion. In this case, however, it is necessary to delay the progressive signal by a frame in order to remove the three-dimensional noise. . Therefore, a separate delay circuit is provided in the noise removal apparatus and the IP conversion process, and the circuit scale of the video signal processing apparatus is increased and the cost is increased.
For this reason, the frame delay signal used in the three-dimensional noise removal process can be used also in the three-dimensional IP conversion so as to obtain a progressive signal with reduced noise.

  Furthermore, in the conventional video signal processing apparatus, the cyclic noise removing unit divides the difference frequency band into a low frequency band and a high frequency band with respect to the difference from the image one field before the motion vector correction. To obtain a motion detection signal for each band and control the amount of feedback to improve the afterimage reduction effect in the high frequency part, remove noise, and obtain an image signal one field before used in this noise removal part There is a proposal of an apparatus used by the delay circuit and the motion vector correction circuit in order to obtain an image signal of one field before used for obtaining a progressive signal (see, for example, Patent Document 3).

JP-A-9-81754 (FIG. 1) Japanese Patent No. 3611773 (FIG. 1) Japanese Patent Laying-Open No. 2004-96628 (FIGS. 2 and 3)

As described above, in the above-described conventional apparatus, cyclic noise removal processing is performed in order to sufficiently remove noise, motion detection is performed based on a difference value between frames, and motion is detected according to the result of motion detection. Reduces the amount of recursion and afterimages by reducing the amount of recursion in parts and reducing the effect of noise removal, and by removing non-circular noise in moving parts based on the motion and level of the video using two-dimensional processing. And noise was removed.
Therefore, in order to reduce tailing and afterimages, it is necessary to weaken the noise removal effect at the part detected as motion, and noise reduction cannot be sufficiently performed. Since tailing and afterimages occur, there is a problem that good noise removal cannot be performed. In addition, when noise is removed in a two-dimensional manner from a portion detected as motion, contour blurring or contour degradation occurs due to filter processing. In this case, noise components having no frame correlation cannot be removed in the motion portion. There is a problem that noise cannot be sufficiently reduced.

  In addition, as described above, in order to obtain a desired noise removal effect, it is necessary to increase the number of frames of the frame delay signal to be used in order to improve the noise removal effect by performing acyclic noise removal. There is a problem that the circuit scale becomes large.

  In addition, as described above, when a progressive signal is obtained by performing three-dimensional IP conversion after performing the cyclic noise removal, if a separate delay circuit is provided in the noise removal device and the IP conversion processing, the video The circuit scale of the signal processing device becomes large and expensive. Further, as in the above-described conventional device, a delay circuit and a motion vector for obtaining an image signal one field before used in the cyclic noise removing unit. Even when the correction circuit is a signal one field before used for three-dimensional IP conversion, the frame delay signal used in the cyclic noise removing unit and the frame delay signal for obtaining the progressive signal are shared, but cyclic noise removal is used. In noise removal, in order to reduce tailing and afterimages, the noise removal effect at the part detected as motion becomes weaker, and the noise removal effect Problem of tailing and afterimage occurs when strong will remain. Therefore, there is a problem that noise cannot be sufficiently reduced, good noise removal cannot be performed, and noise reduction of the obtained progressive signal is not sufficient.

  Accordingly, the present invention has been made to solve the above-described problems, and an object of the present invention is to efficiently implement a delay circuit for a video signal by using a delayed video signal common to noise removal processing and three-dimensional IP conversion processing. It is possible to perform acyclic noise removal that can be sufficiently performed without significant increase in circuit scale, and it has a good noise removal effect that reduces the negative effects of tailing and afterimages. Provided is a video signal processing device, a video signal processing method, and a video signal display device that can obtain a progressive signal from which noise has been removed and obtain a progressive signal from which noise has been removed at a lower cost without a significant increase in circuit scale. There is to do.

The video signal processing apparatus of the present invention is
A video signal processing apparatus for removing a noise component having no correlation between fields or frames of a video signal, and obtaining a sequentially scanned video signal from which the noise component has been removed,
A plurality of delays by delaying an input video signal or a video signal of a current field or frame composed of a video signal obtained by cyclic noise removal processing in units of field periods, which are periods divided by a vertical synchronization signal Field delay means for outputting a video signal;
A video signal of a plurality of fields or frames composed of the video signal of the current field or frame and the plurality of delayed video signals is input, and a motion is obtained from a difference between signals of the plurality of fields or frames. A non-cyclic noise removal processing means for performing a filtering process controlled by movement to remove noise and generating a signal after acyclic noise removal;
When the input video signal is an interlaced scanning video signal,
A scanning line interpolation signal is generated from one or a combination of intra-field interpolation processing and inter-field interpolation processing from the video signals of the plurality of fields or frames and the signal after the non-cyclic noise removal, and interlaced scanning is performed. Three-dimensional scanning line conversion processing means for converting a video signal into a progressive scanning video signal and generating a progressive scanning video signal from which noise components have been removed;
Depending on whether the input video signal is an interlaced scanning video signal or a progressive scanning video signal, either the signal after the non-cyclic noise removal or the progressive scanning video signal from the three-dimensional scanning line interpolation processing means is selected. And a switching means for outputting as a sequentially scanned video signal from which noise components have been removed.

The video signal display device of the present invention is
The above video signal processing device;
Display means;
And display processing means for causing the display means to display an image based on the progressive scanning video signal from which the noise component output from the video signal processing apparatus has been removed.

  According to the video signal processing apparatus of the present invention, the field delay video signal used in the noise removal process and the three-dimensional IP conversion process is obtained by the common field delay means, and the field delay means such as the field memory is not greatly increased. A non-cyclic noise removal signal that removes noise components having no frame correlation over a plurality of fields or frames can be used to perform three-dimensional scanning line interpolation processing. Prevents blurring and deterioration of edges, contours, and signals after filtered acyclic noise removal processing are obtained, the circuit scale does not increase significantly, and the adverse effects of tailing and afterimage are reduced. There is an effect that it is possible to obtain a progressive signal from which noise is well removed.

  In addition, according to the video signal display device of the present invention, the circuit scale does not increase significantly, the edge of the moving portion, the blurring and deterioration of the contour are prevented, the noise removal process is performed, and the negative effects of tailing and afterimage are prevented. A reduced progressive signal can be obtained, and a high-quality video using the progressive signal from which noise has been well removed can be displayed.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a video signal processing apparatus according to Embodiment 1 of the present invention (that is, an apparatus capable of performing the video signal processing method according to Embodiment 1), and is delayed by a field or a frame. A non-recursive three-dimensional noise removal process (non-recursive noise removal process) that removes a noise component from a signal and a current field or frame signal (video signal) using a plurality of field or frame signals. 3D scanning line conversion based on the signal after noise removal, the delayed signal (delayed video signal) used in the noise removal processing, and the current field signal as well as non-cyclic NR or FIR-NR) It is configured to perform (three-dimensional IP conversion).

  The video signal processing apparatus 1 shown in FIG. 1 includes first to seventh field memories 11 to 17, a switching signal generation unit 18, an acyclic noise removal processing unit 20, a three-dimensional IP conversion processing unit 65, And switching means 62.

  In the first to seventh field memories 11 to 17, an input video signal (hereinafter also referred to as an input signal) Di0 is sequentially input as a current field or frame signal (also referred to as an “input field or frame signal”). Each of the first to seventh field memories 11 to 17 is a period (referred to as “vertical period” or “field period”, simply called “field”), which is divided by two vertical synchronization signals that precede and follow the video signal. The field memory is also referred to as “field delay means”.

The first to seventh field memories 11 to 17 as field delay means are memories for delaying each input video signal by one field (one vertical period) and outputting a delayed signal (delayed video signal). The first field memory 11 delays the input video signal Di0 by one field and outputs a one-field delayed signal d1f, and the second field memory 12 receives the one-field delayed signal d1f from the first field memory 11. Delays by one field and outputs a two-field delay signal d2f.
Thereafter, similarly, the delay signal is sequentially delayed by 1 field in each field memory, the 3 field delay signal d3f from the third field memory 13, the 4 field delay signal d4f from the fourth field memory 14 and the fifth field memory. The field memory 15 outputs a 5-field delay signal d5f, the sixth field memory 16 outputs a 6-field delay signal d6f, and the seventh field memory 17 outputs a 7-field delay signal d7f. As a result, a delay signal used in noise removal processing and three-dimensional IP conversion is obtained.

  The first to seventh field memories 11 to 17 may be called “frame memory” or “frame delay means”. This is because one frame of data is written in each vertical period when the input signal is a progressive signal.

  The switching signal generating means 18 generates a switching signal pf indicating whether the input signal is an interlace signal or a progressive signal.

  The acyclic noise removal processing means 20 removes noise components by performing filtering using signals of pixels at the same position in a plurality of fields or frames having pixels at the same position in the screen.

  When the input signal is an interlace signal, the three-dimensional IP conversion processing means 65 generates a scanning line interpolation signal by intra-field interpolation processing and inter-field interpolation processing, and converts the interlace signal into a progressive signal.

  The switching means 62 uses the switching signal pf from the switching signal generating means 18 to perform an IP conversion progressive signal Ips from the three-dimensional IP conversion processing means 65 and noise removal processing (NR processing) from the acyclic noise removal processing means 20. ) Is selected, and an output video signal Do0, which is a progressive signal, is selected and output. That is, when the input signal is an interlace signal, the progressive signal Ips from the three-dimensional IP conversion processing means 65 is selected, and when the input signal is a progressive signal, the signal Rnr from the non-cyclic noise removing means 20 is selected. Select.

  The three-dimensional IP conversion processing unit 65 includes an intra-field interpolation processing unit 30, an interpolation motion detection unit 66, an interpolation signal generation unit 60, and a time axis conversion unit 61.

  The intra-field interpolation processing means 30 is a signal of an interpolation target pixel, that is, an interpolation signal (an “interpolation scanning line signal” or “interpolation scanning line signal”) or “ Interpolated line signal ").

  The interpolation motion detection means 66 includes a motion detection means 40 and a telecine detection means 50 for detecting whether or not the video signal is a telecine video signal. The interpolated motion detection means 66 detects the motion of the signal from the difference between fields (or between frames) of the video signal. Video motion) is detected, and interpolation switching signals mdt and tcif are generated based on the detection result.

In addition, when the difference between the fields which takes the difference between the pixels of the same position of two fields where the pixels are at the same position is detected, the difference between the fields is also a difference between frames. In this sense, “difference between fields” may be simply referred to as “difference between frames” or “frame difference”. When the input video signal is an interlaced signal, the field in which the pixels are at the same position appears once every two field periods. Therefore, “two field periods” is called “one frame period”, and “difference between two fields” is It is called “difference between one frame”, and “difference between four fields” is sometimes referred to as “difference between two frames”.
For the same reason, the correlation obtained by the difference between the fields is also called the frame correlation because the correlation is based on the difference between the fields where the pixels are at the same position.

  The interpolation signal generation unit 60 performs noise removal (abbreviated as “NR”) processing from the non-cyclic type noise removal processing unit 20 based on the interpolation switching signals mdt and tcif output from the interpolation motion detection unit 66. An interpolation signal (interpolation line signal) Im that has been three-dimensionally processed by motion adaptive processing or telecine interpolation is generated from the generated interpolation signals Isnr, Itnr and the signal from the intra-field interpolation processing means 30, and output. Thus, the interpolation switching signals mdt and tcif are used to generate an interpolation signal from the intra-field interpolation process and the inter-field interpolation process.

  The time axis conversion means 61 corresponds to the actual scanning line signal (real line signal) Rnr in the interpolation field from which the interpolation signal Im output from the interpolation signal generation means 60 is denoised from the acyclic noise removal processing means 20. Between the scanning lines to be converted, double-speed converted, converted into a progressive signal, and output.

FIG. 2 is a block diagram showing a configuration example of the non-cyclic noise removal processing means 20.
The non-recursive noise removal processing means 20 shown in the figure, with respect to the signal of the noise removal target pixel in the field or frame (noise removal target field or frame) that is the target of noise removal, Is filtered using the signal of the pixel located at the same position as the above-mentioned noise removal target pixel in the other three fields or frames located at the same position in the screen, that is, the signal of 4 fields or 4 frames in total. The noise is removed by processing, and a signal from which noise has been removed is output.
When the interlace signal is input, the acyclic noise removal processing means 20 performs noise removal processing on the real line signal in the interpolation field, outputs the noise-removed real line signal, and temporally outputs to the interpolation field. Interpolation scan in which noise removal processing is performed on the signal of the real line ("interpolation scan line" or "interpolation line") in the adjacent field at the same position as the interpolation line of the interpolation field. A line signal (interpolation line signal) is output.
This noise-removed interpolation line signal is also called “noise-removed still image interpolation line signal”, and is used for inter-field interpolation in the three-dimensional IP conversion processing means 65.

  The acyclic noise removal processing means 20 shown in FIG. 2 includes first to fourth switching means 201, 202, 203, real line noise removal means 21, and interpolation line noise removal means 22.

  The switching means 201, 202, 203 select the field delay signal based on the switching signal pf from the switching signal generation means 18.

  When the interlace signal is input, the real line noise removing unit 21 outputs the signal of the noise removal target pixel and the noise for the signal of each pixel (noise removal target pixel) of each real line of each field (noise removal target field). Signals of the pixels at the same position in the screen as the noise removal target pixel in the three fields having pixels at the same position in the screen as the noise removal target field, that is, before and after the removal target field, that is, Noise is removed by filtering using a signal of four fields.

  When the progressive signal is input, the real line noise removing unit 21 receives the signal of the noise removal target pixel and the noise for the signal of each pixel (noise removal target pixel) of each scanning line of each frame (noise removal target frame). Noise is removed by performing filter processing using the signals of the pixels at the same position in the screen as the noise removal target pixels in the three frames located before and after the removal target frame, that is, the signals of four frames in total.

  The interpolated line noise removing means 22 receives each pixel (noise removal target pixel) of each interpolation line in fields adjacent to each interpolation field ("interpolation field", "noise removal target field") when an interlace signal is input. The signal of the noise removal target pixel and the noise removal target in three fields that are located before and after the noise removal target field and have pixels at the same position in the screen as the noise removal target field. Noise is removed by filtering using a signal of a pixel and a pixel at the same position in the screen, that is, a signal of four fields in total.

The interpolation line noise removal unit 22 includes a first interpolation line noise removal unit 221 and a second interpolation line noise removal unit 222.
The first interpolation line noise removal means 221 is an interpolation line in a noise removal target field ("first interpolation field" or "first noise removal target field") positioned before the interpolation field in time. The noise of the signal (“first interpolation line” or “first noise removal target line”) is removed.
The second interpolation line noise removing unit 222 performs interpolation lines (“second interpolation field” or “second noise removal target field”) in the noise removal target field (“second interpolation field”) positioned after the interpolation field in time. The noise of the signal of “second interpolation line” or “second noise removal target line”) is removed.

  The field (noise removal target field) and frame (noise removal target frame) that are subject to noise removal by the real line noise removal unit 21, the first interpolation line noise removal unit 221, and the second interpolation line noise removal unit 222 are: They are also called “center field” and “center frame”, respectively. The noise removal target pixel in the noise removal target field or frame is also referred to as a “target pixel”.

  The signal of the interpolation line from which noise has been removed is used in motion conversion processing for IP conversion and telecine interpolation, as will be described later.

  Here, when the input video signal is an interlace signal, data of one field is input in a period divided by two vertical synchronization signals that precede and follow, and when the input video signal is a progressive signal, One frame of data is input during a period delimited by two vertical synchronization signals.

  When an interlace signal is input, one field of data is written in each of the field memories 11 to 17, and each field memory performs a delay in field units. Since one frame is composed of two fields in the interlaced signal, the time interval (or delay) of two fields can be said to be the time interval (or delay) of one frame. In this case, fields having a relationship in which the pixels are at the same position in the screen appear at an interval of one frame (an interval of two fields).

  On the other hand, when a progressive signal is input, one frame of data is written in each of the field memories 11 to 17, and each field memory performs a delay in units of frames. However, as described above, in the present application, a period divided by two adjacent vertical synchronization signals is referred to as one field period, and a delay by each field memory is described as a delay of one field period. In the case of a progressive signal, all the frames have a relationship in which the pixels are at the same position in the screen. In other words, a frame having a relationship in which pixels are at the same position appears at an interval of one frame.

  In the first embodiment described in detail below, the interpolation field to be subjected to the IP conversion, that is, the scanning line interpolation when the interlace signal is input is the field of the two-field delay signal d2f from the second field memory 12, and the two fields It is assumed that a signal corresponding to the delay signal 2df is output as the output video signal Do0, while the output video signal Do0 when the progressive signal is input corresponds to the one-field delay signal d1f from the first field memory 11. It will be explained as being.

  Further, when the input signal is a progressive signal, it is not necessary to perform IP conversion, and it is only necessary to obtain a delay signal used for the non-cyclic noise removal processing by the non-cyclic noise removal processing means 20. Therefore, the output video signal Do0 when the progressive signal is input corresponds to the 1-field delay signal d1f, and the 1-field delay signal d1f and the signals of the three frames before and after it (the current frame signal and the two delay signals). When noise removal processing is performed using a total of four frames of signals, field delay processing is performed by the first to third field memories 11 to 13, and the subsequent fourth to seventh field memories 14 to 14 are performed. The field delay processing by 17 can be omitted.

  The switching signal generating means 18 sets the signal format of the input video signal, for example, by selection by an operation of a user or the like, for example, generates a switching signal pf indicating whether the input signal is an interlace signal or a progressive signal. Output. Instead of setting by selection by the user or the like, it is also possible to automatically determine whether or not the interlace is performed from the period of the synchronization signal of the input signal.

The acyclic noise removal processing means 20 receives the current field or frame signal Di0, the delay signals d1f to d7f, and the switching signal pf from the switching signal generation means 18.
The non-circular noise removal processing means 20 includes the signal of the pixel of interest in the field or frame (center field or frame) to be noise-removed among the signal Di0 of the current field or frame and the delayed signals d1f to d7f, Filter using the signal of the pixel located at the same position as the pixel of interest in the other three fields or frames where the pixel is located at the same position as the central field or frame, that is, the signal of 4 fields or 4 frames in total. To remove noise.

That is, when the interlace signal is input, the noise is removed from the actual line signal to generate the real line signal Rnr after noise removal, and the first interpolation line signal generated by removing the noise from the interpolation line signal. Isnr and the second interpolation line signal Itnr are generated, and when a progressive signal is input, the noise is removed from the signal of each line to generate a real line signal Rnr after noise removal.
The “center field” in noise removal at the time of interlace signal input includes the case where the real line signal Rnr after noise removal is generated, the case where the first interpolation line signal Isnr is generated, and the case where the second interpolation line signal Itnr is generated. It differs from the case of generating.

  The switching signal pf from the switching signal generating means 18 switches between a signal used for noise removal of the actual line signal and a signal used for noise removal for the progressive signal. When a progressive signal is input, since no interpolation of the scanning line is necessary, noise removal is not performed on the interpolation line signal, and the generation of the first interpolation line signal Isnr and the second interpolation line signal Itnr is not performed.

  Here, the signal (the signal of the pixel of interest in the center field) that is the target of the noise removal processing in the non-cyclic type noise removal processing means 20 is the interpolated field signal when the interlace signal is input, 2 field delay signal d2f, and the interpolation line of the interpolation field is an actual line signal (interpolation line signal) of a field temporally adjacent to the interpolation field, that is, 3 field delay signal d3f and 1 field delay signal d1f. is there. On the other hand, when a progressive signal is input, the 1-field delay signal d1f becomes a signal to be subjected to noise removal (a signal of a pixel of interest in the center frame).

  The acyclic noise removal processing means 20 is located in the center field or frame and the center field or frame before and after the signal of the pixel of interest in the center field or frame, and the center field or frame and pixel are within the screen. Noise is removed by filtering using signals of 3 fields or frames in the same position in total, 4 fields (4 fields every other field) or 4 frames (4 consecutive frames). Therefore, the calculation is based on the correlation between 4 fields or 4 frames (frame correlation), and the trailing or afterimage of the motion part can be within the range of 4 fields or 4 frames. Frames that reduce afterimages and span 4 fields (8 vertical periods) or 4 frames (4 vertical periods) Can be removed without noise component Seki, it is possible good noise reduction.

  The signal Di0 in the current field input to the acyclic noise removal processing unit 20 is input to the actual line noise removal unit 21 as the first NR processing signal r1.

  The switching means 201 to 203 in the acyclic noise removal processing means 20 shown in FIG. 2 receive the delay signals d1f to d4f and d6f, perform selection according to the switching signal pf, and select the selected signal as the first signal. The second to fourth NR processing signals r2 to r4 are output.

Different delay signals are selected for the noise removal process for the actual line signal and the noise removal process for the progressive signal.
Specifically, the switching means 201 receives a 1-field delay signal d1f and a 2-field delay signal d2f, selects a 2-field delay signal d2f when an interlace signal is input, and selects a 1-field delay signal d1f when a progressive signal is input. Then, the signal is supplied to the actual line noise removing unit 21 as the second NR processing signal r2.
The switching means 202 receives the 2 field delay signal d2f and the 4 field delay signal d4f, selects the 4 field delay signal d4f when the interlace signal is input, and selects the 2 field delay signal d2f when the progressive signal is input. The signal is supplied to the actual line noise removing unit 21 as the NR processing signal r3.
The switching means 203 inputs the 3 field delay signal d3f and the 6 field delay signal d6f, selects the 6 field delay signal d6f when the interlace signal is input, and selects the 3 field delay signal d3f when the progressive signal is input. The signal is supplied to the actual line noise removing unit 21 as the NR processing signal r4.

  The first to fourth NR processing signals r1 to r4 are used by the real line noise removing unit 21 for noise removal processing for the second NR processing signal r2 that is the center field or frame.

  The delayed signals d1f, d3f, d5f, and d7f are supplied to the interpolation line noise removal unit 22 as first to fourth NR processing signals i1 to i4 used for noise removal processing on the interpolation line signal.

  3 and 4 show the temporal relationship between the field Di0 and the delay signals d1f to d7f of the current field and the positional relationship in the vertical scanning direction of pixels used for noise removal and interpolation. FIG. 3 shows the positional relationship when an interlace signal is input, and FIG. 4 shows the positional relationship when a progressive signal is input. 3 and 4, the earlier field in time is shown on the left side, and the pixels of each field or frame are vertically scanned in order from top to bottom. The symbols r1 to r4 and i1 to i4 may represent pixel signals as described above, but may represent the pixels themselves as shown in FIGS.

  When the interlace signal shown in FIG. 3 is input, the signal in the interpolation field is the two-field delay signal d2f. Therefore, in the noise removal of the real line in the interpolation field, the signal is supplied as the second NR processing signal r2. The two-field delay signal d2f is the signal of the pixel of interest in the center field CFR. Then, the 2 field delay signal d2f, the current field signal Di0, the 4 field delay signal d4f, and the 6 field delay signal d6f (all of which are signals of pixels at the same position as the 2 field delay signal d2f), The first, third, and fourth NR processing signals r1, r3, and r4 are obtained, respectively. FIG. 3 shows that the first to fourth NR processing signals r1 to r4 are signals of pixels r1 to r4 (enclosed by a curve Gr) at the same position in the screen. .

In the noise removal of the interpolation line in the interpolation field, the signals of the adjacent fields (interpolation field) before and after the interpolation field, that is, the three-field delay signal d3f and the one-field delay signal d1f These are the second NR processing signal i2 and the first NR processing signal i1. The 5-field delay signal d5f and the 7-field delay signal d7f become the third NR processing signal i3 and the fourth NR processing signal i4, respectively.
Also in this case, as shown by being surrounded by the curve Gi in FIG. 3, the signals of the pixels at the same position in the screen are the first to fourth NR processing signals i1 to i4. The same NR processing signals i1 to i4 are used for noise removal for the first interpolation field signal (d3f) and for noise removal for the second interpolation field signal (d1f).

  When the progressive signal shown in FIG. 4 is input, since the output video signal Do0 corresponds to the delayed signal d1f from the first field memory 11, it is supplied as the second NR processing signal r2. The one-field delay signal d1f is a signal of the pixel of interest in the center frame. The current frame signal Di0, the two-field delay signal d2f, and the three-field delay signal d3f become the first, third, and fourth NR processing signals r1, r3, and r4, respectively. Also in this case, as shown by being surrounded by the curve Gr in FIG. 4, the signals of the pixels at the same position in the screen are the first to fourth NR processing signals r1 to r4.

  The interpolated line noise removing unit 22 in the non-cyclic type noise removing processing unit 20 receives the delay signals d1f, d3f, d5f, d7f output from the field memories 11, 13, and 15, 17 from the first to the fourth. Input as NR processing signals i1 to i4.

  The interpolation line noise removing unit 22 includes a first interpolation line noise removing unit 221 and a second interpolation line noise removing unit 222. When the interlace signal is input, the input first, second, third, and fourth are input. The NR processing signals i 1, i 2, i 3, i 4 are filtered to remove noise, and the first interpolation line signal Isnr and the second interpolation line signal Itnr are output.

  The first interpolation line signal Isnr and the second interpolation line signal Itnr are derived from an interpolation line signal (a signal in a temporally adjacent field of the interpolation field) used for generating the interpolation line signal at the time of three-dimensional IP conversion. This is a signal from which noise has been removed.

In FIG. 3, the first interpolation line signal Isnr has been subjected to a non-cyclic noise removal processing using the three-field delay signal d3f (the adjacent field temporally preceding the interpolation field) as the signal of the pixel of interest in the center field CFIs. The second interpolation line signal Itnr is a non-recursive noise removal process using the 1-field delay signal d1f (the adjacent field after the interpolation field in time) as the signal of the pixel of interest in the center field CFIt. It is a signal after having performed.
In the IP conversion motion adaptation processing, the first interpolation line signal Isnr is used as the interpolated line signal after noise removal. In the telecine interpolation, the first interpolation line signal Isnr and the second interpolation line signal Itnr are used as the interpolation line signals after noise removal. used.

  Detailed configurations of the real line noise removing unit 21 and the interpolation line noise removing unit 22 in the non-cyclic type noise removing processing unit 20 shown in FIG. 2 will be described later.

  In FIG. 1, to the three-dimensional IP conversion processing means 65, the signal Di0 of the current field also used by the non-cyclic type noise removal processing means 20 and the delayed signals d1f to d7f from the first to seventh field memories 11-17. Are entered. The three-dimensional IP conversion processing means 65 is further supplied with the noise-removed real line signal Rnr, the first interpolation line signal Isnr, and the second interpolation line signal Itnr from the acyclic noise removal processing means 20. The

  In the three-dimensional IP conversion processing means 65, when the input signal is an interlace signal, a scanning line interpolation signal is generated by intra-field interpolation processing and inter-field interpolation processing, and the interlace signal is converted into a progressive signal by three-dimensional scanning line conversion. A progressive signal Ips from which noise has been removed is output. At this time, the first interpolation line signal Isnr and the second interpolation line signal Itnr output from the acyclic noise removal processing unit 20 are used for the inter-field interpolation processing in the three-dimensional IP conversion processing unit 65. .

  Conversion to a progressive signal Ips using the interpolated signal Im obtained from the post-noise-removed real line signal Rnr, the first interpolating line signal Isnr, and the second interpolating line signal Itnr from the non-cyclic type noise removing means 20. Therefore, tailing and afterimages can be reduced, and a progressive signal Ips from which noise is well removed can be obtained.

Specifically, in FIG. 1, the two-field delay signal d2f from the second field memory 12 is input to the intra-field interpolation processing means 30 in the three-dimensional IP conversion processing means 65. At this time, the interpolation field to be subjected to scanning line interpolation by three-dimensional IP conversion is a two-field delay signal d2f, and the progressive signal output after IP conversion is a signal delayed by two fields.
The intra-field interpolation processing means 30 performs arithmetic processing such as filter processing on the pixels on the scanning lines adjacent in the vertical direction in the interpolation field to generate and output the interpolation signal Imv.

A current field signal Di0, a one-field delay signal d1f, a two-field delay signal d2f, and a four-field delay signal d4f are input to the interpolation motion detection means 66 in the three-dimensional IP conversion processing means 65.
The interpolation motion detection means 66 includes a motion detection means 40 and a telecine detection means 50, detects the motion of the signal from the difference between frames or fields of the video signal, and based on the detection result, an interpolation switching signal (motion detection signal). mdt and telecine detection signal tcif) are detected and output.

  The motion detector 40 receives the current field signal Di0, the 2-field delay signal d2f, and the 4-field delay signal d4f. The motion detection means 40 detects difference information between the current field signal Di0 and the two-field delay signal d2f between two fields (one frame) and between the current field signal Di0 and the four-field delay signal d4f between four fields (2 Difference information between frames) is obtained, based on the obtained difference information, the motion of the video signal between 1 frame and 2 frames is detected, and a motion detection signal mdt indicating the degree of motion of the video signal is output.

  The motion detection means 40, for example, divides the motion into a plurality of (for example, nine) grades or stages according to the degree, and outputs a motion detection signal mdt having a value indicating to which stage the motion belongs. For example, when the degree of movement is the maximum, that is, when the video signal is “completely moving”, the first predetermined value, for example, mdt = 8, and the degree of movement is the minimum, That is, in the case of “completely a still image”, a second predetermined value, for example, mdt = 0, for example, is a signal indicating the degree of motion between mdt = 0-8.

  Note that the value of the motion detection signal mdt is not limited to this, and may be a value obtained by another method or a binary value of motion or still image. Further, it is described that difference information indicating a difference (frame difference) between the signal Di0 of the current field and the 2-field delay signal d2f and the 4-field delay signal d4f is obtained, and the motion of the video signal between 1 frame and 2 frames is detected. However, it is also possible to detect the motion based only on the difference information of the video signal for one frame from the difference information of the signal Di0 of the current field or frame and the two-field delay signal d2f.

  Alternatively, the motion may be detected by a motion vector between the previous block (a block forming a part of the field) and a motion detection signal mdt indicating the degree of motion of the video signal can be obtained in some form. That's fine.

  Furthermore, the motion detection means 40 is configured to perform motion detection using the current field signal Di0, the 2-field delay signal d2f, and the 4-field delay signal d4f. However, the position of the interpolation line signal in the interpolation field is considered. Thus, the 1-field delay signal d1f, the 3-field delay signal d3f, and the 5-field delay signal d5f are used to obtain difference information between fields and detect the motion of the video signal between 1 frame and 2 frames. The motion detection may be performed with the same configuration only by changing the field of the signal input to the motion detection means 40.

  Also, the telecine detection means 50 receives the current field signal Di0, the two-field delay signal d2f, and the one-field delay signal d1f. The telecine detection means 50 detects the motion between fields and frames from the difference between frames and between fields, and detects whether or not the input video signal is a telecine video signal. That is, the correlation (frame difference information) between the current field signal Di0 and the two-field delay signal d2f with the field one frame before based on the difference information between the fields, and the field between the current field signal Di0 and the one-field delay signal d1f. The correlation (field difference information) with the previous field by the difference information is obtained. Then, from this frame difference information and field difference information, it is detected whether or not the input video signal is a telecine video signal that satisfies the field sequence repetition condition in the 2-3 pulldown signal or the 2-2 pulldown signal. Based on the detection of the telecine video signal for performing the telecine interpolation, a telecine detection signal tcif indicating the interpolation phase indicating the phase of the sequence (that is, the direction of the converted field from the same frame) is generated and output.

  As described above, the interpolation motion detection unit 66 obtains the motion detection signal mdt output from the motion detection unit 40 and the telecine detection signal tcif output from the telecine detection unit 50, and these are obtained as an interpolation switching signal (motion detection). Signal mdt and telecine detection signal tcif) and output to interpolation signal generation means 60.

The interpolation signal generation means 60 includes a first interpolation line signal Isnr and a second interpolation line signal Itnr from the acyclic noise removal processing means 20, and an intra-field interpolation signal Imv from the intra-field interpolation processing means 30. Then, an interpolation switching signal (motion detection signal mdt and telecine detection signal tcif) from the interpolation motion detection means 66 is input.
The interpolation signal generation means 60 performs motion adaptive IP conversion on the first interpolation line signal Isnr, the second interpolation line signal Itnr, and the intra-field interpolation signal Imv based on the motion detection signal mdt and also performs telecine. When the detection signal tcif indicates a telecine video signal, telecine interpolation is performed and an interpolation line signal Im is output.

Specifically, in the interpolation signal generation means 60, the process of generating the interpolation line signal is performed on the 2-field delay signal d2f, and the motion adaptation process is performed on the normal video signal. In the case of a normal video signal, that is, not a telecine video signal, motion adaptation processing is performed based on the motion detection signal mdt from the motion detection means 40. That is, when the motion detection signal mdt is “0” indicating that it is a complete still image, the first interpolation line, which is a signal obtained by removing noise from the previous three-field delayed signal d3f, is used. Inter-field interpolation is performed using the value of the signal Isnr as the value of the interpolation line signal to obtain an interpolation signal Im.
When the motion detection signal mdt is “8”, indicating that the motion is complete, the intra-field interpolation signal Imv from the intra-field interpolation processing means 30 is set as the interpolation signal Im.
When the motion detection signal mdt is one of “1” to “7” and is between a complete still image (mdt = 0) and a complete motion (mdt = 8), the value of the motion detection signal mdt indicates Depending on the degree of motion, the first interpolation line signal Isnr and the intra-field interpolation signal Imv are mixed to generate an interpolation signal Im.

  As described above, the motion detection signal mdt generates the interpolation signal Im using the first interpolation line signal Isnr in the still image portion. In the motion portion, the intra-field interpolation signal Imv obtained by filtering in the field is used. An interpolation signal Im is generated. Therefore, there is no tailing or the like in the moving part, and the noise removal effect by the filter can be obtained. In this way, the interpolation signal Im generated by the signal from which noise is removed by the non-cyclic type noise removal processing means 20 and the signal by intra-field interpolation by the motion adaptation processing, which reduces tailing and afterimage, and is well noise-removed. Can be obtained.

  Further, the interpolation signal generation means 60 performs a telecine interpolation process on the telecine video signal. That is, when the telecine detection signal tcif output from the telecine detection means 50 indicates that it is a telecine video signal, it corresponds to the interpolation phase of 2-3 or 2-2 pulldown based on the telecine detection signal tcif. A field is selected from the first interpolation line signal Isnr (signal after noise removal of the three-field delay signal d3f) or the second interpolation line signal Itnr (signal after noise removal of the one-field delay signal d1f), and telecine video It is generated as an interpolation signal Im in the case of a signal.

  Also in the telecine interpolation process, the first interpolation line signal Isnr or the second interpolation line signal from which the non-cyclic noise is removed by the non-cyclic noise removal processing means 20 is the same as the still image portion in the motion adaptation process. Itnr becomes an interpolation signal. Therefore, since the interpolation signal Im is generated by the signal from which noise has been removed by the non-cyclic type noise removal processing means 20, tailing and afterimage are reduced and noise is removed satisfactorily.

The time axis conversion means 61 receives the interpolation signal Im from the interpolation signal generation means 60 and the noise-removed real line signal Rnr from the acyclic noise removal processing means 20 and converts the interpolation signal Im into the actual interpolation field. The line signal Rnr is fitted between corresponding scanning lines, double-speed converted, converted to a progressive signal Ips, and output.
Similarly to the above, since the post-noise removal actual line signal Rnr after the noise removal from the non-cyclic type noise removal processing unit 20 and the interpolation signal Im from the interpolation signal generation unit 60 are time-axis converted to the progressive signal Ips, It is possible to obtain a progressive signal from which tailing and afterimages are reduced and noise is removed satisfactorily.

  Then, to the switching means 62, the IP converted progressive signal Ips from the time axis converting means 61, the noise-removed real line signal Rnr from the non-cyclic type noise removing processing means 20, and the switching signal generating means 18 are switched. The signal pf is input. When the switching means 62 indicates that the switching signal pf is an interlace signal input, the IP-converted progressive signal Ips from the time axis conversion means 61 is selected to indicate that the switching signal pf is a progressive signal input. In this case, the real line signal Rnr after noise removal from the non-cyclic type noise removal processing means 20 is selected and output as an output video signal Do0 which is a progressive signal.

  The real line signal Rnr after noise removal, which is the output from the non-cyclic type noise removal processing means 20, is the same as the real line signal when the interlace signal is input with respect to the one-field delay signal d1f of the progressive signal when the progressive signal is input. A non-cyclic noise removal process is performed in the same manner. That is, the filtering process is performed using the signals of the pixels having the same position in the screen for a total of four frames including the one-field delay signal d1f.

  As described above, the delay signal used in the noise removal process and the delay signal used in the three-dimensional IP conversion when the interlace signal is input are obtained from the common field memory both when the interlace signal is input and when the progressive signal is input. Therefore, the number of field memories as delay means can be suppressed, and the adverse effect of tailing and afterimage can be reduced from the switching means 62, and a progressive signal with good noise removed can be obtained. be able to.

  Next, the real line noise removing unit 21 and the interpolation line noise removing unit 22 in the non-cyclic type noise removing processing unit 20 shown in FIG. 2 will be described in detail with reference to FIGS.

  First, in FIG. 2, the actual line noise removing means 21 uses the second field delay signal d2f when the interlace signal is input or the second field signal NR2 configured by the one field delay signal d1f when the progressive signal is input as the central field. Alternatively, the signal of the target pixel of the frame CFR (see FIGS. 3 and 4) is used to remove noise by performing filtering using the input first to fourth NR processing signals r1 to r4 and The line signal Rnr is output.

  The interpolation line noise removal unit 22 performs filtering using the first to fourth NR processing signals i1 to i4 input by the first interpolation line noise removal unit 221 and the second interpolation line noise removal unit 222. The first interpolation line noise removal unit 221 outputs the first interpolation line signal Isnr, and the second interpolation line noise removal unit 222 outputs the second interpolation line signal Itnr. Output.

  In FIG. 3, a first interpolation line signal Isnr is a signal obtained by removing noise from a three-field delay signal d3f (second NR processing signal i2) that is an adjacent field temporally preceding the interpolation field. The first interpolation line noise removing unit 221 uses the three-field delay signal d3f (second NR processing signal i2) as the signal of the pixel of interest in the center field CFIs.

The second interpolation line signal Itnr is a signal obtained by removing noise from the one-field delayed signal d1f (first NR processing signal i1) that is an adjacent field after the interpolation field. In the interpolation line noise removing means 222, the 1-field delay signal d1f (first NR processing signal i1) is used as the signal of the pixel of interest in the center field CFIt.
Hereinafter, the “signal of the pixel of interest in the center field or frame” may be simply referred to as “the signal of the center field or frame”.

  FIG. 5 shows an acyclic noise removal processing means 70 that can be used as the actual line noise removal means 21 or the first interpolation line noise removal means 221 in FIG. When the acyclic noise removal processing means 70 shown in FIG. 5 is used as the real line noise removal means 21, the second NR processing signal among the four input NR processing signals r1 to r4. When r2 is used as the signal of the center field or frame and acyclic noise removal processing is performed and used as the first interpolation line noise removal means 221, the second of the four input NR processing signals i1 to i4. The non-cyclic type noise removal processing is performed using the NR processing signal i2 as the center field signal.

The acyclic noise removing unit 70 shown in FIG. 5 includes a noise removing filter unit 700, an NR motion detecting unit 720, and a filter coefficient generating unit 710.
The noise removal filter unit 700 performs filter processing using the four NR processing signals r1 to r4 or i1 to i4, and outputs a noise-removed signal Rnr or Isnr.
The NR motion detection means 720 detects the motion of the signal based on the four input NR processing signals r1 to r4 or i1 to i4 with respect to the signal of the center field or frame, and indicates a signal (NR) Motion degree signal) mds is output.
The filter coefficient generation unit 710 generates and outputs filter coefficients kn1 to kn4 of the noise removal filter based on the NR motion degree signal mds generated by the NR motion detection unit 720.

  The noise removal filter means 700 adds coefficient multiplication means 701, 702, 703, 704 for multiplying the respective NR processing signals by the filter coefficients kn1 to kn4 from the filter coefficient generation means 710, and the signals multiplied by the coefficients. And adding means 705 for outputting the filtered signal after noise removal.

The NR motion detection unit 720 includes a frame difference detection unit 730, an edge adjustment unit 740, and a conversion unit 750.
The frame difference detection unit 730 includes difference detection units 731, 732, and 733 that detect a frame difference, and a combination unit 734. The frame difference detection unit 730 performs calculation to obtain a frame difference and detects a motion difference signal that represents motion.
The input of the frame difference detection means 730 is a “field signal” when an interlace signal is input, and is a “frame signal” when a progressive signal is input. The detection means 730 calculates a difference between fields in which the pixels are at the same position (a field every two fields when an interlace signal is input), and is referred to as “determining a frame difference”.
The edge adjustment unit 740 detects an edge part in the signal of the center field or frame (CFR or CFIs), adjusts the motion difference signal of the edge part, and averages the target pixel and the surrounding pixels. To generate a motion difference signal Dfe.
The conversion unit 750 converts the motion difference signal Dfe from the edge adjustment unit 740 into a motion degree signal mds.

  To the non-cyclic type noise removal processing means 70, the first to fourth NR processing signals r1 to r4 shown in FIGS. 3 and 4 or the first to fourth NR processing signals i1 shown in FIG. To i4 are input via the input terminals 751 to 754, and the non-cyclic noise removal processing is performed by using the second NR processing signal r2 or i2 as a central field or frame (CFR or CFIs) signal.

  In the following, the case where the acyclic noise removal processing means 70 is mainly used as the actual line noise removal means 21 will be described in more detail. The operation when the acyclic noise removal processing means 70 is used as the first interpolation line noise removal means 221 is the same, but the signals i1, i2, i3, i4 are replaced with the signals r1, r2, r3, r4. It differs in that it is input.

  The non-cyclic type noise removing unit 70 detects the motion of the signal between four fields or four frames based on the NR processing signals r1 to r4 input from the NR motion detecting unit 720, and is obtained as a result of the detection. Based on the NR motion degree signal mds, the filter coefficient generation means 710 generates filter coefficients kn1 = kn4, and controls the noise removal strength to perform acyclic noise removal processing.

  Here, the processing performed by the non-cyclic type noise removing means 70 is filter processing using frame correlation between 4 fields (every other field) or 4 frames (consecutive), and the tail of the motion part is processed. It is possible to reduce the trailing and afterimages within 4 fields or 4 frames within the range of 4 fields or 4 frames, and noise without frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods). Components can be removed. However, since it is a filtering process, it is conceivable that tailing and afterimages appear as edges and blurring or degradation of the contour portion in the contour portion of the motion. Therefore, in the acyclic noise removing unit 70, the NR motion detecting unit 720 detects a signal motion between four fields or four frames and generates a filter coefficient based on the NR motion degree signal mds obtained as a result of the detection. By performing the filtering process using kn1 to kn4, the strength of noise removal is controlled, thereby preventing the blur and deterioration of the edge and contour of the moving part.

The NR motion detection means 720 receives the first to fourth NR processing signals r1 to r4.
In the NR motion detection means 720, the difference (frame) between the second NR processing signal r2 as the center field or frame signal and the first, third, and fourth NR processing signals r1, r3, r4. Difference) is obtained, the obtained frame differences are combined to detect the motion in the video signal, and the NR motion degree signal mds obtained as a result of motion detection is output.

The NR motion detection means 720 divides the motion into a plurality of grades or stages of (MD1 + 1), for example, according to the degree, and outputs an NR motion degree signal mds indicating to which stage the motion belongs. For example, when the motion is the maximum level, that is, when the video signal is “complete motion”, the first predetermined value, for example, mds = MD1, and when the motion level is the minimum level, that is, “complete still image” If this is the case (this includes the case where the difference is due to a noise component), the second predetermined value, for example, mds = 0 is set, and the value of mds increases as the degree of motion increases. . MD1 is defined as “4”, for example.
The NR motion degree signal mds generated by the NR motion detector 720 is sent to the filter coefficient generator 710.

  The first to fourth NR processing signals r1 to r4 are input to the frame difference detecting unit 730 in the NR motion detecting unit 720, and the second NR processing signal r2 and the first, third, and fourth NR processing signals r2 to r4 are input. The difference between the NR processing signals r1, r3, r4 (three frame differences) is calculated, and by combining these frame differences, a frame difference indicating the degree of signal movement is detected and motion is detected. Output the difference signal. The frame difference includes a “motion” component in the video signal. When the frame difference is 0, the frame difference is a completely stationary portion, and when there is a motion or noise component, the value increases.

  In FIG. 5, the difference detection means 731, 732, 733 are input with the second NR processing signal r2 and the first, third, and fourth NR processing signals r1, r3, r4, respectively. Is calculated. That is, the second NR processing signal r2 and the first NR processing signal r1 are input to the difference detection means 731 to calculate the frame difference and output the frame difference dif1. Similarly, the second NR processing signal r 2 and the third NR processing signal r 3 are input to the difference detection means 732, the frame difference is calculated and the frame difference dif 2 is output, and the difference detection means 733 is output. Receives the second frame r2 and the fourth NR processing signal r4, calculates the frame difference, and outputs the frame difference dif3. Three frame differences dif1, dif2, and dif3 are obtained from the difference detection means 731, 732, and 733.

  When the non-cyclic type noise removal processing means 70 is used as the real line noise removal means 21, that is, the second NR composed of the two-field delay signal 2df of the interlace signal or the one-field delay signal 1df of the progressive signal. In the noise removal of the processing signal r2, as shown in FIGS. 3 and 4, the frame difference dif1 is equal to the second NR processing signal r2 and the first NR after one frame (after two or one vertical period). This is a difference between one frame of the processing signal r1, and the frame difference dif2 is between one frame of the second NR processing signal r2 and the third NR processing signal r3 one frame before (two or one vertical period before). The frame difference dif3 is the second NR processing signal r2 and the fourth NR processing signal r two frames before (four or two vertical periods before). The difference between the two frames.

  When the acyclic noise removal processing means 70 is used as the first interpolation line noise removal means 221, that is, for a three-field delay signal d3f that is a signal of a field adjacent to the interpolation field at the time of interlace signal input. In the noise removal performed, as shown in FIG. 3, the frame difference dif1 is a difference between two fields (one frame) of the second NR processing signal i2 and the first NR processing signal i1 after two fields. Yes, the frame difference dif2 is the difference between two fields (one frame) of the second NR processing signal i2 and the third NR processing signal i3 two fields before, and the frame difference dif3 is the second NR This is the difference between the four fields (two frames) of the processing signal i2 and the fourth NR processing signal i4 four fields before.

  Since the difference detection means 731, 732, and 733 can have the same configuration, the difference detection means 731 will be described below. For example, as shown in FIG. 6, the difference detection unit 731 includes a difference calculation unit 735, an absolute value calculation unit 736, and a motion sensitivity conversion unit 737.

  The difference calculation means 735 receives the two NR processing signals r2 and r1, and calculates the difference between the two input NR processing signals. The signal r1 is subtracted from the signal r2 to obtain the frame difference Dn. The frame difference Dn includes a “motion” component in the video signal. When the frame difference Dn is 0, the frame difference Dn is a completely stationary part. If there is a motion or noise component, its value is growing.

  The absolute value calculation means 736 receives the frame difference Dn from the difference calculation means 735. The absolute value calculating means 736 calculates the absolute value of the frame difference Dn and outputs the absolute value (difference absolute value) Dabs of the frame difference.

  The difference absolute value Dabs from the absolute value calculation means 735 is input to the motion sensitivity conversion means 737. In the motion sensitivity conversion means 737, for example, the difference absolute value Dabs is multiplied by a predetermined sensitivity magnification, the offset value is subtracted, and the value is limited to a predetermined value (for example, 0 to dM). Is subjected to nonlinear conversion, sensitivity conversion of the frame difference is performed, and output as the frame difference dif1 subjected to sensitivity conversion.

  FIG. 7 shows the relationship between an absolute difference value Dabs (input shown on the horizontal axis), which is an example of input / output characteristics in the motion sensitivity conversion means 737, and an output frame difference dif1 (output shown on the vertical axis). Yes. In the illustrated example, when the difference absolute value Dabs is equal to or greater than Tm, the output frame difference dif1 is a constant value dM, and when the difference absolute value Dabs is equal to or less than the offset value Tof, it is treated as a minute noise component and output. The frame difference dif1 is 0.

  In the range where the difference absolute value Dabs is from Tof to Tm, the frame difference dif1 gradually increases from 0 to dM as the difference absolute value Dabs increases, and the value indicates the degree of motion. By increasing the sensitivity magnification, the frame difference absolute value Dabs is adjusted to represent a larger movement, and when the offset value is increased, the difference value up to that value becomes a minute noise component (or a stationary part). ) Will be detected. Therefore, the sensitivity of motion detection is adjusted by the sensitivity magnification and the offset value.

  A low-pass filter (LPF) or the like for extracting a low frequency component is inserted on the output side of the difference calculation means 735 or the output side of the absolute value calculation means 736, thereby filtering the frame difference Dn or the difference absolute value Dabs. The result of the filtering process may be supplied to a circuit (absolute value calculation means 736 or motion sensitivity conversion means 737) in the next stage.

  The configuration and operation of the difference detection means 732 and 733 are the same as the difference detection means 731. However, the difference detection means 732 receives the signals r2 and r3 instead of the signals r2 and r1 and outputs the frame difference dif2, and the difference detection means 733 receives the signals r2 and r4 instead of the signals r2 and r1. The frame difference dif3 is output.

  By the operation as described above, three frame differences dif1, dif2, and dif3 between the frames are obtained from the difference detecting means 731, 732, and 733.

The frame differences dif1, dif2, and dif3 output from the difference detection units 731, 732, and 733 are input to the combining unit 734 shown in FIG. The synthesizing unit 734 synthesizes the frame differences dif1, dif2, and dif3 to obtain a frame difference (motion difference) indicating the degree of signal motion, and outputs a motion difference signal Daf.
As a synthesis method, for example, the maximum values of the frame differences dif1, dif2, and dif3 are obtained, and the obtained maximum values are detected as motion differences between the four input frames, and set as the motion difference signal Daf. Note that the synthesis method is not limited to obtaining the maximum value, and an average value may be obtained.
The motion difference signal Daf output from the synthesizing unit 734 is sent to the edge adjustment unit 740.

  The edge adjustment unit 740 receives the motion difference signal Daf from the frame difference detection unit 730 and the second NR processing signal r2. The edge adjustment means 740 detects an edge portion (or contour portion) in the second NR processing signal r2 (by determining whether or not the pixel of interest forms an edge), and determines that the edge is formed. The motion difference signal Daf at the selected pixel is adjusted so as to represent a larger motion, and the pixel that has not been determined to constitute an edge is isolated by performing averaging processing with surrounding pixels. The point is removed, and the adjusted motion difference signal Dfe is output.

FIG. 8 is a block diagram showing a configuration example of the edge adjustment unit 740. In FIG. 8, the edge adjustment unit 740 includes an edge determination unit 744, an averaging unit 742, an adjustment unit 743, and a selection unit. 745.
The edge determination means 744 determines whether or not the target pixel constitutes an edge based on the second NR processing signal r2.
The averaging means 742 obtains the average value of the motion difference signal Daf with the peripheral pixels centered on the target pixel.
The adjustment unit 743 adjusts the motion difference signal Daf in the pixel determined to constitute the edge according to the determination result from the edge determination unit 744 so as to represent a larger motion.

  Here, the edge and outline of a signal are often the boundary between a pixel determined to be motion and a pixel determined to be a still image, and erroneous detection of motion is likely to occur. In some cases, the detection omission of flickering flickers at the switching portion of the noise removal strength, resulting in degradation of image quality. Therefore, the edge adjustment unit 740 determines the edge of the signal and the boundary to be the outline, and adjusts the value of the motion difference signal Daf in the pixel determined to constitute the edge to a value indicating a larger movement (for example, The output Dcn is set to a value larger than the motion difference signal Daf), and the isolated points are removed by averaging the pixels that have not been determined to constitute the edge with the surrounding pixels. As a result, flickering due to the detection omission of the motion difference signal can be prevented, and blurring and deterioration of the edges and contours of the motion part can be prevented.

  The edge determination means 744 receives the second NR processing signal r2. Based on the input second NR processing signal r2, the edge determination means 744 extracts the high frequency components in the horizontal and vertical directions of the target pixel in the field, and the target pixel determines the edge from the extracted component value. It is determined whether to configure (edge determination). Then, an edge determination result edflg and an adjustment value mxed for the motion difference signal are obtained and output.

Specifically, the horizontal and vertical high frequency components in the edge determination means 744 are extracted by a bandpass filter or the like, and the horizontal high frequency components are compared with the vertical high frequency components. By selecting the maximum value, horizontal and vertical high frequency components in each pixel are extracted.
Here, the high frequency component in the horizontal direction is centered on the target pixel position, for example,
High frequency components between adjacent pixels in the horizontal direction (for example, a value obtained by absoluteizing the bandpass filter output), or high between adjacent pixels in the horizontal direction (adjacent adjacent pixels) A band frequency component (for example, a value obtained by converting the bandpass filter output into an absolute value) is extracted.
On the other hand, the high-frequency component in the vertical direction is obtained by the line delay process or the like between the target pixel (for example, r2 in FIG. 3 or 4) and pixels on other lines aligned in the vertical direction with respect to the target pixel. A signal (for example, r2_0 in FIG. 3 or r2_2 in FIG. 4) is obtained, and a high-frequency component (for example, a value obtained by converting the bandpass filter output into an absolute value) between lines is extracted. Note that a high frequency component (a value obtained by converting the band-pass filter output into an absolute value) between both the pixel located above and the pixel located below the pixel position of interest as the center may be extracted.

The edge determination means 744 further compares the horizontal and vertical high-frequency component values extracted as described above with a predetermined threshold value, detects the pixels constituting the edge, and results in the binarized value as a result. Output as edge determination result edflg. In this comparison, for example, if the horizontal and vertical high-frequency components are larger than a predetermined threshold, it is determined that the edge is formed (determined that the target pixel forms an edge), and is smaller than other predetermined thresholds. Is determined not to constitute an edge.
Further, the edge determination means 744 obtains an adjustment value mxed for the motion difference signal by nonlinearly transforming the values of the extracted horizontal and vertical high frequency components. For example, when the value of the high frequency component is larger than a predetermined threshold, the adjustment value mxed can be set to a predetermined multiplication value for the motion difference, and the motion difference at the edge portion can be adjusted to a value representing a larger motion. Like that.
Note that the adjustment value mxed may be a multiplication value for a motion difference that changes in proportion to the value of the high frequency component, or may be an addition value. It is only necessary to output an adjustment value that can adjust the difference to a value representing a larger movement.

  The adjustment unit 743 receives the motion difference signal Daf output from the frame difference detection unit 730 and the adjustment value mxed for the motion difference signal output from the edge determination unit 744. The adjustment unit 743 performs adjustment based on the adjustment value mxed with respect to the motion difference signal Daf, and outputs the adjusted motion difference signal Dcnt. For example, when the adjustment value mxed sets a predetermined multiplication value for the motion difference, the adjustment unit 743 multiplies the motion difference signal Daf by a value based on the adjustment value mxed. As a result, the motion difference signal Daf can be adjusted to indicate a larger motion. On the other hand, when the adjustment value mxed is an addition value for the motion difference, the adjustment unit 743 adds the value based on the adjustment value mxed to the motion difference signal Daf.

  The motion difference signal Daf from the frame difference detection means 730 is input to the averaging means 742. The averaging means 742 calculates an average value Davg of the difference signal within a predetermined peripheral pixel range for the motion difference signal Daf at the target pixel, averages the difference signal, and removes isolated points. The motion difference signal Daf at the pixel is corrected. That is, in order to eliminate the variation with the values of the pixels around the motion difference signal Daf, the values are unified by averaging, the detection of motion detection is prevented, and the averaged motion difference signal (average motion difference signal) ) Output Davg.

  Here, the range of peripheral pixels for which the averaging means 742 obtains the average value is, for example, a range of 5 pixels in the horizontal direction centering on the target pixel. Note that the peripheral pixel range is not limited to this, and may be set in a two-dimensional range, for example, a total of 15 pixels including 3 lines in the vertical direction and 5 pixels in the horizontal direction centering on the target pixel. .

  The edge difference adjusted motion difference signal Dcnt output from the adjustment unit 743 and the average motion difference signal Davg output from the averaging unit 742 are sent to the selection unit 745.

  The selection unit 745 receives the motion difference signal Dcnt after the edge adjustment, the average motion difference signal Davg, and the determination result edflg from the edge determination unit 744. In the selection unit 745, when the determination result edflg indicates that the target pixel constitutes an edge, the motion difference signal Dcnt after the edge adjustment from the adjustment unit 743 is selected for the pixel, If the determination result edflg indicates that the pixel of interest does not constitute an edge, the average motion difference signal Davg from the averaging means 742 is selected for the pixel, and the edge portion is adjusted and averaged. The motion difference signal Dfe from which the isolated points have been removed is output.

  As described above, the edge adjustment unit 740 performs adjustment of the motion difference signal of the edge part and obtains the motion difference signal Dfe subjected to the averaging process, and prevents flicker due to the detection omission of the motion difference signal, Blur and deterioration of edges and contours of moving parts can be prevented.

  The motion difference signal Dfe output from the edge adjustment unit 740 is input to the conversion unit 750 in the NR motion detection unit 720 in FIG. 5, and the NR is calculated from the motion difference signal Dfe subjected to the edge adjustment and averaging. A movement degree signal mds is generated and output.

  In the conversion means 750, the input motion difference signal Dfe is subjected to a conversion process by limiting the amplitude, and converted into, for example, a five-step NR motion degree signal mds from 0 to 4 indicating the degree of motion. That is, if the input motion difference signal Dfe is 0 or less than a predetermined value (a value close to 0), mds = 0, and if the value is greater than or equal to the predetermined value, conversion is performed so that mds = 4. NR movement degree signal mds indicating the degree of movement with a value between 0 and 4 is generated.

  This NR motion degree signal mds has a large motion difference signal value, and when the frame difference is detected as complete motion, mds = 4 (complete motion pixel), and the motion difference signal value is small (close to 0 or 0). Or less), when it is detected that the image is a complete still image (this includes a case where there is a difference but it is due to a noise component), mds = 0 (complete still pixel), and 0 to 4 A value between the values indicates the degree of movement.

  The setting of the movement degree signal mds has been described as five stages from 0 to 4 indicating the degree of movement. However, the setting is not limited to this, and any value indicating the degree of movement may be set in five stages or more. Even lower than that is possible. In addition, although it has been described that the conversion process based on the amplitude limit is performed, the motion difference signal Dfe is subjected to conversion by multiplying a predetermined value, addition or subtraction of a predetermined value, and non-linear conversion to limit the amplitude, and the degree of NR motion The signal mds may be obtained.

  As described above, the NR motion detection unit 720 performs signal motion detection by generating the NR motion degree signal mds in the conversion unit 750, and the NR motion degree signal mds is sent to the filter coefficient generation unit 710.

  In the filter coefficient generation means 710, filter coefficients kn1 to kn4 corresponding to the NR motion degree signal mds are obtained. The filter coefficients kn1 to kn4 are used in the acyclic noise removal filter processing in the noise removal filter means 700.

  The filter coefficients kn1 to kn4 generated by the filter coefficient generation means 710 are used to set the coefficients of a filter that performs noise removal, and are each determined to be 0 or more and 1 or less, and the sum thereof is 1. The strength (effect) of noise removal is determined by assigning filter coefficients kn1 to kn4 to the NR processing signal.

  The filter coefficients kn1 to kn4 for the first, second, third, and fourth NR processing signals r1, r2, r3, and r4 are determined using the second NR processing signal r2 as a central field or frame signal. It is done. When the filter coefficients kn1 to kn4 have the same value 1/4 (= 0.25), the noise component having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, if the value of the coefficient kn2 for the signal r2 of the center field or frame CFR is large, kn2 = 1, and the other filter coefficients kn1, kn3, kn4 are zero, Is multiplied by 0, the noise cyclic value becomes zero, and the signal is output as it is, and noise removal is not performed.

  The filter coefficient generation means 710 controls the values of the filter coefficients kn1 to kn4 in accordance with the NR motion degree signal mds from the NR motion detection means 720 in order to prevent blurring and deterioration of edges and contours of moving parts in the video signal. By doing so, the noise removal strength of the moving part is controlled. The filter coefficient generating means 710 is configured by, for example, a ROM (Read Only Memory) that inputs the NR motion degree signal mds as an address, and generates a filter coefficient whose value is determined according to the value of the NR motion degree signal mds). .

  That is, when the NR motion degree signal mds is MD1 (for example, 4) and indicates complete motion, kn2 = 1, kn1 = kn3 = kn4 = 0 (hereinafter, (kn1, kn2, kn3, kn4) = (0, 1, 0, 0)), and when the degree of motion is the smallest and mds = 0 detected as a complete still image (including the case where the difference is due to noise), the filter The coefficients kn1 to kn4 are set to the same value. That is, (kn1, kn2, kn3, kn4) = (0.25, 0.25, 0.25, 0.25). When the NR motion degree signal mds has an intermediate value (mds = 1, 2, 3), for example, (kn1, kn2, kn3, kn4) = (2/8, 3/8, 2/3, 1) / 3), (1/3, 1/3, 1/3, 0), (0.5, 0.5, 0, 0), etc., to control the strength of noise removal. .

  As the degree of motion (value of the NR motion degree signal mds) increases, the value of the coefficient kn2 for the signal r2 of the center field or frame CFR increases, and the values of the other coefficients decrease (for example, (kn1, kn2, kn3). , Kn4) = (2/8, 3/8, 2/3, 1/3)), or the number of fields or frames used for filtering (the number of NR processing signals with coefficients other than 0). 3 (for example, (kn1, kn2, kn3, kn4) = (1/3, 1/3, 1/3, 0)) or 2 (for example, (kn1, kn2, kn3, kn4) = (0.5, 0.5, 0, 0)), the filter coefficient considering the motion can be obtained.

  Note that the filter coefficient generation means 710 may be configured by a circuit that generates a filter coefficient by calculation processing from the value of the NR motion degree signal mds, instead of being configured by the ROM.

  FIG. 9 is an example conceptually showing the relationship between the value of the NR motion degree signal mds and the strength of noise removal in the processing by the filter coefficient generation means 710. The NR motion degree signal mds (horizontal axis) and noise removal The relationship of the effect (vertical axis) is shown. When the NR motion level signal mds is MD1 (for example, 4) having the maximum value and indicates complete motion, the effect of noise removal is minimized, the motion level is the minimum, and it is detected that the image is a complete still image. In other words, when the NR motion degree signal mds is 0, which is the minimum value (this includes the case where the difference is due to noise), kn1, kn2, kn3, kn4) = (0.25, 0.25, 0.25, 0.25), and the motion is taken into account by changing so that the noise removal effect becomes smaller as the NR motion degree signal mds increases. Filter coefficients are obtained.

  The filter coefficients kn1 to kn4 generated by the filter coefficient generation unit 710 are supplied to the coefficient multiplication units 701, 702, 703, and 704.

  The noise removal filter means 700 receives the first to fourth NR processing signals r1 to r4 and the filter coefficients kn1 to kn4, and uses the signals of the pixels at the same position in 4 fields or 4 frames to use the acyclic type. Perform noise removal processing.

  Coefficient multiplication means 701, 702, 703, and 704 shown in FIG. 5 perform an operation of multiplying the NR processing signals r1 to r4 by filter coefficients kn1 to kn4. That is, the first NR processing signal r1 and the filter coefficient kn1 are input to the coefficient multiplying unit 701, and the first NR processing signal r1 is multiplied by the filter coefficient kn1 to obtain a multiplication result r1 × kn1. Similarly, the coefficient multiplication means 702 receives the second NR processing signal r2 and the filter coefficient kn2, and obtains a multiplication result r2 × kn2. Similarly, the coefficient multiplying unit 703 receives the third NR processing signal r3 and the filter coefficient kn3, and obtains a multiplication result r3 × kn3. Similarly, the fourth NR processing signal r4 and the filter coefficient kn4 are input to the coefficient multiplying means 704, and the multiplication result r4 × kn4 is obtained.

  Coefficient multiplication results r1 × kn1, r2 × kn2, r3 × kn3, and r4 × kn4 output from the coefficient multipliers 701, 702, 703, and 704 are input to the adding unit 705. The adding means 705 adds the four input multiplication results, and outputs a signal obtained by the addition as a real line signal Rnr after noise removal.

  The filter coefficients kn1 to kn4 are generated according to the value of the NR motion degree signal mds, and the filter coefficient is set so that the effect of noise removal becomes smaller as the degree of motion indicated by the NR motion degree signal mds increases. Values of kn1 to kn4 are determined. Note that “when the degree of motion indicated by the NR motion degree signal mds is small” includes “when the frame difference is detected as being caused by noise”.

  The case where the acyclic noise removal processing unit 70 is used as the actual line noise removal unit 21 has been described above. However, as described above, when the non-cyclic type noise removal processing unit 70 is used as the first interpolation line noise removal unit 221, The operation is the same except that signals i1 to i4 are input instead of signals r1 to r4.

  As described above, the acyclic noise removal processing means 70 detects the movement of the signal between 4 fields (every other field) or 4 frames (consecutive), and the NR motion degree signal obtained as a result of the detection. Filtering is performed using the filter coefficient generated based on mds, and a signal after acyclic noise removal in which the strength of noise removal is controlled is output. In this way, blurring and deterioration of the edges and contours of the moving part can be prevented, the trailing part and afterimage of the moving part can be accommodated within the range of 4 fields or 4 frames, and the trailing and afterimage can be reduced. Further, noise components having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed.

  In this way, the real line noise removal means 21 constituted by the non-cyclic type noise removal processing means 70 outputs the non-cyclic type noise-removed real line signal Rnr, which is also the non-cyclic type noise removal process. A first interpolation line signal Isnr from which acyclic noise has been removed is output from the first interpolation line noise removal means 221 configured by the means 70.

  Next, the configuration of the second interpolation line noise removing unit 222 in the interpolation line noise removing unit 22 in FIG. 2 will be described in detail.

  The second interpolation line noise removing unit 222 performs filtering using the input first to fourth NR processing signals i1 to i4, and performs interpolation processing on the interpolation line signal used at the time of three-dimensional IP conversion. A non-cyclic noise removal process is performed, and a second interpolation line signal Itnr is output. As described above, the second interpolation line signal Itnr is converted into a 1-field delay signal d1f (first NR processing signal i1), which is an adjacent field after the interpolation field, as shown in FIG. On the other hand, it is a signal from which noise is removed, and the first NR processing signal i1 is a signal in the center field.

  FIG. 10 shows an acyclic noise removal processing means 71 that can be used as the second interpolation line noise removal means 222 of FIG. The non-cyclic noise removal processing means 71 shown in FIG. 10 uses the first NR processing signal i1 among the four input NR processing signals i1 to i4 as a signal of the center field CFIt, and performs non-cyclic noise removal processing. I do. The acyclic noise removing unit 71 shown in FIG. 10 is generally configured in the same manner as in FIG. 5, and the same or corresponding components as those in FIG. 5 are given the same reference numerals.

In FIG. 10, the acyclic noise removal processing means 71 includes a noise removal filter means 700, an NR motion detection means 721, and a filter coefficient generation means 711.
The noise removal filter unit 700 has the same configuration as that shown in FIG.
The NR motion detection means 721 detects the motion of the signal based on the four input NR processing signals i1 to i4 with respect to the signal of the center field CFIt, and outputs an NR motion degree signal mdsb.
The NR motion detection unit 721 includes a frame difference detection unit 730, an edge adjustment unit 741, and a conversion unit 750.
The edge adjustment unit 741 detects an edge part in the signal of the center field CFIt, adjusts the motion difference signal of the edge part, and averages the target pixel and surrounding pixels.
The filter coefficient generation unit 711 generates and outputs a filter coefficient based on the NR motion degree signal mdsb generated by the NR motion detection unit 721.
The configuration and operation of the noise removal filter means 700 in FIG. 10 and the frame difference detection means 730 and conversion means 750 in the NR motion detection means 721 are the same as those shown in the acyclic noise removal processing means 70 in FIG. Detailed description thereof will be omitted.

  The first to fourth NR processing signals i1 to i4 shown in FIG. 3 are input to the non-cyclic type noise removal processing means 71 via the input terminals 751 to 754, of which the first NR processing signal i1. Is used as a signal in the center field, and acyclic noise removal processing is performed.

  The non-cyclic noise removing unit 71 detects the motion of the signal between the four fields based on the NR processing signals i1 to i4 input by the NR motion detecting unit 721, and the NR motion level obtained as a result of the detection. Based on the signal mdsb, the filter coefficient generation means 711 generates a filter coefficient, and controls the noise removal strength to perform the acyclic noise removal processing.

  Here, the processing performed by the non-cyclic type noise removing means 71 is a filtering process using frame correlation between four fields (every other field) as in the case of FIG. The means 721 detects the motion of the signal between the four fields, generates a filter coefficient based on the NR motion degree signal mdsb obtained as a result of the detection, and controls the strength of noise removal, so that the edge and contour of the motion portion Can prevent blur and deterioration.

The first to fourth NR processing signals i1 to i4 are input to the NR motion detection means 721.
In the NR motion detection means 721, a difference (frame difference) between the first NR processing signal i1 as the center field signal and the second to fourth NR processing signals i2 to i4 is obtained and obtained. The frame difference is synthesized to detect the motion in the video signal, and the NR motion degree signal mdsb is output.

  The NR motion degree signal mdsb is generated in the same manner as the NR motion degree signal mds in FIG. 5 and indicates the degree of motion with a value between 0 and MD1 (for example, MD1 = 4). The NR motion degree signal mdsb from the NR motion detector 721 is sent to the filter coefficient generator 711.

  The frame difference detection means 730 of the NR motion detection means 721 has the same configuration as that in FIG. 5 and its detailed configuration is omitted, but the first to fourth NR processing signals i1 to i4 are input, The difference (three frame differences) between the NR processing signal i1 and the second to fourth frames i2 to i4 is calculated, and the motion difference signal Dafb is output by combining these frame differences. To do.

  Here, as shown in FIG. 3, the frame difference dif1 obtained by the difference detection means 731 is 1 between the first NR processing signal i1 and the second NR processing signal i2 one frame before (two vertical periods before). This is a difference between frames (2 fields), and the frame difference dif2 obtained by the difference detection means 732 is the difference between the first NR processing signal i1 and the third NR processing signal i3 two frames before (four vertical periods before). It is a difference between two frames (four fields), and the frame difference dif3 obtained by the difference detection means 733 is for the first NR processing signal i1 and the fourth NR processing three frames before (six vertical periods before). This is the difference between 3 frames (6 fields) of the signal i4. The input signals to the difference detection means 731, 732, and 733 are different from those in FIG. 5, but the difference detection means 731, 732, and 733 can have the same configuration, and their operations are the same as each other. is there.

The synthesizing unit 734 synthesizes the frame differences dif1, dif2, and dif3 (for example, outputs the maximum values of the frame differences dif1, dif2, and dif3 as synthesis results), and outputs the motion difference signal Dafb. The synthesizing unit 734 may have the same configuration as that shown in FIG. 5 and operates in the same manner as that shown in FIG.
The motion difference signal Dafb output from the frame difference detection unit 730 is sent to the edge adjustment unit 741.

  The edge adjustment unit 741 receives the motion difference signal Dafb from the frame difference detection unit 730 and the first NR processing signal i1. The edge adjustment unit 741 adjusts the motion difference signal Dafb in the pixel determined to constitute the edge so as to represent a larger movement, and for the pixel not determined to constitute the edge. The isolated point is removed by averaging the target pixel and the surrounding pixels, and the adjusted motion difference signal Dfeb is output. The configuration of the edge adjustment means 741 is generally the same as that of the edge adjustment means 740 shown in FIG. 8, but the first NR processing signal i1 is sent to the edge determination means 744 instead of the second NR processing signal r2. The difference is that the motion difference signal Dafb is input to the averaging means 742 and the adjustment means 743 instead of the motion difference signal Daf.

  The first NR processing signal i1 is input to the edge determination unit 744 in the edge adjustment unit 741. The edge adjustment unit 741 extracts the high frequency components in the horizontal and vertical directions of the pixel of interest in the field based on the input first NR processing signal i1, and the pixel of interest is edged from the value of the extracted component. Is determined (edge determination). In this case, the high frequency component in the vertical direction is converted into a pixel on another line (for example, i1_0 in FIG. 3) aligned in the vertical direction with respect to the target pixel (for example, i1 in FIG. 3) by line delay processing or the like. And a high frequency component between lines (for example, a value obtained by converting the bandpass filter output into an absolute value) is extracted. Note that a high frequency component (a value obtained by converting the band-pass filter output into an absolute value) between both the pixel located above and the pixel located below the pixel position of interest as the center may be extracted.

  The conversion unit 750 in the edge adjustment unit 741 receives the motion difference signal Dfeb from the edge adjustment unit 741, and converts the NR motion degree signal mdsb from the motion difference signal Dfeb that has been subjected to edge adjustment and averaging. Generate and output.

  Next, the NR motion degree signal mdsb from the NR motion detector 721 is input to the filter coefficient generator 711 shown in FIG. The filter coefficient generation unit 711 obtains filter coefficients ko1 to ko4 corresponding to the NR motion degree signal mdsb. The filter coefficients ko1 to ko4 are used in the acyclic noise removal filter processing in the noise removal filter means 700.

  The filter coefficients ko1 to ko4 can be generated by the same method as the filter coefficients kn1 to kn4 in FIG. 5 described above, and set the coefficients of the filter that performs noise removal. The noise removal strength (effect) is determined by assigning the values of the filter coefficients ko1 to ko4 for the four NR processing signals.

  The filter coefficients ko1, ko2, ko3, and ko4 for the first, second, third, and fourth NR processing signals i1, i2, i3, and i4 are values having the first NR processing signal i1 as a signal in the center field. Is determined. When the filter coefficients ko1 to ko4 have the same value 1/4 (= 0.25), noise components having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, if the value of the coefficient ko1 with respect to the first NR processing signal i1 that is the signal in the center field is large, ko1 = 1, and the other coefficients ko2, ko3, and ko4 are zero, the signal in the previous field is On the other hand, when 0 is multiplied, the noise cyclic value becomes zero, and the signal is output as it is, and noise removal is not performed.

  When the NR motion level signal mdsb = 4 and indicates complete motion, (ko1, ko2, ko3, ko4) = (1, 0, 0, 0), the motion level is the minimum, and a complete still image When mdsb = 0 is detected (this includes the case where the difference is due to noise), (ko1, ko2, ko3, ko4) = (0.25, 0.25, 0. 25, 0.25). When the NR motion degree signal mdsb has an intermediate value (mds = 1, 2, 3), for example, (ko1, ko2, ko3, ko4) = (7/16, 4/16, 3/16, 2) / 16), (1/3, 1/3, 1/3, 0), (0.5, 0.5, 0, 0), etc., to control the strength of noise removal .

  As the degree of motion (value of the NR motion degree signal mdsb) increases, the value of the coefficient ko1 for the signal i1 of the center field CFIt increases and the values of the other coefficients decrease (for example, (ko1, ko2, ko3, ko4). ) = (7/16, 4/16, 3/16, 2/16)), or the number of fields used for filter processing (the number of NR processing signals with coefficients other than 0) is reduced to 3 (for example, ((Ko1, ko2, ko3, ko4) = (1/3, 1/3, 1/3, 0)) or reduced to 2 (for example, (ko1, ko2, ko3, ko4) = (0.5 , 0.5, 0, 0)), the filter coefficient considering the motion can be obtained.

  The filter coefficients ko1 to ko4 generated by the filter coefficient generation unit 711 are supplied to the coefficient multiplication units 701, 702, 703, and 704.

The noise removal filter means 700 receives the first to fourth NR processing signals i1 to i4 and the filter coefficients ko1 to ko4, and uses the signals of the pixels at the same position in the four fields to perform acyclic noise removal processing. I do.
The noise removal filter means 700 is shown in FIG. 5 except that the input NR processing signals are i1 to i4 instead of r1 to r4, and the filter coefficients are ko1 to ko4 instead of kn1 to kn4. The first and fourth NR processing signals i1 to i4 are multiplied by the filter coefficients ko1 to ko4 and added to output a filtered signal.

  The filter coefficients ko1 to ko4 are generated according to the value of the NR motion degree signal mdsb, and the filter coefficient is set such that the greater the degree of motion indicated by the NR motion degree signal mdsb, the smaller the noise removal effect. Values of ko1 to ko4 are determined. Note that “when the degree of motion indicated by the NR motion degree signal mdsb is small” includes “when the frame difference is detected as being caused by noise”.

  As described above, the acyclic noise removal processing means 71 detects the signal motion between four fields (every other field) and generates the filter coefficient based on the NR motion degree signal mdsb obtained as a result of the detection. Is used to perform a filtering process to output a signal after acyclic noise removal in which the strength of noise removal is controlled. In this way, as with the non-recursive noise removal processing means 70, blurring and deterioration of the edges and contours of the moving parts are prevented, and the trailing and afterimages of the moving parts are kept within the range of 4 fields. Afterimages can be reduced, and noise components having no frame correlation over 4 fields (8 vertical periods) can be removed.

  In this way, the second interpolation line noise removal unit 222 configured by the non-cyclic type noise removal processing unit 71 outputs the second interpolation line signal Itnr from which the non-cyclic type noise has been removed.

  As described above, the real line noise removing unit 21 and the interpolated line noise removing unit 22 shown in FIG. 2 perform filtering using signals of 4 fields (every other field) or 4 frames (consecutive). It is possible to obtain a signal after non-cyclic noise removal processing, and to keep the tail and afterimage of the moving part within the range of 4 fields or 4 frames, and to reduce the tail and afterimage. Noise component having no frame correlation over 8 vertical periods) or 4 frames (4 vertical periods), and after noise removal, the real line signal Rnr, the first interpolation line signal Isnr, and the second interpolation line As the signal Itnr, a signal from which the noise component is removed can be obtained.

  As described above, in the video signal processing apparatus 1 according to the first embodiment, the delayed signals d1f to d7f are obtained from the field memories 11 to 17, and the acyclic noise removal process is performed using the delayed signals. When the input video signal is an interlaced signal, three-dimensional IP conversion is performed using the delayed signal.

  FIG. 11 is a flowchart showing an operation in the video signal processing apparatus 1 for obtaining the delayed signals d1f to d7f, performing acyclic noise removal processing, performing three-dimensional IP conversion, and outputting a noise-removed progressive signal. FIG. 12 is a flowchart showing the operation of the non-cyclic noise removal process among the above. Hereinafter, a description will be given with reference to FIGS. 11 and 12.

  First, it is assumed that the input video signal to the video signal processing device 1 is an interlace signal, and the input interlace signal Di0 is subjected to non-cyclic noise removal processing and subjected to three-dimensional IP conversion to be denoised. The operation of outputting a progressive signal will be described with reference to FIG. In the following description, it is assumed that the signal of the interpolation field to be subjected to IP conversion, that is, scanning line interpolation when the interlace signal is input is the two-field delay signal d2f.

  When an interlace signal is input, the switching signal pf from the switching signal generation means 18 indicates that the input signal is an interlace signal.

  An input signal Di0, which is an interlaced signal input to the video signal processing device 1, is input to the field memory 11, and the field memory 11 to the field memory 17 sequentially delays the signal by one field, and the one field delayed signal d1f to the seven fields. A delay signal of a total of 7 fields of the delay signal d7f is output (S1). As a result, a delay signal used in noise removal processing and three-dimensional IP conversion is obtained.

The input signal Di0 and the delayed signals df1 to df7 are sent to the acyclic noise removal processing means 20, and noise removal is performed by filtering each of the actual line signal and the interpolation line signal using a signal of 4 fields. I do.
For the real line signal (2-field delay signal d2f), the input signal Di0, the 2-field delay signal d2f, the 4-field delay signal d4f, and the 6-field delay signal d6f are the first to fourth NR processing signals r1 to r4. Is supplied to the actual line noise removing means 21 and the actual line noise removing means 21 performs filtering using these four NR processing signals r1 to r4 to obtain a real line signal Rnr after noise removal ( S2).

  The field of the interpolation line signal is a signal of a field adjacent before and after the interpolation field, and the interpolation line signal is a three-field delay signal d3f and a one-field delay signal d1f. For these interpolation line signals, the 1-field delay signal d1f, the 3-field delay signal d3f, the 5-field delay signal d5f, and the 7-field delay signal d7f are the first interpolation line noise removal means 221 and the second interpolation line. The first to fourth NR processing signals i1 to i4 are input to the noise removing unit 222, and noise is removed by performing filtering using the four input NR processing signals i1 to i4, respectively. The first interpolation line signal Isnr and the second interpolation line signal Itnr are obtained (S3).

  Detailed operations in the real line noise removing unit 21 and the interpolation line noise removing unit 22 will be described later with reference to FIG.

  The input signal Di0 and the delayed signals df1, d2f, and df4 are also used for generating a three-dimensional scanning line interpolation signal (generation of an interpolation line signal by three-dimensional IP conversion). The first interpolation line signal Isnr and the second interpolation line signal Itnr are used to generate a three-dimensional scanning line interpolation signal.

  The two-field delay signal d2f, which is a signal of the interpolation field, is input to the intra-field interpolation processing unit 30. The intra-field interpolation processing unit 30 performs filtering processing or the like by the pixels in the scanning lines adjacent in the vertical direction in the interpolation field. An arithmetic process is performed to generate an interpolation signal Imv (S4).

  The current field signal Di0 and the two-field delay signal d2f and the four-field delay signal d4f are input to the motion detection means 40, and during one frame (for two vertical periods) between the current field signal Di0 and the two-field delay signal d2f. Difference information and difference information between the current field signal Di0 and the four-field delay signal d4f between two frames (four vertical periods) are obtained, and a video signal between one frame and two frames is obtained based on the obtained difference information. The motion detection signal mdt is obtained (S5).

  Further, the telecine detection means 50 receives the current field signal Di0 and the two-field delay signal d2f and the one-field delay signal d1f, and the frame difference information between the current field signal Di0 and the two-field delay signal d2f; The field difference information between the field Di0 of the current field and the 1-field delay signal d1f is obtained, and it is detected whether the input video signal is a telecine video signal by 2-3 pulldown signal or 2-2 pulldown signal. Based on the detection result, a telecine detection signal tcif for performing telecine interpolation is obtained (S6).

  The generation of the scanning line interpolation signal is performed in the interpolation signal generation means 60, and the telecine detection means 50 outputs the first interpolation line signal Isnr and the second interpolation line signal Itnr and the intra-field interpolation signal Imv. When the telecine detection signal tcif does not indicate that it is a telecine video signal, an interpolated signal is generated by motion adaptation processing based on the motion detection signal mdt from the motion detection means 40, while the telecine detection from the telecine detection means 50 is detected. When the signal tcif indicates that it is a telecine video signal, telecine interpolation is performed to obtain an interpolation line signal Im (S7).

  The time axis conversion means 61 receives the interpolation signal Im from the interpolation signal generation means 60 and the noise-removed real line signal Rnr from the acyclic noise removal processing means 20, and converts the interpolation signal into the real line of the interpolation field. The signal is fitted between corresponding scanning lines, double-speed converted, and converted into a progressive signal Ips (S8).

  Since the switching means 62 indicates that the switching signal pf is an interlace signal input, the progressive signal Ips from the time axis conversion means 61 is selected and the output video signal Do0 is obtained (S9).

When a progressive signal is input, since it is not necessary to perform IP conversion, the operations from step S3 to step S7 are omitted, and the input signal delay processing (S1) and the noncyclic noise removal processing by the noncyclic noise removal processing means 20 are omitted. (S2), the switching means 62 selects the real line signal Rnr after noise removal, and becomes the output video signal Do0.
Even when a progressive signal is input, a non-cyclic noise removal process is performed on the one-field delayed signal d1f to reduce tailing and afterimages, and a progressive signal from which noise has been successfully removed can be obtained.

  Next, detailed operations in the real line noise removing unit 21 and the interpolated line noise removing unit 22 in the acyclic noise removing processing unit 20 will be described in detail with reference to FIG.

  In the non-cyclic noise removal processing step S2 described with reference to FIG. 11, a real line signal Rnr after noise removal is obtained, and in step S3, the first interpolation line noise removal means 221 and the second interpolation line noise removal means are obtained. Through 222, a first interpolation line signal Isnr and a second interpolation line signal Itnr are obtained.

  There is a difference between the signal of each central field or frame being the second NR processing signal r2 or i2 or the first NR processing signal i1, and the value of the filter coefficient for the 4 field or 4 frame signal used, Although the method of taking the frame difference at the time of motion detection is different, the basic noise removal processing is the same. Hereinafter, the case where the signal of the center field or frame is the second NR processing signal r2 will be described with reference to FIG.

  In FIG. 12, the signals of the first to fourth NR processing signals r1 to r4 are input to the acyclic noise removal processing means 70, and the second NR processing signal r2 becomes the signal of the central field or frame. (S11).

  The non-cyclic noise removing means 70 detects the signal motion between the four fields or four frames input by the NR motion detecting means 720, and based on the NR motion degree signal mds obtained as a result of the detection, the filter coefficient And acyclic noise removal processing is performed by controlling the strength of noise removal.

  The NR motion detection means 720 receives the first to fourth NR processing signals r1 to r4, the second NR processing signal r2 as a center field or frame signal, and the first, third, and fourth signals. The difference (frame difference) between the NR processing signals r1, r3, and r4 is calculated, and three frame differences dif1, dif2, and dif3 are detected (S12).

  Next, the synthesizing unit 734 synthesizes the frame differences dif1, dif2, and dif3 (for example, obtains the maximum value) to obtain the motion difference signal Daf (S13).

  The edge adjustment unit 740 detects an edge portion in the signal r2 of the center field or frame, and the pixel determined to form an edge has a larger motion than the motion difference signal Daf obtained by the frame difference detection unit 730. In addition to performing adjustment so as to represent the pixel, the pixel that has not been determined to constitute an edge is averaged with surrounding pixels to remove isolated points and obtain an adjusted motion difference signal Dfe (S14). In this way, the edge of the signal and the boundary to be the contour portion are determined, the motion difference signal of the edge portion is adjusted, and the motion difference signal Dfe subjected to the averaging process is obtained. Flickering due to detection omission can be prevented, and blurring and deterioration of edges and contours of moving parts can be prevented.

  The conversion unit 750 generates an NR motion degree signal mds from the motion difference signal Dfe subjected to the edge adjustment and averaging from the edge adjustment unit 740 (S15).

  That is, the input motion difference signal Dfe is subjected to conversion processing by amplitude limitation or the like, for example, the degree of motion is set to the (MD1 + 1) stage (for example, MD1 = 4), and it is detected as a complete motion. If the value is MD1 (for example, MD1 = 4, complete motion pixel) and it is detected that the image is a complete still image (this may be detected as a noise component, not a difference due to motion). NR motion degree signal mds indicating the degree of motion with a value between 0 and MD1 (for example, MD1 = 4) is generated as zero (mds = 0, complete still pixel).

  As in steps S11 to S15, motion detection is performed by the NR motion detection unit 720, and the NR motion degree signal mds is sent to the filter coefficient generation unit 710.

  The filter coefficient generation means 710 receives the NR motion degree signal mds from the NR motion detection means 720, and obtains filter coefficients kn1 to kn4 corresponding to the NR motion degree signal mds (S16).

  The filter coefficients kn1 to kn4 are used in a filter that performs non-cyclic noise removal. Each of the filter coefficients kn1 to kn4 is determined to be 0 to 1 and the sum is 1, and filter coefficients kn1 to NR1 for four NR processing signals are used. The noise removal strength (effect) is determined by assigning the value of kn4.

  The values of the filter coefficients kn1 to kn4 are determined using the second NR processing signal r2 as a central field or frame signal. When the filter coefficients kn1 to kn4 have the same value 1/4 (= 0.25), the noise component having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, when the value of the coefficient kn2 for the signal r2 in the center field or frame is large, kn2 = 1, and the other filter coefficients kn1, kn3, and kn4 are zero, 0 for the signals in the preceding and succeeding fields or frames. No noise removal is performed.

Therefore, the filter coefficient generation means 710 controls the values of the filter coefficients kn1 to kn4 in accordance with the NR motion degree signal mds in order to prevent blurring and deterioration of the edges and contours of the moving part in the video signal. Controls the strength of the noise removal of the part.
That is, when the NR motion degree signal mds is MD1 (for example, 4) and indicates complete motion, (kn1, kn2, kn3, kn4) = (0, 1, 0, 0), and the degree of motion is the minimum level. When mds = 0, which is detected as a complete still image (this includes a case where the difference is due to a noise component), (kn1, kn2, kn3, kn4) = (0.25, 0 .25, 0.25, 0.25). When the NR motion degree signal mds has an intermediate value, the value of the coefficient kn2 with respect to the signal r2 of the center field or frame increases as the degree of motion (value of the NR motion degree signal mds) increases, and other coefficients (For example, (kn1, kn2, kn3, kn4) = (2/8, 3/8, 2/3, 1/3)) or the number of fields or frames used for filtering (the coefficient is Reduce the number of NR processing signals to a value other than 0 to 3 (for example, (kn1, kn2, kn3, kn4) = (1/3, 1/3, 1/3, 0)) or 2 By changing so as to reduce (for example, (kn1, kn2, kn3, kn4) = (0.5, 0.5, 0, 0)), a filter coefficient considering movement is obtained, and noise removal is performed. Control strength

  In the noise removal filter unit 700, four NR processing signals (signals of pixels at the same position in four fields or four frames) are obtained by the first to fourth NR processing signals r1 to r4 and the filter coefficients kn1 to kn4. Is used to perform acyclic noise removal processing (S17), and the resulting noise-removed signal Rnr is output (S18).

  In this filter processing, the first to fourth NR processing signals r1 to r4 are multiplied by filter coefficients kn1 to kn4, and the sum of the multiplication results is obtained.

  The case where the non-recursive noise removing unit 70 used as the real line noise removing unit 21 generates the real line signal Rnr after noise removal has been described above. However, the non-cyclic type noise removing unit 221 used as the first interpolation line noise removing unit 221 has been described. The processing for obtaining the first interpolation line signal Isnr by the mold noise removing means 70 is similarly performed.

The processing in the case of obtaining the second interpolation line signal Itnr by the acyclic noise removal processing means 71 used as the second interpolation line noise removal means 222 is generally the same. However, when obtaining the second interpolation line signal Itnr, the first NR processing signal i1 is used as the signal of the center field. That is, the difference between the first NR processing signal i1 as the center field signal and the second, third, and fourth NR processing signals i2, i3, i4 in the difference calculation in step S12 ( In step S14, the edge adjustment unit 741 detects an edge portion in the first NR processing signal i1, and in step S16, the first NR processing signal i1 is used as a central field signal. The difference is that filter coefficients ko1 to ko4 having values for performing filtering are used.
In addition, the basic noise removal processing described in steps S11 to S18 is the same as in FIG.

  As described above, according to the video signal processing apparatus 1 of the first embodiment, the delayed signals d1f to d7f are obtained from the field memories 11 to 17, and the 4 fields or 4 frames of the delayed signals and the current field or frame signals are obtained. The non-cyclic noise removal processing means 20 performs filtering using the signal to perform non-cyclic noise removal processing. For interlaced signals, the signal after noise removal and the non-cyclic noise removal processing are also used. The three-dimensional IP conversion processing unit 65 performs three-dimensional IP conversion based on the delayed signal and the current field signal, and outputs a progressive signal. Then, the non-cyclic noise removal processing means 20 detects a motion between four fields or four frames, generates a filter coefficient based on the detection result, and a non-cyclic noise removal whose noise removal strength is controlled. It is carried out. Therefore, the delay signal used in the acyclic noise removal processing and the three-dimensional IP conversion processing is obtained by the common field memory, and the field memory as the field delay means is efficiently used without significantly increasing the field memory. A noise component having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed, and a three-dimensional scanning line interpolation is performed by a signal subjected to such acyclic noise removal. Processing can be performed. The non-cyclic noise removal process can prevent blurring and deterioration of the edges of moving parts, contours, and reduce adverse effects such as tailing and afterimage without significantly increasing the circuit scale such as field memory. Noise removal can be performed satisfactorily, and a progressive signal from which noise has been removed can be obtained.

  The non-cyclic noise removal means 70 and 71 constituting the non-cyclic noise removal processing means 20 use the input signal Di0 and the field delay signals d1f to d7f and perform filter processing using signals of 4 fields or 4 frames. However, it may be configured to perform filtering using more than 4 fields or frames in the non-cyclic type noise removal processing, and to perform filtering processing using a larger number of fields or frames. If this is done, the calculation is performed between frames that are more distant from each other in time, noise components having no frame correlation in the time axis are removed, and the noise removal effect can be further increased.

  Further, in the noise removal processing in the non-cyclic type noise removal processing means 20, a filter process using three fields or frames, or two fields or frames may be performed. In short, it is only necessary to perform a filter process using two or more, that is, a plurality of fields or frames, of the current field signal and the delayed signal.

  In FIG. 1 and FIG. 11, in the interpolation motion detection means 66 in the three-dimensional IP conversion processing means 65, the motion detection means 40 detects the motion of the signal and the telecine detection means 50 detects the telecine, Although it has been described that the generation unit 60 generates an interpolation signal that has been three-dimensionally processed by motion adaptation processing or telecine interpolation, the interpolation signal generation by the three-dimensional IP conversion processing unit 65 may be only the motion adaptation processing, or the interpolation motion detection unit 66. May be composed of only the motion detection means 40. In that case, if the three-dimensional IP conversion is performed by using the first interpolation line signal Isnr in the non-cyclic type noise removal processing means 20, the second interpolation line signal Itnr becomes unnecessary, and the second interpolation The line signal Itnr can be omitted.

  In the video signal processing apparatus 1 according to the first embodiment, it is assumed that the output video signal Do0 when the progressive signal is input corresponds to the one-field delay signal d1f. As shown in FIG. 4, the line noise removal means 21 performs noise removal processing by filtering with a signal of four frames that include the current frame and are continuous (at intervals of one frame).

  However, instead of doing the above, as shown in FIG. 13, the two-field delay signal d2f is a signal of the center frame CFR, includes the center frame, and is positioned every two frames (at intervals of two frames). You may comprise so that a noise removal process may be performed by the filter process using the signal of the flame | frame.

Further, as shown in FIG. 14, a filtering process using a signal of four frames, which includes the center frame CFR as the signal Di0 of the current frame and is located at each frame (at intervals of one frame), including the center frame. It may be configured to remove noise.
Thus, in the configuration of the real line noise removal means 21 in the non-cyclic type noise removal processing means 20, even if the input signal at the time of progressive signal input is changed, the same effect as described with reference to FIG. Is obtained.

  In addition, when performing noise removal processing by filtering using signals of 4 frames located every 2 frames, calculation is performed between frames that are further apart in time, and frame correlation is performed in a more distant time axis. Noise components without noise can be obtained, and the noise removal effect can be further increased.

Embodiment 2. FIG.
As shown in FIGS. 5 and 10, the real line noise removing unit 21 and the interpolation line noise removing unit 22 in the video signal processing apparatus 1 according to the first embodiment are NR motion degree signals mds in filter coefficient generating units 710 and 711, respectively. Filtering is performed using a filter coefficient that changes according to mdsb, but instead of this configuration, a predetermined (constant) noise removal effect strength is set as shown in FIGS. In order to obtain a signal after noise removal by mixing a signal that has been subjected to noise removal filter processing with a filter coefficient to be processed and a signal that has not been subjected to noise removal at a mixing ratio according to the values of the NR motion degree signals mds and mdsb. Cyclic noise removal means can also be configured.

The overall configuration of the video signal processing apparatus according to the first embodiment (that is, the apparatus that can implement the video signal processing method according to the second embodiment) is the same as that shown in FIG. However, as each of the actual line noise removing unit 21 and the first interpolation line noise removing unit 221, the acyclic noise removal processing unit 80 shown in FIG. 15 is used, and as the second interpolation line noise removing unit 222, 16 is used.
In FIG. 15, the same reference numerals are given to the same or corresponding components as those in FIG. 5. In FIG. 16, the same or corresponding components as those in FIG. 10 are denoted by the same reference numerals.

  In the video signal processing apparatus of the second embodiment, the configuration for the field delay processing, the signal input to the acyclic noise removal processing means 20, the processing for generating the interpolation line signal and outputting the progressive signal is described in the embodiment. 1 is the same as that described with reference to FIG. 1 and FIG. 2, and detailed description thereof is omitted. In the following, the actual line noise removing unit 21, the first interpolation line noise removing unit 221 and the second Non-cyclic noise removal processing means 80 and 81 used as the interpolation line noise removal means 222 will be described.

  Below, the case where the acyclic noise removal process means 80 is mainly used as the real line noise removal means 21 is demonstrated. The operation when the acyclic noise removal processing unit 80 is used as the first interpolation line noise removal unit 221 is the same, but the signals i1, i2, i3, i4 are replaced with the signals r1, r2, r3, r4. It differs in that it is input.

  The non-cyclic noise removal processing means 80 shown in FIG. 15 is the same as the non-cyclic noise removal processing means 70 of FIG. 5, and the second NR processing signal among the four input NR processing signals r1 to r4. A non-recursive noise removal process is performed using r2 as a central field or frame signal, and the configuration and operation other than those relating to the mixing means 801 and the filter coefficient setting means 820 are generally the same as those shown in FIG. .

The acyclic noise removal processing means 80 includes a noise removal filter means 700, an NR motion detection means 720, a filter coefficient setting means 820, and a mixing means 801.
The noise removal filter unit 700 and the NR motion detection unit 720 are configured in the same manner as those in the non-cyclic type noise removal processing unit 70 of FIG.

That is, the noise removal filter unit 700 performs filter processing using the first to fourth NR processing signals r1 to r4. A noise removal filter output signal output from the noise removal filter means 700 is indicated by a symbol dnr.
The NR motion detection means 720 detects the motion of the signal and outputs an NR motion degree signal mds.

The filter coefficient setting unit 820 sets a predetermined value as the filter coefficient of the noise removal filter. Here, the “predetermined value” means a value that is kept constant unless a separate process for setting (or changing) the value as described later is performed.
The mixing unit 801 outputs the NR from the NR motion detection unit 720 to the noise removal filter output signal dnr output from the addition unit 705 in the noise removal filter unit 700 and the signal r2 of the center field or frame that has not been subjected to noise removal. A mixing process is performed at a mixing ratio corresponding to the value of the movement degree signal mds, and a signal Rnr after noise removal is output.

  The first to fourth NR processing signals r1 to r4 shown in FIG. 3 and FIG. 4 are input to the non-cyclic type noise removal processing means 80 through the input terminals 751 to 754, and the second NR Acyclic noise removal processing is performed using the processing signal r2 as the signal of the center field or frame CFR.

  The acyclic noise removing unit 80 detects the motion of the signal between four fields or four frames based on the NR processing signals r1 to r4 input from the NR motion detecting unit 720, and is obtained as a result of the detection. Based on the NR motion degree signal mds, the noise removal filter output signal dnr and the signal r2 of the center field or frame where noise removal is not performed are determined, and after the noise removal by performing a mixing process at a predetermined mixture ratio Signal Rnr is obtained. In this way, the acyclic noise removal process in which the strength of noise removal is controlled is performed.

The first to fourth NR processing signals r1 to r4 are input to the NR motion detection means 720.
In the NR motion detection means 720, the difference (frame difference) between the second NR processing signal r2 and the first, third, and fourth NR processing signals r1, r3, r4 is obtained and obtained. The frame difference is combined to detect motion in the video signal, and an NR motion level signal mds indicating the level of motion is output as a result of motion detection.

The NR motion detection means 720 operates in the same manner as that shown in FIG. 5 and generates the NR motion degree signal mds.
The configuration so far is the same as that in FIG. 5, but in FIG. 15, the NR motion degree signal mds from the NR motion detection means 720 is a signal that has not been subjected to noise removal with the noise removal filter output signal dnr. Is sent to the mixing means 801 as a movement degree signal indicating the degree of movement for controlling the mixing ratio.

  Next, the filter coefficient setting unit 820 in the non-cyclic type noise removal processing unit 80 shown in FIG. 15 sets and outputs the filter coefficients kf1 to kf4 according to, for example, an operation of the user or the like. Instead of setting according to the operation by the user or the like, the noise amount in the input video signal may be detected, and the filter coefficient may be automatically set based on the detected noise amount.

  The filter coefficients kf1 to kf4 are set to values at which a noise removal effect having a preset strength can be obtained.

  The filter coefficients kf1 to kf4 set in the filter coefficient setting unit 820 are the same as those described above with reference to FIG. 5 regarding the filter coefficients of the filter coefficient generation units 710 and 711 in the first embodiment. A coefficient is set, and the strength (effect) of noise removal is determined by assigning the values of the filter coefficients kf1 to kf4 for the four NR processing signals, and the filter coefficients kf1 to kf4 are 0 or more and 1 or less, respectively. The sum is determined to be 1.

  The filter coefficients kf1 to kf4 for the first to fourth NR processing signals r1 to r4 are determined by using the second NR processing signal r2 as a central field or frame signal. When the filter coefficients kf1 to kf4 are the same value 1/4 (= 0.25), the noise component having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, if the value of the coefficient kf2 for the signal r2 in the central field or frame is large, kf2 = 1, and the other coefficients kf1, kf3, kf4 are zero, 0 for the signals in the preceding and succeeding fields or frames. , The noise cyclic value becomes zero, and the input signal is output as it is, and noise removal is not performed.

  The effect of noise removal is to increase the value of the coefficient kf2 for the signal r2 in the center field or frame and decrease the values of the other coefficients (for example, (kf1, kf2, kf3, kf4) = (2/8, (3/8, 2/3, 1/3)) or the number of fields or frames used for filtering (the number of NR processing signals with coefficients other than 0) is reduced to 3 (for example, ((Kf1, kf2, kf3, kf4) = (1/3, 1/3, 1/3, 0)) or reduced to 2 (for example, (kf1, kf2, kf3, kf4) = (0.5 , 0.5, 0, 0)), and can be made weaker by changing the filter coefficient.

  That is, when maximizing the noise removal effect, the filter coefficients kf1 to kf4 are set to the same value ((kf1, kf2, kf3, kf4) = (0.25, 0.25, 0.25, 0.25). )). When setting the effect of noise removal to an intermediate strength or weaker, for example, (kn1, kn2, kn3, kn4) = (2/8, 3/8, 2/3, 1/3) or , (1/3, 1/3, 1/3, 0), (0.5, 0.5, 0, 0) are set to set the strength of noise removal.

  The filter coefficients kf1 to kf4 generated by the filter coefficient setting unit 820 are supplied to the coefficient multiplication units 701, 702, 703, and 704 in FIG.

  The noise removal filter unit 700 receives the first to fourth NR processing signals r1 to r4 and the filter coefficients kf1 to kf4. As in FIG. 5, the noise removal filter unit 700 performs filter processing using the four input NR processing signals and outputs a noise removal filter output signal dnr.

  Coefficient multiplying means 701, 702, 703, 704 perform operations for multiplying the first to fourth NR processing signals r1 to r4 by filter coefficients kf1 to kf4, and the multiplication results r1 × kf1, r2 × kf2, r3 × kf3, r4 × kf4 is obtained.

  Coefficient multiplication results r1 × kf1, r2 × kf2, r3 × kf3, and r4 × kf4 output from the coefficient multipliers 701, 702, 703, and 704 are input to the adding unit 705. The adding means 705 adds the four input multiplication results and outputs a signal obtained by the addition as a noise removal filter output signal dnr.

  The above configuration in the noise removal filter unit 700 is the same as that in FIG. Note that the filter coefficients kf1 to kf4 are coefficients for obtaining a noise removal effect having a preset strength, and the filter output signal dnr output from the adding means 705 is a predetermined value with respect to the center field or frame signal r2. This signal is noise-removed with strength (predetermined noise removal effect).

  To the mixing unit 801, a noise removal filter output signal dnr output from the addition unit 705 in the noise removal filter unit 700, a center field or frame signal r2, and an NR motion degree signal output from the NR motion detection unit 720. mds is input. The mixing unit 801 performs a mixing process on the noise removal filter output signal dnr and the central field or frame signal r2 at a mixing ratio corresponding to the value of the NR motion degree signal mds, thereby obtaining a signal after acyclic noise removal. Rnr is output.

  Here, the noise removal filter means 700 is a filtering process between 4 fields (every other field) or 4 frames (successive), and the tailing and afterimage of the moving part is in the range of 4 fields or 4 frames. The uncorrelated noise component over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed.

  However, since it is a filtering process, it is conceivable that tailing and afterimages appear as edges and blurring or degradation of the contour portion in the contour portion of the motion. Therefore, in the mixing unit 801, the noise removal filter output signal dnr and the signal r2 of the center field or frame from which noise removal has not been performed are performed based on the mixing ratio determined based on the NR motion degree signal mds from the NR motion detection unit 720. By mixing, the ratio of the signal from which noise is removed (occupying ratio) is controlled by the mixing ratio according to the movement of the signal, and the strength of noise removal is controlled. Can be prevented.

  Specifically, the mixing unit 801 determines a mixing ratio between the noise removal filter output signal dnr and the center field or frame signal r2 from the value of the NR motion degree signal mds from the NR motion detecting unit 720, and uses this mixing ratio. The noise removal filter output signal dnr and the center field or frame signal r2 are mixed.

  When the NR motion degree signal mds indicates the maximum level of motion (complete motion), the ratio (occupying ratio) of the signal r2 of the center field or frame is maximized so that noise removal is not performed (that is, the noise removal filter). (The mixing ratio of the output signal dnr and the center field or frame signal r2 is 0: 100). This is to prevent blurring and deterioration of the contour portion in a motion portion where the NR motion degree signal mds shows a large motion. On the other hand, when the degree of motion indicated by the NR motion degree signal mds is the minimum level and indicates a complete still image, the ratio occupied by the noise removal filter output signal dnr is maximized (that is, the noise removal filter). (The mixing ratio of the output signal dnr and the center field or frame signal r2 is 100: 0). When the NR motion level signal mds indicates an intermediate state between the complete motion and the complete still image, as the NR motion level signal mds increases, the ratio of the signal r2 in the center field or frame is changed so as to reduce noise. The removal effect is reduced.

  By the above processing, a signal from which noise is removed is obtained. As a result, blurring and deterioration of edges and contours are likely to occur in the moving portion, but the noise-removed signal is obtained by mixing the noise removal filter output signal dnr and the signal r2 of the central field or frame where noise removal is not performed. Therefore, the noise removal effect can be increased when setting the filter coefficient by the filter coefficient setting means 820, and the noise removal effect by the filter processing is changed step by step according to the degree of motion, Edge and contour blur and deterioration can be prevented.

  FIG. 17 is an example showing the relationship between the value of the NR motion degree signal mds and the mixing ratio at the time of mixing in the processing by the mixing unit 801. The mixing ratio is 0 to 1 (100%), and the NR motion degree signal mds ( The relationship between the horizontal axis), the noise removal filter output signal dnr, and the mixing ratio (vertical axis) of the center field or frame signal r2 is shown. When the NR motion level signal mds is MD1 (for example, 4) and indicates complete motion, the ratio of the signal r2 in the center field or frame to 1 is set to 1 so that noise removal is not performed, and the noise removal filter output signal dnr. The ratio occupied by 0 is assumed to be 0. When the degree of motion is minimum (when the NR motion degree signal mds is 0), the ratio of the noise removal filter output signal dnr filtered by a predetermined noise removal effect is set to 1, and the signal of the center field or frame The ratio of r2 is assumed to be 0.

  When the NR motion degree signal mds is larger than 0 and smaller than MD1, the ratio of the central field or frame to the signal r2 is reduced so that the noise reduction effect is reduced as the NR motion degree signal mds becomes larger. Let

For example, the mixing ratio of the center field or frame signal r2 and the noise removal filter output signal dnr is (MD1-mds) vs. mds,
Rnr = r2 * (MD1-mds) / MD1 + dnr * mds / MD1
Rnr given by is output. Here, MD1 is the maximum value that mds can take as described above.

  By doing in this way, the real line signal Rnr after the noise removal in which the signal after the filtering process is mixed stepwise according to the degree of motion is obtained.

  As described above, the acyclic noise removal processing means 80 detects the motion of the signal between 4 fields (every other field) or 4 frames (consecutive), and generates the NR motion degree signal mds obtained as a result of the detection. Based on this, the noise removal filter output signal dnr and the central field or frame signal r2 are mixed by the mixing unit 801, and the filtered signal is mixed in stages according to the degree of motion. This prevents the blurring and deterioration of the edges and contours of the moving parts, and the trailing and afterimages of the moving parts are within the range of 4 fields or 4 frames, thereby reducing the trailing and afterimages. In addition, noise components having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed.

  The actual line noise removing unit 21 configured by the non-cyclic type noise removing processing unit 80 has been described above, but the first interpolation line noise removing unit 221 configured by the non-cyclic type noise removing processing unit 80 operates in the same manner. The second interpolation line noise removing unit 222 operates in the same manner, and outputs the first interpolation line signal Isnr from which the acyclic noise is removed.

  The non-cyclic noise removal processing means 81 shown in FIG. 16 is the same as the non-cyclic noise removal processing means 71 of FIG. A non-cyclic noise removal process is performed using i1 as a central field signal, and the configuration and operation other than those relating to the mixing unit 811 and the filter coefficient setting unit 821 are the same as those shown in FIG. The detailed explanation is omitted.

The non-cyclic type noise removal processing unit 81 is configured similarly to the non-cyclic type noise removal processing unit 80 shown in FIG. 15, and includes a noise removal filter unit 700, an NR motion detection unit 721, and a filter coefficient setting unit 821. And a mixing means 811.
The noise removal filter means 700 and the NR motion detection means 721 are configured in the same manner as those in the non-cyclic noise removal processing means 70 of FIG. 5 and the non-cyclic noise removal processing means 71 of FIG.

That is, the noise removal filter unit 700 performs filter processing using the first to fourth NR processing signals i1 to i4, and outputs a signal after noise removal (indicated by a symbol dnrb).
The NR motion detection means 721 detects the motion of the signal based on the four input NR processing signals i1 to i4 with respect to the signal in the center field, and outputs an NR motion degree signal mdsb.

The filter coefficient setting unit 821 sets filter coefficients kg1 to kg4.
The mixing unit 811 applies the degree of NR motion from the NR motion detection unit 721 to the noise removal filter output signal dnrb output from the addition unit 705 in the noise removal filter unit 700 and the signal i1 in the center field that has not been subjected to noise removal. A mixing process is performed at a mixing ratio corresponding to the value of the signal mdsb, and a signal Itnr after noise removal is output.

  The first to fourth NR processing signals i1 to i4 shown in FIG. 3 are input to the non-cyclic noise removal processing means 81 via the input terminals 751 to 754, and the first NR processing signal i1 is included. Is used as a signal of the center field CFR, and acyclic noise removal processing is performed.

  The non-cyclic noise removing unit 81 detects the motion of the signal based on the four NR processing signals i1 to i4 input in the NR motion detecting unit 721, and based on the NR motion degree signal mdsb as a detection result, The noise removal filter output signal dnrb and the signal i1 of the center field that has not been subjected to noise removal are mixed to obtain a signal Itnr after noise removal. In this way, the acyclic noise removal process in which the strength of noise removal is controlled is performed.

  Here, the acyclic noise removing means 81 is a filtering process using frame correlation between four fields (every other field), like the acyclic noise removing means 80 of FIG. The detection means 721 detects the movement of the signal during the four fields, and the noise reduction filter output signal dnrb and the central field signal i1 are mixed by the mixing means 811 based on the NR movement degree signal mdsb, thereby increasing the degree of movement. Accordingly, a signal after acyclic noise removal in which the signal after filtering is mixed step by step is output. As a result, the noise removal effect by the filtering process can be changed step by step according to the degree of movement, and the edge of the moving part, the blurring and deterioration of the contour are prevented, and the filtering process is performed with a signal of four fields. A signal with components removed is obtained.

The NR motion detection means 721 receives the first to fourth NR processing signals i1 to i4.
In the NR motion detection means 721, the difference (frame difference) between the first NR processing signal i1 as the center field signal and the second, third, and fourth NR processing signals i2, i3, i4. And the obtained frame difference is combined to detect a motion in the video signal, and an NR motion degree signal mdsb indicating the degree of motion as a result of motion detection is output.

  The NR motion degree signal mdsb is generated in the same manner as the NR motion degree signal mds in FIG. 5, the NR motion degree signal mdsb in FIG. 10, and the NR motion degree signal mds in FIG. The value between indicates the degree of movement. The NR motion degree signal mdsb from the NR motion detection means 721 is a mixing means 811 as a motion degree signal indicating the degree of motion for controlling the mixing ratio between the noise removal filter output signal dnrb and the signal that has not been subjected to noise removal. Sent to.

  Next, in the filter coefficient setting means 821 in the acyclic noise removal processing means 81 shown in FIG. 16, the filter coefficients kg1 to kg4 are set in accordance with, for example, an operation by a user or the like, similarly to the filter coefficient setting means 820 in FIG. Set and output. Instead of setting according to the operation by the user or the like, the noise amount in the input video signal may be detected, and the filter coefficient may be automatically set based on the detected noise amount.

  The filter coefficients kg1 to kg4 are set so as to obtain a noise removal effect having a preset strength.

  As described above with reference to FIG. 15, the filter coefficients filter coefficients kg1 to kg4 set in the filter coefficient setting means 821 also set the coefficient of the filter for removing noise in the same manner as the filter coefficients in the filter coefficient generation means 710 and 711. The strength (effect) of noise removal is determined by assigning the values of the filter coefficients kg1 to kg4 to the four NR processing signals, and the filter coefficients kg1 to kg4 are each 0 to 1 and the sum is 1. It is determined to be.

  However, the filter coefficient setting means 820 in FIG. 15 defines the filter coefficients kf1 to kf4 using the second NR processing signal as the center field or frame signal, whereas the filter coefficient setting means 821 The values of filter coefficients kg1 to kg4 are determined by using one NR processing signal i1 as a central field signal. When the filter coefficients kg1 to kg4 are the same value 1/4 (= 0.25), noise components having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, if the value of the coefficient kg1 for the signal i1, which is the central field, is large, kg1 = 1, and the other filter coefficients kg2, kg3, kg4 are zero, that is, (kg1, kg2, kg3, kg4) = (1 , 0, 0, 0), the signals in the preceding and succeeding fields are multiplied by 0, the noise cyclic value becomes zero, and the signal is output as it is, and noise removal is not performed. .

  The effect of noise removal is to increase the value of the coefficient kg2 for the signal i1, which is the central field, and decrease the values of the other coefficients (for example, (kg1, kg2, kg3, kg4) = (7/16, 4 / 16, 3/16, 2/16)), or the number of fields used for filtering (the number of NR processing signals with coefficients other than 0) is reduced to 3 (for example, ((kg1 , Kg2, kg3, kg4) = (1/3, 1/3, 1/3, 0)) or reduced to 2 (for example, (kg1, kg2, kg3, kg4) = (0.5, 0. (5, 0, 0)), it can be made weaker by changing the filter coefficient.

  That is, in order to maximize the noise removal effect, the filter coefficients kg1 to kg4 are set to the same value ((kg1, kg2, kg3, kg4) = (0.25, 0.25, 0.25, 0.25). )). When setting the effect of noise removal to a medium strength or weak, for example, (kg1, kg2, kg3, kg4) = (7/16, 4/16, 3/16, 2/16) or , (1/3, 1/3, 1/3, 0), (0.5, 0.5, 0, 0) are set to set the strength of noise removal.

  The filter coefficients kg1 to kg4 generated by the filter coefficient setting unit 821 are supplied to the coefficient multiplication units 701, 702, 703, and 704 in FIG.

  The first to fourth NR processing signals i1 to i4 and the filter coefficients kg1 to kg4 are input to the noise removal filter unit 700. Similarly to FIG. 15, the noise removal filter means 700 performs the filter processing using the first to fourth NR processing signals i1 to i4 and outputs the noise removal filter output signal dnrb.

  First to fourth NR processing signals i1 to i4 and filter coefficients kg1 to kg4 are input to coefficient multiplication means 701, 702, 703, and 704 in the noise removal filter means 700, and filter coefficients are input to the input signal. Multiplying operations are performed to obtain multiplication results i1 × kg1, i2 × kg2, i3 × kg3, i4 × kg4.

  Coefficient multiplication results i1 × kg1, i2 × kg2, i3 × kg3, and i4 × kg4 output from the coefficient multipliers 701, 702, 703, and 704 are input to the adding unit 705. The adding means 705 adds the four input multiplication results and outputs a signal obtained after the addition as a noise removal filter output signal dnrb.

  The configuration of the coefficient multiplying means 701, 702, 703, 704 and the adding means 705 up to this point is the same as that shown in FIGS. Note that the filter coefficients kg1 to kg4 are coefficients for obtaining a noise removal effect having a preset strength, and the filter output signal dnrb output from the adding means 705 has a predetermined strength (with respect to the signal i1 in the center field). This is a signal that has been subjected to noise removal filter processing with a predetermined noise removal effect.

  Next, to the mixing unit 811, the noise removal filter output signal dnrb output from the addition unit 705 in the noise removal filter unit 700, the center field signal i 1, and the NR motion degree output from the NR motion detection unit 721. The signal mdsb is input. The mixing unit 811 performs a mixing process on the noise removal filter output signal dnrb and the central field signal i1 at a mixing ratio according to the value of the NR motion degree signal mdsb, and outputs a signal Itnr after noise removal.

  In the mixing unit 811, the mixing ratio between the noise removal filter output signal dnrb and the central field signal i 1 is determined from the value of the NR motion degree signal mdsb from the NR motion detection unit 721, and the noise removal filter output signal dnrb is determined based on this mixing ratio. The configuration for mixing the signal i1 in the center field is the same as the configuration described in FIGS. 15 and 17, and detailed description thereof will be omitted. However, as the NR motion degree signal mdsb increases, the effect of noise removal increases. The ratio of the central field signal i1 is changed so as to be small, and the noise-removed filter output signal dnrb and the central-field signal i1 that has not been subjected to noise removal are mixed to obtain a signal after noise removal. Therefore, when setting the filter coefficient by the filter coefficient setting means 821, It is possible to increase the effect of removing's stepwise change the effect of removing filtering noise in accordance with the degree of movement, the edge of the moving portion, the contour of blurring and deterioration can be prevented.

  As described above, the non-cyclic noise removal processing means 81 detects the signal motion between four fields (every other field), and the mixing means 811 removes the noise based on the NR motion degree signal mdsb obtained as a result of the detection. By mixing the filter output signal dnrb and the signal i1 of the center field, a signal after acyclic noise removal in which the signal after filter processing is mixed stepwise according to the degree of motion is output. Edge and contour blur and deterioration can be prevented, and the tail and afterimage of the moving part can be kept within the range of 4 fields, and the tail and afterimage can be reduced. Also, noise components having no frame correlation over 4 fields (8 vertical periods) can be removed.

  Then, the second interpolation line signal Itnr from which the non-cyclic noise is removed is output from the second interpolation line noise removing unit 222 configured by the non-cyclic noise removal processing unit 81.

  As described above, in the acyclic noise removal means 80 and 81 shown in FIGS. 15 and 16, the signal subjected to the noise removal filter processing and the signal not subjected to noise removal are represented by the NR motion degree signals mds and mdsb. A mixing process is performed at a mixing ratio according to the value, and a signal after acyclic noise removal is obtained in which the filtered signal is mixed step by step according to the degree of movement, and the edges and contours of the moving part are blurred. And a non-recursive noise-removed signal that has been filtered using a signal of 4 fields (every other field) or 4 frames (consecutive), and preventing the degradation. Trailing and afterimages can be accommodated within the range of 4 fields or 4 frames, and tailing and afterimages can be reduced, and they extend over 4 fields (8 vertical periods) or 4 frames (4 vertical periods). A noise component having no frame correlation can be removed, and the noise component is removed as the real line signal Rnr after noise removal, the first interpolation line signal Isnr, and the second interpolation line signal Itnr. A signal can be obtained.

  In the second embodiment, the configuration and operation other than those relating to the non-cyclic noise removing means 80 and 81 constituting the non-cyclic noise removing processing means 20 are the same as those in the first embodiment, and the detailed description thereof is as follows. Although omitted, when an interlace signal is input, three-dimensional IP conversion is performed on the signal after noise removal, the delayed signal used in the noise removal processing, and the signal in the current field, and a progressive signal is output.

The overall operation of the video signal processing apparatus according to the second embodiment is the same as that described with reference to FIG. 11 regarding the first embodiment. The video signal processing apparatus of the second embodiment is different from the first embodiment in the operation of the noise removing means 21, 221 and 222.
Hereinafter, the operation of the acyclic noise removing unit 80 constituting the noise removing unit 21 will be described with reference to FIG. The operations of the noise removing units 221 and 222 are the same.

  In FIG. 18, steps having the same contents as those in FIG. The first to fourth NR processing signals r1 to r4 are input to the non-cyclic type noise removal processing means 80, and the second NR processing signal r2 is the signal of the center field or frame (S11).

  The acyclic noise removing means 80 performs filtering using the four input NR processing signals r1 to r4 to obtain a noise removal filter output signal dnrb, and performs noise removal with the noise removal filter output signal dnrb. The non-recursive noise removal process is performed by performing a mixing process on the signal r2 of the center field or frame that has not been performed at a mixing ratio according to the value of the NR motion degree signal mds from the NR motion detecting unit 720.

  The NR motion detection means 720 receives the first to fourth NR processing signals r1 to r4, and in step S12 to step S15 in FIG. 18, the second NR processing signal r2 as a center field or frame signal. And a difference (frame difference) between the first, third, and fourth NR processing signals r1, r3, r4, and combining the obtained frame differences to detect a motion in the video signal, An NR motion level signal mds indicating the level of motion as a result of motion detection is generated.

  Since this operation is the same as the operation described in steps S12 to S15 in FIG. 12, the detailed description thereof is omitted, but the second NR processing signal r2 as the center field or frame signal and the first, A frame difference between the third and fourth NR processing signals r1, r3, r4 is calculated, and three frame differences dif1, dif2, dif3 are detected (S12).

  Next, the synthesizing unit 734 synthesizes the frame differences dif1, dif2, and dif3 (for example, obtains the maximum value) to obtain the motion difference signal Daf (S13).

  The edge adjustment unit 740 detects an edge portion in the signal r2 of the center field or frame, and the pixel determined to form an edge has a larger motion than the motion difference signal Daf obtained by the frame difference detection unit 730. In addition to performing adjustment so as to represent the pixel, the pixel that has not been determined to constitute an edge is averaged with surrounding pixels to remove isolated points and obtain an adjusted motion difference signal Dfe (S14).

  The conversion unit 750 generates an NR motion degree signal mds from the motion difference signal Dfe subjected to the edge adjustment and averaging from the edge adjustment unit 740 (S15).

  That is, the input motion difference signal Dfe is subjected to conversion processing by amplitude limitation or the like, for example, the degree of motion is set to the (MD1 + 1) stage (for example, MD1 = 4), and it is detected that there is complete motion. In this case, the value is MD1 (for example, MD1 = 4, complete motion pixel), and it is detected that the image is a complete still image (this includes the case where the difference is detected as a noise component rather than due to motion). NR motion level signal mds indicating the degree of motion is generated with a value between 0 and MD1 (for example, MD1 = 4) as zero (mds = 0, complete still pixel).

As in steps S12 to S15, motion detection is performed by the NR motion detection unit 720, and the NR motion degree signal mds obtained as a result of the detection is sent to the mixing unit 801.
The filter coefficient setting unit 820 sets the filter coefficients kf1 to kf4 (S26).
The filter coefficients kf1 to kf4 are used to set the coefficients of a filter that performs acyclic noise removal. Each of the filter coefficients kf1 to kf4 is determined to be 0 to 1 and the sum is 1, and filter coefficients for four NR processing signals. The strength (effect) of noise removal is determined by the allocation of the values of kf1 to kf4.

  The values of the filter coefficients kf1 to kf4 are determined using the second NR processing signal r2 as a central field or frame signal. When the filter coefficients kf1 to kf4 are the same value 1/4 (= 0.25), the noise component having no frame correlation on the time axis can be most strongly removed, and the noise removal effect is maximized. . On the other hand, if the value of the coefficient kf2 for the signal r2 of the center field or frame is large, kf2 = 1, and the coefficients kf1, kf3, kf4 for other fields or frames are zero, In other words, 0 is multiplied so that the cyclic value of noise is zero, and the input signal is output as it is, and noise removal is not performed.

  The effect of noise removal is to increase the value of the coefficient kf2 for the signal r2 in the center field or frame and decrease the values of the other coefficients (for example, (kf1, kf2, kf3, kf4) = (2/8, (3/8, 2/3, 1/3)) or the number of fields or frames used for filtering (the number of NR processing signals with coefficients other than 0) is reduced to 3 (for example, ((Kf1, kf2, kf3, kf4) = (1/3, 1/3, 1/3, 0)) or reduced to 2 (for example, (kf1, kf2, kf3, kf4) = (0.5 , 0.5, 0, 0)), and can be made weaker by changing the filter coefficient.

In the noise removal filter unit 700, four NR processing signals (signals of pixels at the same position in four fields or four frames) are obtained by the first to fourth NR processing signals r1 to r4 and the filter coefficients kf1 to kf4. Is used to obtain a noise removal filter output signal dnr (S27).
The filter coefficients kf1 to kf4 are coefficients with which a noise removal effect having a preset strength can be obtained. Therefore, the obtained noise removal filter output signal dnr is denoised with a predetermined strength (predetermined noise removal effect). Signal.

That is, the mixing ratio between the noise removal filter output signal dnr and the central field or frame signal r2 is determined from the value of the NR motion degree signal mds, and the noise removal filter output signal dnr and the central field or frame signal r2 are determined based on this mixing ratio. Mixing is performed (S28), and a noise-removed signal Rnr obtained by the mixing is output (S18). As the NR motion degree signal mds becomes larger, the mixing ratio is changed so that the ratio of the signal r2 in the center field or frame becomes smaller so that the effect of noise removal becomes smaller (see FIG. 17).
Since the signal after noise removal is obtained by mixing the noise removal filter output signal dnr and the signal r2 of the center field or frame that has not been subjected to noise removal, when the filter coefficient is set by the filter coefficient setting means 820, the noise The removal effect can be increased, and the noise removal effect by the filtering process can be changed stepwise according to the degree of movement, and blurring and deterioration of the edges and contours of the moving part can be prevented.

  Although the case where the non-recursive noise removing unit 80 used as the real line noise removing unit 21 generates the real line signal Rnr after noise removal has been described above, the acyclic type used as the first interpolation line noise removing unit 221 has been described. The processing for obtaining the first interpolation line signal Isnr by the mold noise removing means 80 is similarly performed.

The processing in the case of obtaining the second interpolation line signal Itnr by the acyclic noise removal processing means 81 used as the second interpolation line noise removal means 222 is generally the same. However, when obtaining the second interpolation line signal Itnr, the first NR processing signal i1 is used as the signal of the center field. That is, the difference (frame difference) between the first NR processing signal i1 as the center field signal and the second to fourth NR processing signals i2, i3, i4 in the difference calculation in step S12. In step S14, the edge adjustment means 741 detects the edge portion of the first NR processing signal i1, and in step S26, filtering using the first NR processing signal as a central field signal is performed. This is different in that filter coefficients kg1 to kg4 having values for use are used.
The other basic noise removal processing operations described in steps S11 to S18 are the same as those in FIG.

  As described above, in the non-cyclic noise removal processing means 80 and 81 in the non-cyclic noise removal processing means 20 of the second embodiment, the motion between 4 fields (every other field) or 4 frames (continuous) is detected. Based on the NR motion degree signal mds obtained as a result of the detection, the noise removal filter output signal dnr after filtering by the signal of 4 fields or 4 frames and the signal of the center field where noise removal is not performed are mixed. A non-cyclic noise removal with controlled noise removal strength is performed by obtaining a signal after acyclic noise removal in which the signal after filtering is mixed stepwise depending on the degree. Then, three-dimensional IP conversion is performed on the signal after the non-cyclic noise removal, the delayed signal used in the non-cyclic noise removal processing, and the signal in the current field, so that a progressive signal from which noise is removed can be obtained. ing. Therefore, the field delay signal used in the noise removal process and the three-dimensional IP conversion process is obtained by a common field memory, and 4 fields (8 vertical periods) or 4 frames (4 vertical periods) without significantly increasing the field memory. ) Over which non-cyclic noise removal is performed to remove noise components having no frame correlation, and a three-dimensional scanning line interpolation process can be performed using the noise-removed signal. . Further, in the non-cyclic noise removal processing, it is possible to obtain a filtered signal after the non-cyclic noise removal processing by preventing the edge and outline of the moving portion from being blurred and deteriorated. Without significantly increasing the circuit scale, it is possible to reduce the adverse effects of tailing and afterimages and to perform good noise removal, and to obtain a progressive signal from which noise has been removed.

Embodiment 3 FIG.
In the video signal processing apparatus 1 according to the first embodiment, when the interlace signal is input, the interpolation field is the field of the two-field delay signal d2f, and a signal delayed by two fields is output as the output video signal Do0, and when the progressive signal is input. Is output as an output video signal Do0 corresponding to the 1-field delay signal d1f. However, as in the video signal processing device 3 shown in FIG. 19, the interpolation field at the time of interlace signal input is the field of the 3-field delay signal d3f. The output video signal Do0 may correspond to the three-field delay signal d3f, and the output video signal Do0 when the progressive signal is input may be configured to correspond to the one-field delay signal d1f. For each signal in the acyclic noise removal processing means 20 As the noise removal filter (corresponding to the real line noise removal means 21, the first interpolation line noise removal means 221 in FIG. 5, and the second interpolation line noise removal means 222 in FIG. 15), those having the same internal configuration are used. be able to.

  FIG. 19 is a block diagram showing the configuration of the video signal processing apparatus according to the third embodiment of the present invention (that is, an apparatus that can implement the video signal processing method according to the third embodiment). Similarly to the above, a non-recursive noise removal process for removing a noise component by performing filtering using a signal of a plurality of fields or frames, for example, a signal of 4 fields or 4 frames, is performed. In FIG. 19, the same or corresponding components as those in the first embodiment shown in FIG.

  In FIG. 19, the video signal processing apparatus 3 according to the third embodiment uses the non-cyclic noise removal processing means 20 in the configuration of the video signal processing apparatus 1 (see FIG. 1) according to the first embodiment described above. It is replaced by means 200, the interpolation field at the time of interlace signal input is the field of the three-field delay signal d3f, the output video signal Do0 at the time of input of the progressive signal corresponds to the one-field delay signal d1f, A three-dimensional IP conversion process is performed. Configurations and operations other than those relating to the acyclic noise removal processing unit 200 are the same as those shown in the first embodiment, and detailed description thereof is omitted.

  The non-cyclic noise removal processing means 200 in FIG. 19 is located in the field or frame and the front and back of the signal of the central field or frame, similarly to the non-cyclic noise removal processing means 20 shown in FIG. Noise is removed by filtering using a signal of a total of 4 fields or 4 frames of the other 3 fields at the same position in the screen, and a signal from which noise has been removed is output. When an interlace signal is input, a noise removal process is performed on the actual line signal in the interpolation field to output a noise-removed actual line signal, and the interpolation line in the field temporally adjacent to the interpolation field is output. This is an example of outputting a line signal for interpolation with noise removal applied to the signal. If configured as shown in FIG. 20.

  In FIG. 20, this acyclic noise removal processing means 200 includes first to fourth switching means 204 to 207, actual line noise removal means 21, and interpolation line noise removal means 23.

The actual line noise removing means 21 is the same as that shown in FIG.
The switching units 204 to 207 select a signal based on the switching signal pf from the switching signal generation unit 18.
The interpolation line noise removing means 23 removes noise by performing filter processing on the interpolation line signal using the signals of the pixels at the same position in the four fields.
The interpolation line noise removal unit 23 includes a first interpolation line noise removal unit 221 and a second interpolation line noise removal unit 223.
The first interpolation line noise removing unit 221 is the same as that shown in FIG.
The second interpolation line noise removing unit 223 is generally the same as the second interpolation line noise removing unit 222 shown in FIG. 2, but there are differences as described below.

  In FIG. 19, first to seventh field memories 11 to 17 sequentially output video signals after being delayed by 1 field, and output 1-field delayed signals d1f to 7-field delayed signals d7f.

  The switching signal generator 18 generates and outputs a switching signal pf indicating whether the input signal is an interlace signal or a progressive signal.

In the three-dimensional IP conversion processing means 65, when the input signal is an interlace signal, a scanning line interpolation signal is generated by intra-field interpolation processing and inter-field interpolation processing, and the interlace signal is converted into a progressive signal by three-dimensional scanning line conversion. A progressive signal Ips from which noise has been removed is output. In the three-dimensional IP conversion processing unit 65, the intra-field interpolation processing unit 30 performs intra-field interpolation by arithmetic processing such as filter processing on the three-field delay signal d3f which is an interpolation field to generate an interpolation signal Imv. The motion detection means 40 in the motion detection means 66 detects a motion based on the current field signal Di0, the 2-field delay signal d2f, and the 4-field delay signal d4f and outputs a motion detection signal mdt.
Further, the telecine detection means 50 detects whether it is a telecine video signal based on the current field signal Di0, the two-field delay signal d2f, and the one-field delay signal d1f, and outputs a telecine detection signal tcif. The configuration so far is the same as that shown in the first embodiment, and a detailed description thereof will be omitted.

  In the example shown in FIG. 19, the motion detection means 40 in the three-dimensional IP conversion processing means 65 receives the current field signal Di0, the two-field delay signal d2f, and the four-field delay signal d4f based on them. Although motion detection is performed, in consideration of the position of the interpolation line signal in the interpolation field, the input to the motion detection means 40 is a 2-field delay signal d2f, a 4-field delay signal d4f, and a 6-field delay signal d6f. Or a 1-field delay signal d1f, a 3-field delay signal d3f, and a 5-field delay signal d5f. The difference information is obtained between these fields, and the motion of the video signal between 1 frame and 2 frames. May be configured to detect the field or frame of the signal input to the motion detection means 40. Only by changing the beam may be carried out motion detection in the same configuration. Similarly, the input to the telecine detection means 50 can also be a 1-field delay signal d1f, a 2-field delay signal d2f, and a 3-field delay signal d3f.

The acyclic noise removal processing means 200 shown in FIG. 20 receives the current field or frame signal Di0, the delayed signals d1f to d7f, and the switching signal pf from the switching signal generating means 18.
The acyclic noise removal processing means 200 is located in the center field or frame signal of the current field or frame signal Di0 and the delayed signals d1f to d7f, and before or after the field or frame, and the pixel is within the screen. Noise is removed by performing filtering using signals of other 3 fields or frames in the same position, and a total of 4 fields or 4 frames, and after noise removal, the real line signal Rnr and the first interpolation line signal Isnr , And a second interpolation line signal Itnr.

  The switching signal pf from the switching signal generating means 18 switches between a signal used for noise removal of the actual line signal and a signal used for noise removal for the progressive signal. When a progressive signal is input, since no interpolation of the scanning line is necessary, noise removal is not performed on the interpolation line signal, and the generation of the first interpolation line signal Isnr and the second interpolation line signal Itnr is not performed.

Here, in the non-cyclic type noise removal processing means 200, when the interlace signal is input, the signal of the center field is the signal of the interpolation field, that is, the three-field delay signal d3f for the real line of the interpolation field. For the lines, the actual line signals (interpolation line signals) of the field temporally adjacent to the interpolation field, that is, the 4-field delay signal d4f and the 2-field delay signal d2f.
On the other hand, when a progressive signal is input, the 1-field delay signal d1f becomes the signal of the center frame.

  The switching means 204 to 207 in the acyclic noise removal processing means 200 shown in FIG. 20 receives the current field or frame signal Di0 and the delayed signals d1f to d3f, d5f, and d7f, and responds to the switching signal pf. The selected signal is output as the first to fourth NR processing signals r1 to r4.

Specifically, the switching means 204 inputs the current field or frame signal Di0 and the one-field delay signal d1f, receives the one-field delay signal d1f when the interlace signal is input, and receives the signal Di0 of the current frame when the progressive signal is input. This is selected and supplied to the actual line noise removing means 21 as the first NR processing signal r1.
The switching means 205 receives the 1-field delay signal d1f and the 3-field delay signal d3f, selects the 3-field delay signal d3f when the interlace signal is input, and selects the 1-field delay signal d1f when the progressive signal is input. The signal is supplied to the actual line noise removing unit 21 as the NR processing signal r2 (as the center field or frame signal).
The switching means 206 inputs the 2 field delay signal d2f and the 5 field delay signal d5f, selects the 5 field delay signal d5f when the interlace signal is input, and selects the 2 field delay signal d2f when the progressive signal is input. The signal is supplied to the actual line noise removing unit 21 as the NR processing signal r3.
The switching means 207 receives the 3 field delay signal d3f and the 7 field delay signal d7f as input, selects the 7 field delay signal d7f when the interlace signal is input, and selects the 3 field delay signal d3f when the progressive signal is input. The signal is supplied to the actual line noise removing unit 21 as the NR processing signal r4.

  The first to fourth NR processing signals r <b> 1 to r <b> 4 are used by the actual line noise removing unit 21 for noise removal processing for the second NR processing signal r <b> 2.

  The current field signal Di0 and the delayed signals d2f, d4f, and d6f are interpolated line noise removal means 22 as first to fourth NR processing signals i1 to i4 used for noise removal processing for the interpolation line signal. To be supplied.

  FIG. 21 shows the temporal relationship between fields and the positional relationship in the vertical scanning direction of the delayed signals d1f to d7f when an interlace signal is input. The earlier field in time is shown on the left side, and the pixels in each field or frame are vertically scanned in order from top to bottom. The positional relationship when a progressive signal is input is as shown in FIG.

  When the interlace signal shown in FIG. 21 is input, the signal in the interpolation field is a three-field delay signal d3f, and therefore is supplied as the second NR processing signal r2 in the noise removal of the actual line in the interpolation field. The three-field delay signal d3f is the signal of the center field CFR. The 3 field delay signal 3df, the 1 field delay signal d1f, the 5 field delay signal d5f, and the 7 field delay signal d7f (all of which are signals of pixels at the same position as the 3 field delay signal d3f) are respectively The first, third, and fourth NR processing signals r1, r3, and r4 are obtained.

  In the noise removal of the interpolation line in the interpolation field, the signals of the interpolation field adjacent before and after the interpolation field, that is, the 4-field delay signal d4f and the 2-field delay signal d2f are the signals of the center fields CFIs and CFIt, respectively. 2 NR processing signal i2 and first NR processing signal i1. The 6-field delay signal d6f and the 8-field delay signal d8f become the third NR processing signal i3 and the fourth NR processing signal i4, respectively.

  When the progressive signal shown in FIG. 4 is input, the output video signal Do0 corresponds to the delayed signal d1f from the first field memory 11 as described in the first embodiment. The 1-field delay signal d1f supplied as the second NR processing signal r2 is the signal of the center frame, and the current frame signal Di0, the 2 field delay signal d2f, and the 3 field delay signal d3f are respectively the first, third, The fourth NR processing signals r1, r3, r4 are obtained.

  In FIG. 20, the real line noise removing unit 21 in the non-cyclic type noise removing processing unit 200 performs noise removal by performing filtering using the input first to fourth NR processing signals r1 to r4. The real line signal Rnr is output after noise removal.

  The actual line noise removing unit 21 performs non-cyclic noise removal processing using the second NR processing signal r2 of the four input NR processing signals r1 to r4 as a central field or frame signal. The non-cyclic noise removal processing means 70 described with reference to FIG. 5 with respect to the first embodiment can be used. Although the detailed description of the configuration is omitted, the actual line noise removing unit 21 detects the motion of the signal between four fields or four frames by adopting the configuration of FIG. 5, and the NR motion obtained as a result of the detection. Based on the degree signal mds, a filter coefficient of a noise removal filter is generated, and a signal after acyclic noise removal in which the strength of noise removal is controlled is output. This prevents blurring and deterioration of the edges and contours of the moving part, allows the trailing part and afterimage of the moving part to fall within the range of 4 fields or 4 frames, and reduces the trailing and afterimage. A noise component having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed.

  As the real line noise removing means 21, the non-cyclic type noise removing processing means 80 described with reference to FIG. 15 in connection with the second embodiment may be used. Signal motion between them, and based on the NR motion degree signal mds obtained as a result of the detection, the noise removal filter output signal dnr and the central field or frame signal r2 are mixed, and stepwise according to the degree of motion. The filtered signal is mixed with the non-recursive noise-removed signal to prevent blurring and deterioration of the edges and contours of the moving parts, and the trailing and afterimages of the moving parts. Within 4 fields or 4 frames to reduce tailing and afterimages, and 4 fields (8 vertical periods) or 4 frames (4 vertical periods). It can be removed without noise component frame correlation over.

  The interpolated line noise removal means 23 in the acyclic noise removal processing means 200 shown in FIG. 20 includes the current field signal Di0 and the delayed signals d2f, d4f, and d6f as the first to fourth NR processing signals. Input as i1 to i4.

  The interpolation line noise removal unit 23 includes a first interpolation line noise removal unit 221 and a second interpolation line noise removal unit 223. When the interlace signal is input, the first to fourth NR processing signals are input. Noise is removed by filtering using i1 to i4, and a first interpolation line signal Isnr and a second interpolation line signal Itnr are output.

  The first interpolation line signal Isnr and the second interpolation line signal Itnr are derived from an interpolation line signal (a signal in a temporally adjacent field of the interpolation field) used for generating the interpolation line signal at the time of three-dimensional IP conversion. This is a signal from which noise has been removed.

As the first interpolation line noise removing means 221, one having the same internal configuration as shown in FIG. 5 can be used. However, the relationship between each input terminal and the signal supplied to the input terminal is different. That is, as shown in FIG. 22, the first, second, third, and fourth input terminals connected to the first, second, third, and fourth coefficient multipliers 701, 702, 703, and 704, respectively. The fourth, third, second, and first NR processing signals i4, i3, i2, and i1 are supplied to 751, 752, 753, and 754, respectively. The filter coefficient generation unit 710 generates the same coefficient as the filter coefficient generation unit 710 of FIG. As a result, the noise removal filter unit 700 performs acyclic noise removal using the third NR processing signal i3 supplied to the second input terminal 752 as a central field.
As described above, by changing the relationship between the input terminals 751, 752, 753, and 754 and the signals i4, i3, i2, and i1 supplied to the respective input terminals, the coefficient generator can be used as shown in FIG. 21) can be used.
As described above, the current field signal Di0 and the delayed signals d2f, d4f, and d6f are supplied as the first, second, third, and fourth NR processing signals i1, i2, i3, and i4, respectively. Therefore, the first interpolation line signal Isnr output from the first interpolation line noise removal means 221 configured by the non-cyclic noise removal processing means 70 shown in FIG. A 4-field delayed signal d4f (third NR processing signal i3 in FIG. 21) which is an adjacent field is a signal from which noise has been removed using a signal in the center field.

  The second interpolation line noise removing unit 223 is generally the same as the second interpolation line noise removing unit 222 of FIG. 2, but the input field signal Di0, the two field delay signal d2f, the four field delay signal d4f, and the six field. The first, second, third, and fourth NRs of the delayed signal d6f (in the relative positional relationship as they are, that is, without reversing the order as in the first interpolation line noise removing unit 221 in FIG. 20). The processing signals i1, i2, i3, i4 are received, and the second NR processing signal i2 is used as a signal in the center field to perform noise removal processing. The second interpolation line signal Itnr output from the second interpolation line noise removing means 223 removes noise by using a 2-field delay signal d2f, which is a signal of an adjacent field temporally after the interpolation field, as a central field signal. Signal.

The second interpolation line noise removing unit 223 can be configured by the non-cyclic type noise removing processing unit 70 shown in FIG.
As described above, the second interpolation line noise removing unit 223 includes the non-cyclic type noise removing processing unit 70 shown in FIG. 5, and as described above, the first interpolation line noise removing unit 221 is also configured as shown in FIG. (As described with reference to FIG. 22, only the relationship between the input terminal and the signal supplied from the outside is different). Therefore, the first interpolation line noise removal unit 221 and the second interpolation line noise removal unit 223 are configured by circuits having the same internal configuration, and input signals (input wirings) supplied to the input terminals are different. You can do it.

  Even when the first interpolation line noise removing unit 221 and the second interpolation line noise removing unit 223 are used, the motion of the signal between four fields (every other field) or four frames (consecutive) is detected. Then, based on the NR motion degree signal mds obtained as a result of the detection, a filter coefficient can be generated and a signal after acyclic noise removal in which the strength of noise removal is controlled can be output. This prevents blurring and deterioration of the edges and contours of the moving part, allows the trailing part and afterimage of the moving part to fall within the range of 4 fields or 4 frames, and reduces the trailing and afterimage. A noise component having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed.

  Note that the first interpolation line noise removal unit 221 and the second interpolation line noise removal unit 223 may be configured by the acyclic noise removal processing unit 80 described with reference to FIG. 15 with respect to the second embodiment. In this case as well, signal motion between 4 fields (every other field) or 4 frames (successive) is detected, and the noise removal filter output signal dnr is based on the NR motion degree signal mds obtained as a result of the detection. And the signal r2 of the central field or frame can be output in a non-cyclic type noise-removed signal in which the filtered signal is mixed in stages according to the degree of motion, Prevents blurring and deterioration of edges and contours of moving parts, and tails and afterimages of moving parts are kept within the range of 4 fields or 4 frames. It is possible to reduce the afterimage, also can be removed without noise component frame correlation over four fields (8 vertical period) or 4 frames (4 vertical interval).

  As described above, the actual line noise removing unit 21 and the interpolated line noise removing unit 23 in the non-cyclic type noise removing processing unit 200 shown in FIG. 20 have four fields (every other field) or four (continuous). A non-recursive noise-removed signal that has been filtered using the frame signal can be obtained, and the tailing and afterimage of the moving part are stored within the range of 4 fields or 4 frames, reducing the tailing and afterimage. In addition, noise components having no frame correlation over 4 fields (8 vertical periods) or 4 frames (4 vertical periods) can be removed, and the real line signal Rnr after noise removal is used for the first interpolation. As the line signal Isnr and the second interpolation line signal Itnr, a signal from which noise is removed can be output.

  Hereinafter, in FIG. 19, the noise-removed real line signal Rnr, the first interpolation line signal Isnr, and the second interpolation line signal Itnr that have been subjected to noise removal processing from the acyclic noise removal processing unit 200, From the intra-field interpolation signal Imv from the interpolation processing unit 30, the motion detection signal mdt from the motion detection unit 40, and the telecine detection signal tcif from the telecine detection unit 50, the interpolation signal generation unit 60 and the time axis conversion unit 61 The configuration for performing the three-dimensional IP conversion and outputting the output video signal Do0 from the switching means 62 is the same as the configuration of FIG. 1 of the first embodiment, and thus detailed description is omitted, but from the switching means 62, In both cases of interlace signal input and progressive signal input, both noise reduction processing and 3D IP conversion processing are used. Using the delayed signal, it is possible to reduce the harmful effect tailing and afterimages can be obtained as a progressive signal, those acyclic noise removal process has good noise removal effect is applied.

  In the video signal processing apparatus 3 according to the third embodiment, the non-cyclic noise removal processing by the non-cyclic noise removal processing means 200 and the three-dimensional IP conversion by the three-dimensional IP conversion processing means 65 are performed by delaying the interpolation field by three fields. The first embodiment will be described with reference to FIGS. 11 and 12 except that the field of the signal d3f is used and the signal in the center field is the second NR processing signal, or the second embodiment. The operation is the same as that described with reference to FIG.

  As described above, according to the video signal processing device 3 of the third embodiment, the interpolation field when the interlace signal is input is the field of the 3-field delay signal d3f, and the output video signal Do0 when the progressive signal is input is the 1-field delay signal d1f. The non-cyclic noise removal processing means 200 performs filtering using a 4-field or 4-frame signal in the non-cyclic noise removal processing means 200 and performs non-cyclic noise removal processing. The three-dimensional IP conversion processing unit 65 performs three-dimensional IP conversion using the delayed signal used in the noise removal processing and the current field signal, and outputs a progressive signal. Therefore, the delay signal used in the noise removal process and the three-dimensional IP conversion process is obtained by a common field memory, and 4 fields (8 vertical periods) or 4 frames (4 vertical periods) without significantly increasing the field memory. A three-dimensional scanning line interpolation process can be performed using a signal that has been subjected to acyclic noise removal that removes noise components having no frame correlation. Also, in the non-cyclic noise removal processing, the edge of the moving part, the blurring and deterioration of the contour are prevented, and the filtered signal after the non-cyclic noise removal processing is obtained. Without significantly increasing the scale, it is possible to reduce the adverse effects of tailing and afterimage and to obtain a progressive signal from which noise has been successfully removed.

  Further, as the real line noise removing unit 21, the first interpolation line noise removing unit 221, and the second interpolation line noise removing unit 223 in the non-cyclic type noise removing processing unit 200, the non-cyclic type noise removing unit having the same internal configuration is used. Since the device can be used, it is possible to simplify the design of the apparatus, and in each signal, it is possible to obtain a signal after removal of acyclic noise in which the strength of noise removal subjected to similar processing is controlled. .

In the video signal processing apparatus 3 according to the third embodiment, the interpolation field at the time of interlace signal input is the field of the three-field delay signal d3f, and the non-recursive noise removal processing means 200 uses the current field of the positional relationship shown in FIG. In the video signal processing apparatus 1 according to the first embodiment, when the interlace signal is input, the two-field delay signal d2f is used. As a signal of the interpolation field, the non-cyclic noise removal processing means 20 performs the non-cyclic noise removal processing in the positional relationship shown in FIG.
However, the first embodiment is modified as in the third embodiment, and the non-recursive noise removal processing is performed using the positional relationship signal shown in FIG. 23 in which the 1-field delay signal d1f is the interpolation field signal. The configuration to perform the three-dimensional IP conversion is to perform the non-circular noise removal processing using the positional relationship signal shown in FIG. 24 using the current field signal Di0 as the interpolation field signal, and perform the three-dimensional IP conversion. In the non-cyclic type noise removal processing means 200, 20, the input signal is switched to the corresponding signal, and the non-cyclic type noise removal process is performed according to the position of the central field or frame as shown in FIG. 15) or the configuration shown in FIG. 10 (or FIG. 16), and the same effects as those of the third embodiment can be obtained.

  In FIG. 19, in the interpolated motion detecting means 66 in the three-dimensional IP conversion processing means 65, the motion detecting means 40 detects the motion of the signal, the telecine detecting means 50 detects the telecine, and the interpolated signal generating means 60. In the above description, the interpolation signal generated by the three-dimensional processing is generated by the motion adaptation process or the telecine interpolation. However, the interpolation signal generation by the three-dimensional IP conversion processing unit 65 may be only the motion adaptation process, and the interpolation motion detection unit 66 detects the motion. You may comprise only the means 40. FIG. In this case, if the three-dimensional IP conversion is performed by using the first interpolation line signal Isnr in the non-cyclic type noise removal processing unit 200, the second interpolation line signal Itnr becomes unnecessary, and the second interpolation The line signal Itnr can be omitted.

Further, in the video signal processing apparatus 3 according to the third embodiment, it is assumed that the output video signal Do0 when the progressive signal is input corresponds to the one-field delay signal d1f, and the actual signal in the acyclic noise removal processing unit 200 is used. As shown in FIG. 4, the line noise removal means 21 performs noise removal processing by filtering with a signal of four frames that include the current frame and are continuous (at intervals of one frame).
However, instead of doing the above, as shown in FIG. 13, the two-field delay signal d2f is a signal of the center frame CFR, includes the center frame, and is positioned every two frames (at intervals of two frames). You may comprise so that a noise removal process may be performed by the filter process using the signal of the flame | frame.

Embodiment 4 FIG.
In the video signal processing apparatuses 1 and 3 in the first to third embodiments, the acyclic noise removal processing is performed using the input video signal and the field delay signal, and the noise-removed signal, the field delay signal, and the noise are removed. The configuration is such that three-dimensional IP conversion is performed based on the current field signal before removal. However, the signal obtained by performing the cyclic noise removal process on the input video signal and the field delay signal obtained by delaying the signal are used in the non-cyclic noise removal process and the three-dimensional IP conversion. It can also be configured.

FIG. 25 is a block diagram showing a configuration of a video signal processing apparatus according to Embodiment 4 of the present invention (that is, an apparatus capable of executing the video signal processing method of Embodiment 4).
The video signal processing apparatus shown in the figure performs a cyclic noise removal process, delays the signal after the cyclic noise removal process, and uses the signal after the cyclic noise removal process. A non-recursive noise removal process and three-dimensional IP conversion are performed using a signal of four fields or four frames of the signal obtained by field delay.
In FIG. 25, the same or corresponding components as those in the first embodiment shown in FIG.

  The video signal processing device 4 shown in FIG. 25 is the same as the video signal processing device 1 shown in FIG. 1, but differs in that a cyclic noise removal processing means 10 is added. An input video signal Di0 is sequentially input to this cyclic noise removing means 10, and noise components are removed by the difference between frames.

The input video signal Di0 is supplied to the cyclic noise removal means 10, and the signal dl0f after the cyclic noise removal processing output from the cyclic noise removal processing means 10 is input to the first field memory 11, and the cyclic noise is inputted. The signal dl0f after the removal processing is sequentially delayed one field at a time in the first to seventh field memories 11 to 17, and field delay signals d1f to d7f are output. Of the field delay signals d1f to d7f, the two-field delay signal d2f from the second field memory 12 is fed back to the cyclic noise removal means 10 and used for cyclic noise removal.
A part or all of the signal dl0f after the cyclic noise removal processing and the field delay signals d1f to d7f are processed by the non-cyclic noise removal processing in the non-cyclic noise removal processing means 20 and the three-dimensional IP in the three-dimensional IP conversion processing means 65. Used in conversion.
The configuration and operation other than those relating to the cyclic noise removing means 10 are the same as those shown in FIG. 1 described above, and detailed description thereof will be omitted. Therefore, the effect of noise removal can be further strengthened.

  As described above, the cyclic noise removing means 10 receives the signal Di0 of the current field or frame and the two-field delayed signal d2f from the second field memory 12 (the output signal dl0f after noise removal is divided into two fields (two vertical periods). ) Delayed signal) is input. In the cyclic noise removing means 10, a difference value (frame difference) Diff between the signal Di0 of the current field or frame and the two-field delay signal d2f is obtained, and whether the difference corresponds to “motion” from the frame difference Diff. The motion between frames is detected by determining whether it corresponds to “noise”, and a motion degree signal is obtained. A cyclic coefficient for a noise component is obtained based on the motion degree signal, and noise removal processing is performed from the difference and the cyclic coefficient. And output a signal dl0f after cyclic noise elimination. The cyclic noise-removed signal dl0f corresponds to the field of the input video signal Di0 and is sent to the first field memory 11.

  For example, as shown in FIG. 26, the cyclic noise removing unit 10 includes a subtracting unit 101, an amplitude limiting unit 102, an interframe motion detecting unit 103, a cyclic coefficient generating unit 104, a multiplying unit 105, and an arithmetic operation. Means 106.

In FIG. 26, the current field signal Di0 and the two-field delay signal d2f are input to the subtracting means 101, and the current field signal Di0 is subtracted from the two-field delay signal d2f, and the difference (frame difference) Diff between them is obtained. Desired.
The frame difference Diff includes a “motion” component and a “noise” component in the video signal. When the frame difference Diff is 0, the frame difference Diff is a portion that is completely stationary and has no noise component. Or, if there is a noise component, the value becomes large.

  The amplitude limiting unit 102 limits the amplitude within a predetermined value (for example, within ± dTh) with respect to the frame difference Diff from the subtracting unit 101, and the amplitude-limited difference value Dfn in the pixel of the input signal Di0. Output as a noise component.

The inter-frame motion detection unit 103 receives the frame difference Diff from the subtraction unit 101 and detects the inter-frame motion in the video signal from the frame difference Diff.
Specifically, the movement is divided into a plurality of grades or stages, for example, (FMD1 + 1) according to the degree, and a movement degree signal fmds indicating which stage the movement belongs to is output.
For example, when the motion is at the maximum level, that is, when the video signal is “complete motion”, the first predetermined value, for example, fmds = FMD1, and when the motion level is at the minimum level, that is, “complete still image. In the case of “Yes” (this includes the case where the difference is due to a noise component), the second predetermined value, for example, fmds = 0, and the value of fmds increases as the degree of motion increases. FMD1 is determined to be “8”, for example. However, it may be larger than FMD1 or 8, and may be smaller than 8.
The motion degree signal fmds generated by the inter-frame motion detection unit 103 is sent to the cyclic coefficient generation unit 104.

  The cyclic coefficient generation unit 104 receives the motion degree signal fmds from the inter-frame motion detection unit 103. The cyclic coefficient generation unit 104 generates a cyclic coefficient Km for noise corresponding to the motion degree signal fmds. The cyclic coefficient Km sets a cyclic value of noise when performing cyclic noise removal, and is set to 0 ≦ Km ≦ 1. The closer the value of the cyclic coefficient Km is to 1, the greater the cyclic value of noise for noise removal, and the higher the noise removal effect. On the other hand, when the value of the cyclic coefficient Km is zero (Km = 0), the cyclic value of noise is zero and noise removal is not performed.

That is, when the motion level signal fmds = FMD1 (for example, 8) and complete motion is indicated, the cyclic coefficient Km = 0 is set, and the motion level is the minimum level, and it is detected that the motion image is a complete still image (to this). Includes a case where the difference is detected as a noise component), that is, when fmds = 0, the degree of motion (the value of the motion degree signal fmds) increases as Km0 = Kmax (Kmax ≦ 1). Accordingly, the value of the cyclic coefficient Km0 is changed to be small.
The cyclic coefficient generation unit 104 includes, for example, a unit that performs multiplication with the value of the motion level signal fmds, and a ROM (Read Only Memory) that inputs the value of the motion level signal fmds as an address.

  The cyclic coefficient Km from the cyclic coefficient generation means 104 is supplied to the multiplication means 105 of the cyclic noise removal means 10 in FIG.

  To the multiplication means 105, the difference value Dfn from the amplitude limiting means 102 and the cyclic coefficient Km from the coefficient generation means 104 are sent. The multiplication means 105 multiplies the difference value Dfn by a cyclic coefficient Km (performs the calculation of Km × Dfn) to obtain a noise cyclic amount Nd. The obtained noise circulation amount Nd is supplied to the computing means 106.

  The computing means 106 receives the noise circulation amount Nd from the multiplication means 105 and the current field signal Di0 inputted thereto. The computing means 106 subtracts the noise circulation amount Nd from the signal Di0 in the current field. That is, if the sign of the noise circulation amount Nd from the multiplication means 105 is positive, the absolute value is subtracted, and if the sign is negative, the absolute value is added.

  Through the above processing, noise in the video signal is removed, and a signal dl0f after cyclic noise removal is obtained. As described above, the cyclic coefficient Km becomes a value corresponding to the value of the motion degree signal fmds, and therefore, the noise removal effect is changed stepwise according to the degree of motion.

  Next, refer to FIG. 25 again. The cyclic noise-removed signal dl0f output from the cyclic noise removal means 10 is a field signal of the input video signal Di0 and is sent to the first field memory 11.

  The first to seventh field memories 11 to 17 sequentially delay the video signal by one field (one vertical period) and output a one-field delay signal d1f to a seven-field delay signal d7f. The signal dl0f after cyclic noise removal is input to the first field memory 11, and the signal dl0f after cyclic noise removal is sequentially delayed by the first to seventh field memories 11-17.

Hereinafter, for field delay processing, non-cyclic noise removal processing in the non-cyclic noise removal processing means 20, and three-dimensional IP conversion processing in the three-dimensional IP conversion processing means 65, refer to FIGS. The detailed description is omitted.
In Embodiment 4, at least one of the cyclic noise removing means 10, the non-cyclic noise removing processing means 20, and the three-dimensional IP processing means 65 is used for both interlace signal input and progressive signal input. A delay signal output from the same field memory is used.

  Next, in the video signal processing apparatus 4 according to the fourth embodiment, the cyclic noise removal by the cyclic noise removal means 10, the non-cyclic noise removal by the non-cyclic noise removal processing means 20, and the three-dimensional IP conversion processing means 65. 3D IP conversion will be described with reference to the flowchart of FIG. 27, the same reference numerals as those in FIG. 11 indicate the same or similar processing.

  First, the operation when the input video signal to the video signal processing device 4 is an interlace signal will be described. The 2-field delay signal d2f is an interpolation field signal in IP conversion.

  When the input video signal Di0 is input to the video signal processing device 4 (S100), the input video signal Di0 is sent to the cyclic noise removing means 10 together with the 2-field delay signal d2f. The cyclic noise removing means 10 obtains a frame difference Diff between the current field signal Di0 and the two-field delayed signal d2f, obtains a motion level signal fmds of the motion from the frame difference Diff, and generates noise based on the motion level signal fmds. A cyclic coefficient Km for the component is obtained, noise removal processing based on the difference Diff and the cyclic coefficient Km is performed, and a signal dl0f after cyclic noise removal is obtained (S101).

The cyclic noise-removed signal dl0f is sent to the first field memory 11 in the same field as the input video signal Di0 (a signal in which the input video signal Di0 is not delayed), and the first to seventh field memories 11 to 11 are sent. The field delay signals d1f to d7f are sequentially delayed by 17 in one field (one vertical period) (S1). Since the signal dl0f after cyclic noise removal is field-delayed, the delayed signal is also a signal subjected to cyclic noise removal processing.
As in the first embodiment, the delay signal used in the acyclic noise removal process and the delay signal used in the three-dimensional IP conversion are obtained from a common field memory.

  Hereinafter, the operations in steps S2 to S9 in FIG. 27, that is, the non-cyclic noise removal processing by the non-cyclic noise removal processing means 20, and the three-dimensional IP conversion by the three-dimensional IP conversion processing means 65, the signal in the current field is cyclic. Since it is the same as that described with reference to FIG. 11 with respect to the first embodiment except for the signal after noise removal dl0f, detailed description thereof will be omitted.

When the progressive signal is input, the cyclic noise removal processing in steps S100, S101, and S1 and the processing for obtaining the delayed signal of the signal dl0f after cyclic noise removal are the same as when the interlace signal is input. On the other hand, since it is not necessary to perform IP conversion, the processes in steps S3, S4, S5, S6, and S7 are omitted.
A field delay process (S1) of the input signal and a non-cyclic noise removal process (S2) by the non-cyclic noise removal processing means 20 are performed, and the real line signal Rnr after noise removal is selected by the switching means 62 and the output video signal Do0. Is obtained.

  When a progressive signal is input, the cyclic noise removal process may be performed using the 1-field delay signal d1f instead of the 2-field delay signal d2f. This is because in the case of a progressive signal, a frame difference can be obtained based on the one-field delay signal d1f and the input video signal Di0.

  The operation in the acyclic noise removal processing means 20 is the same as the operation described with reference to FIG. 12 for the first embodiment and the operation described with reference to FIG. 18 for the second embodiment. Detailed description thereof is omitted.

  As described above, according to the video signal processing device 4 of the fourth embodiment, the cyclic noise removal processing is performed, the post-cyclic noise removal signal dl0f is field-delayed, and the field delay signals d1f to d7f are obtained. Of these, the two-field delayed signal d2f is used by the cyclic noise removing means 10. Further, the non-cyclic noise removal processing means 20 performs non-cyclic noise removal processing using the four-field or four-frame signal among the cyclic noise-removed signal dl0f and the delayed signals d1f to d7f. Three-dimensional IP conversion is performed in the dimensional IP conversion processing means 65. As described above, the field delay signal used in the noise removal process and the three-dimensional IP conversion process is obtained by a common field memory, and a progressive signal from which noise is satisfactorily removed can be obtained without significantly increasing the field memory. it can.

  Further, the cyclic noise-removed signal dl0f subjected to the cyclic noise removal processing is subjected to field delay, and the non-cyclic noise removal processing and the three-dimensional IP conversion are performed using the cyclic noise-removed signal dl0f and the field delay signal. Therefore, in the cyclic noise removal process, a minute noise is removed, a noise is removed by setting a cyclic coefficient to such an extent that there is no harmful effect such as tailing or afterimage of the moving part, and then the non-cyclic noise removal process and the three-dimensional noise removal process are performed. IP conversion can be performed, and noise removal can be performed more effectively.

  In the video signal processing apparatus 4 according to the fourth embodiment, the cyclic noise removal processing means 10 is configured as shown in FIG. 26, and a signal obtained by delaying the current field or frame signal and the cyclic noise-removed signal dl0f by two fields. The cyclic coefficient generation unit 104 generates a cyclic coefficient for noise corresponding to the motion detection result based on the motion between frames. The configuration of the cyclic noise removal processing unit 10 is shown in FIG. For example, it may be configured as a cyclic noise removing unit 10b shown in FIG.

  28, the same reference numerals are given to the same or corresponding components as those in FIG. The cyclic noise removing unit 10b shown in FIG. 28 includes a cyclic coefficient setting unit 107 instead of the cyclic coefficient generation unit 104 in FIG. 26, sets a predetermined cyclic coefficient Km by the cyclic coefficient setting unit 107, and further includes a frame. In accordance with the motion degree signal fmds from the inter-motion detection means 103, a mixing means 108 for mixing the cyclic noise-removed signal lnf output from the arithmetic means 106 and the input signal Di0 is provided, whereby the motion degree signal The signal Inf after noise removal with a predetermined cyclic coefficient Km and the signal (input video signal) Di0 not subjected to noise removal are mixed at a mixing ratio according to fmds, and after the cyclic noise removal controlled by the motion is performed. The signal dl0fb is obtained. Since the configuration and operation other than the cyclic coefficient setting unit 107 and the mixing unit 108 are the same as those of the cyclic noise removal processing unit 10 described above, detailed description thereof will be omitted.

  In FIG. 28, the cyclic coefficient setting unit 107 in the cyclic noise removing unit 10b sets a predetermined cyclic coefficient Km. The frame difference Diff is obtained from the subtracting unit 101 and the amplitude is limited by the amplitude limiting unit 102. The value Dfn is obtained, the motion degree signal fmds is output from the inter-frame motion detection means 103, the multiplication amount 105 is used to determine the cyclic amount Nd by Km × Dfn, and the arithmetic means 106 calculates the noise cyclic amount Nd from the current field signal Di0. Is subtracted to remove noise in the video signal, and a signal Inf after cyclic noise removal is obtained. The steps so far are the same as those in FIG.

  The mixing means 108 mixes the cyclic noise-removed signal lnf and the input video signal Di0 not subjected to noise removal at a mixing ratio according to the motion degree signal fmds. Here, the cyclic noise-removed signal lnf is a signal that has been subjected to noise removal processing with a cyclic coefficient Km that produces a predetermined noise removal effect, and corresponds to the value of the motion degree signal fmds from the inter-frame motion detection means 103. The mixing ratio is changed to mix the signal lnf after cyclic noise removal and the input video signal Di0 not subjected to noise removal. As the degree of motion (the value of the motion degree signal fmds) increases, the signal dl0fb after removing cyclic noise considering the motion is obtained by changing the ratio of the signal lnf after cyclic noise removal to be smaller. It will be.

For example, the mixing ratio of the input video signal Di0 and the cyclic noise-removed signal Inf is set to fmds: (FMD1-fmds),
dl0fb = Di0 × fmds / FMD1 + Inf × (FMD1−fmds) / FMD1
To obtain d10fb.
Here, FMD1 is the maximum value that can be taken by the motion degree signal fmds. Further, the minimum value that the motion degree signal fmds can take is zero.

  Further, in the video signal processing apparatus 4 in the fourth embodiment, the input to the first field memory 11 in the configuration of the video signal processing apparatus 1 (see FIG. 1) in the first embodiment is received from the cyclic noise removing means 10. In the configuration of the video signal processing apparatus 3 of Embodiment 3 (see FIG. 19), the input to the first field memory 11 in the configuration of the video signal processing apparatus 3 of Embodiment 3 (see FIG. 19) is the cyclic noise removal means 10. It is also possible to configure so as to be a signal after a cyclic noise removal process. Further, instead of the non-cyclic type noise removal processing unit 20, the non-cyclic type noise removal processing unit 200 in the second embodiment may be used. In any case, if the input to the first field memory 11 can be configured to be a signal after cyclic noise removal processing from the cyclic noise removal means 10, the same effect as in the fourth embodiment can be obtained.

Modified example of the first to fourth embodiments.
In the second to fourth embodiments, as described with respect to the first embodiment, filter processing using more than four fields or frames may be performed in the acyclic noise removal processing. If filtering is performed using a larger number of fields or frames, the calculation will be performed between frames that are more distant from each other in time, and noise components that have no frame correlation in the time axis will be removed, thus increasing the noise removal effect. Can do.

  Furthermore, a filter process using three fields or frames, or two fields or frames may be performed. In short, it is only necessary to perform a filter process using two or more, that is, a plurality of fields or frames, of the current field signal and the delayed signal.

  As described above, in the first to fourth embodiments, each component of the video signal processing apparatuses 1, 3, and 4 is described as being configured by hardware. However, these are programmed by software processing, that is, programmed. You may comprise so that it may implement | achieve by a computer.

Embodiment 5 FIG.
In the first to fourth embodiments, the video signal processing apparatus and the apparatus capable of performing the video signal processing method have been described. However, the present invention performs processing by removing noise from the input video signal and displays it with high image quality. It can also be applied to a video signal display device. In the following, a TV broadcast signal, a video signal input from a recording / playback device such as a DVD or VTR, a TV broadcast receiver, or the like is processed, and the video signal processing devices 1, 3, and 4 of the first to fourth embodiments are treated as one of them. A video signal display device that is provided as a unit and displays a video signal will be described as a fifth embodiment.

FIG. 29 is a block diagram showing an example of a video signal display apparatus according to Embodiment 5 of the present invention. The illustrated video signal display device 5 includes an input terminal 111, an input signal processing unit 112, a video signal processing unit 110, a display processing unit 113, and a display unit 114, and is a video that is a progressive signal from which noise has been removed. The signal is displayed on the display means 114.
The video signal processing means 110 is configured like the video signal processing device 1 of the first embodiment or the second embodiment, the video signal processing device 3 of the second embodiment, or the video signal processing device 4 of the fourth embodiment. Is done.

  The input terminal 111 receives a TV broadcast signal, a signal from a recording / playback device such as a DVD or VTR, a TV broadcast receiver, or the like. A signal input to the input terminal 111 is sent to the input signal processing means 112.

  The input signal processing means 112 performs input signal processing on a TV broadcast signal, a signal from a recording / playback device such as a DVD or VTR, a TV broadcast receiver, etc. For example, when an analog signal is input Input signal processing such as conversion of a signal into a digital signal and separation of a synchronization signal is performed. When MPEG data is received, processing such as decoding of the MPEG data is performed. Then, the video signal subjected to the input signal processing is output to the video signal processing means 110. Here, the video signal subjected to the input signal processing may be an interlace signal or a progressive signal.

  The video signal processing unit 110 performs noise removal processing and three-dimensional IP conversion, and outputs a progressive signal after noise removal to the display processing unit 113.

  The display processing unit 113 performs signal processing for converting the progressive signal from which noise has been removed into a display signal such as scaling processing, and outputs the display signal to the display unit 114 as a display signal. The display unit 114 displays an image based on the display signal from the display processing unit 113.

  As described above, according to the video signal display device of the fifth embodiment, since the video signal processing device of the first to fourth embodiments is used, the circuit scale such as a field memory as field delay means is greatly increased. High-quality video that uses progressive signals that are well noise-removed by preventing noise and blurring and deterioration of moving parts, reducing noise, reducing the effects of tailing and afterimages, etc. Can be displayed.

It is a block diagram which shows an example of a structure of the video signal processing apparatus 1 by Embodiment 1 of this invention. FIG. 3 is a block diagram showing a configuration example of an acyclic noise removal processing unit 20 in the video signal processing device 1 according to the first embodiment. FIG. 7 is a diagram illustrating an example of a positional relationship between a signal Di0 in a current field and a delay signal in a time direction and a vertical scanning direction when an interlace signal is input in the video signal processing device 1 according to the first embodiment. It is a figure which shows an example of the positional relationship of the signal Di0 of the present field at the time of a progressive signal input in the video signal processing apparatus 1 by Embodiment 1, and a delay signal in the time direction and a vertical scanning direction. The non-cyclic noise removal processing means 70 used as the real line noise removal means 21 and the first interpolation line noise removal means 221 in the non-cyclic noise removal processing means 20 in the video signal processing apparatus 1 according to Embodiment 1 is shown. It is a block diagram. It is a block diagram which shows one structural example of the difference detection means 731 in the acyclic noise removal process means 70 shown by FIG. It is a figure which shows an example of the input-output characteristic of the motion sensitivity conversion means 737 shown by FIG. It is a block diagram which shows one structural example of the edge part adjustment means 740 in the non-cyclic type noise removal process means 70 shown by FIG. FIG. 5 conceptually shows the relationship between the value of the NR motion degree signal mds in the acyclic noise removal processing means 70 shown in FIG. 5 and the noise removal strength of the filter coefficient generated by the filter coefficient generation means 710 in accordance therewith. FIG. 7 is a block diagram showing a non-cyclic noise removal processing means 71 used as the second interpolation line noise removal means 222 in the non-cyclic noise removal processing means 20 in the video signal processing apparatus 1 according to Embodiment 1. FIG. 3 is a flowchart showing an operation of the video signal processing apparatus 1 according to the first embodiment. 4 is a flowchart showing an operation of acyclic noise removal processing in the video signal processing apparatus 1 according to Embodiment 1. It is a figure which shows the other example of the positional relationship of the signal Di0 of the present field at the time of a progressive signal input in the video signal processor 1 by Embodiment 1 and a delay signal in the time direction and a vertical scanning direction. It is a figure which shows the further another example of the positional relationship of the signal direction Di0 of the present field at the time of a progressive signal input in the video signal processing apparatus 1 by Embodiment 1, and a delay signal in the time direction and a vertical scanning direction. FIG. 9 is a block diagram showing non-cyclic noise removal processing means 80 used as actual line noise removal means 21 and first interpolation line noise removal means 221 in non-cyclic noise removal processing means 20 in Embodiment 2 of the present invention. is there. FIG. 10 is a block diagram showing a non-cyclic noise removal processing means 81 used as the second interpolation line noise removal means 222 in the non-cyclic noise removal processing means 20 in the second embodiment. FIG. 16 is a diagram illustrating an example of a relationship between a value of an NR motion degree signal mds and a mixing ratio in a mixing unit 801 in the non-cyclic noise removal processing unit 80 illustrated in FIG. 15. 10 is a flowchart showing an operation of acyclic noise removal processing in the second embodiment. It is a block diagram which shows an example of a structure of the video signal processing apparatus 3 by Embodiment 3 of this invention. FIG. 10 is a block diagram showing a configuration example of an acyclic noise removal processing unit 200 in the video signal processing device 3 according to Embodiment 3. It is a figure which shows an example of the positional relationship of the signal direction Di0 of the present field at the time of the input of the interlace signal and delay signal in the video signal processing apparatus 3 by Embodiment 3 in a time direction and a vertical scanning direction. FIG. 10 is a block diagram showing an example of the configuration of first interpolation line noise removal means 221 in acyclic noise removal processing means 20 in video signal processing apparatus 3 according to Embodiment 3. FIG. 10 is a diagram illustrating another example of the positional relationship between the current field signal Di0 and the delay signal in the time direction and the vertical scanning direction when the interlace signal is input in the video signal processing device 3 according to the third embodiment. FIG. 16 is a diagram illustrating still another example of the positional relationship between the current field signal Di0 and the delay signal in the time direction and the vertical scanning direction when the interlace signal is input in the video signal processing device 3 according to the third embodiment. It is a block diagram which shows an example of a structure of the video signal processing apparatus 4 by Embodiment 4 of this invention. FIG. 10 is a block diagram illustrating a configuration example of a cyclic noise removal processing unit 10 in a video signal processing device 4 according to a fourth embodiment. 14 is a flowchart showing an operation of the video signal processing device 4 according to the fourth embodiment. FIG. 26 is a block diagram illustrating another configuration example of the cyclic noise removal processing unit 10 b that can be used in place of the cyclic noise removal processing unit 10 in the video signal processing device 4 of FIG. 25. It is a block diagram which shows an example of a structure of the video signal display apparatus by Embodiment 5 of this invention.

Explanation of symbols

1,3,4 video signal processing apparatus, 11-17 field memory, 18 switching signal generation means, 20,200 non-cyclic noise removal processing means, 65 three-dimensional IP conversion processing means, 66 interpolation motion detection means, 30 in field Interpolation processing means, 40 motion detection means, 50 telecine detection means, 60 interpolation signal generation means, 61 time axis conversion means, 62 switching means, 21 real line noise removal means, 22 interpolation line noise removal means, 221 first interpolation line Noise removal means 222, 223 Second interpolation line noise removal means 201-203, 204-207 switching means 70, 71 Acyclic noise removal processing means 700 Noise removal filter means 710 711 Filter coefficient generation means 720, 721 NR motion detection means, 701-704 coefficient multiplication Means, 705 addition means, 730 frame difference detection means, 740, 741 edge adjustment means, 750 conversion means, 731 to 733 difference detection means, 734 composition means, 735 difference calculation means, 736 absolute value calculation means, 737 motion sensitivity conversion Means, 742 averaging means, 743 adjustment means, 744 edge determination means, 745 selection means, 80, 81 acyclic noise removal processing means, 820, 821 filter coefficient setting means, 801, 811 mixing means, 10, 10b cyclic type Noise removal processing means, 101 subtraction means, 102 amplitude limiting means, 103 inter-frame motion detection means, 104 cyclic coefficient generation means, 105 multiplication means, 106 arithmetic means, 107 cyclic coefficient setting means, 108 mixing means, 5 video signal display device 110 video signal processing means, 111 Power terminal, 112 an input signal processing unit, 113 display processing unit, 114 display unit.

Claims (11)

A video signal processing apparatus for removing a noise component having no correlation between fields or frames of a video signal, and obtaining a sequentially scanned video signal from which the noise component has been removed,
A plurality of delays by delaying an input video signal or a video signal of a current field or frame composed of a video signal obtained by cyclic noise removal processing in units of field periods, which are periods divided by a vertical synchronization signal Field delay means for outputting a video signal;
A video signal of a plurality of fields or frames composed of the video signal of the current field or frame and the plurality of delayed video signals is input, and a motion is obtained from a difference between signals of the plurality of fields or frames. A non-cyclic noise removal processing means for performing a filtering process controlled by movement to remove noise and generating a signal after acyclic noise removal;
When the input video signal is an interlaced scanning video signal,
A scanning line interpolation signal is generated from one or a combination of intra-field interpolation processing and inter-field interpolation processing from the video signals of the plurality of fields or frames and the signal after the non-cyclic noise removal, and interlaced scanning is performed. Three-dimensional scanning line conversion processing means for converting a video signal into a progressive scanning video signal and generating a progressive scanning video signal from which noise components have been removed;
Depending on whether the input video signal is an interlaced scanning video signal or a progressive scanning video signal, either the signal after the non-cyclic noise removal or the progressive scanning video signal from the three-dimensional scanning line interpolation processing means is selected. And a switching means for outputting as a sequentially scanned video signal from which noise components have been removed. The acyclic noise removal processing means is:
A motion is obtained from a difference between a signal of a pixel of interest in each of one or more central fields or frames to be denoised, and a signal of each central field or frame and the adjacent field or frame before or after the central field or frame. And the pixel of interest in each central field or frame, and the pixel of interest in one or more fields or frames in which the central field or frame and pixels are at the same position, controlled by the obtained movement. Noise is removed by filtering using a signal of a pixel at the same position, and the signal after the acyclic noise removal is output,
Receiving the actual scanning line signal of the interlaced scanning video signal or the progressive scanning video signal;
When an actual scanning line signal of the interlaced scanning video signal is received, the interpolation field is the central field, the interpolation target pixel in the central field is the target pixel, and the non-cyclic noise removal is performed on the signal of the target pixel. To output the actual scanning line signal after acyclic noise removal,
When the progressive scanning video signal is received, the actual scanning is performed by performing the non-cyclic noise removal on the signal of the pixel of interest in the center frame and outputting the actual scanning line signal after the non-cyclic noise removal. Line noise removal means;
An actual scanning line signal of the interlaced scanning video signal is received, and a field temporally adjacent to the interpolation field (hereinafter referred to as “interpolation field”) is set as the central field and is in the interpolation field, and the interpolation is performed. The pixel at the same position as the pixel to be interpolated in the field is the target pixel, the non-cyclic noise is removed from the signal of the target pixel, and the interpolation scanning line signal after the non-cyclic noise is removed is output. Interpolated scanning line noise removing means,
The actual scanning line signal after removal of the non-cyclic type noise and the interpolation scanning line signal after removal of the non-cyclic type noise are output as signals after the removal of the non-cyclic type noise. The video signal processing apparatus described. Each of the actual scanning line noise removing unit and the interpolating scanning line noise removing unit includes:
A signal of a pixel at the same position in a plurality of fields or frames is received, and a difference between a signal of the target pixel in the central field or frame and a signal of a pixel at the same position in the other field or frame is calculated. A noise removal motion detection means for calculating, detecting the motion of the signal from the difference, and outputting the motion detection result;
Filter coefficient generation means for generating and outputting a filter coefficient for each field or frame signal according to the motion detection result from the noise removal motion detection means;
Noise removal filter means for performing filter processing using filter coefficients from the filter coefficient generation means using signals of pixels at the same position in the plurality of fields or frames,
The video signal processing apparatus according to claim 2, wherein the signal after the filter processing from the noise removal filter means is output as the signal after the acyclic noise removal. Each of the actual scanning line noise removing unit and the interpolating scanning line noise removing unit receives a signal of a pixel at the same position in a plurality of fields or frames, and receives a signal of the central field or frame and another field or frame. Calculating a difference with the signal, detecting a motion of the signal from the difference, and outputting a motion detection result;
Filter coefficient setting means for setting and outputting a filter coefficient for each field or frame signal giving a predetermined noise removal effect;
Noise removal filter means for performing a filtering process using a filter coefficient from the filter coefficient setting means using the signals of pixels at the same position in the plurality of fields or frames, and outputting a noise removal filter output signal;
The noise removal filter output signal and the signal of the central field or frame that has not been noise-removed are mixed by the mixing ratio according to the motion detection result from the noise removal motion detection means, and the mixed signal is processed by the non-cyclic type. The video signal processing apparatus according to claim 2, further comprising a mixing unit that outputs the signal after noise removal. The noise removal motion detection means receives signals of pixels at the same position in the plurality of fields or frames, obtains a difference between the signal of the center field or frame and a signal of another field or frame as a frame difference, A frame difference detection means for detecting and outputting a motion difference signal representing movement from the obtained frame difference;
The edge portion in the signal of the center field or frame is detected, the motion difference signal of the edge portion is adjusted with respect to the motion difference signal from the frame difference detecting means, and the motion difference of pixels around the pixel of interest is adjusted. Edge adjustment means for performing an averaging process with the signal and outputting the adjusted motion difference signal;
Conversion means for converting the adjusted motion difference signal into a motion degree signal indicating the degree of motion, and outputting as a motion detection result;
The video signal processing apparatus according to claim 3 or 4, wherein the motion detection result is output as a motion between signals of a plurality of input fields or frames. The edge adjustment means is
Determining whether or not the pixel of interest constitutes an edge, and determining and outputting an adjustment value for the motion difference signal, and a determination result;
Averaging means for receiving the motion difference signal, obtaining an average value of the motion difference signal for the pixel of interest and the motion difference signal for the surrounding pixels, and outputting an averaged motion difference signal;
Adjusting means for adjusting the motion difference signal output from the frame difference detection means by the adjustment value, and outputting an adjusted motion difference signal;
The averaged motion difference signal and the adjusted motion difference signal are received by the determination result output from the edge determination means,
When it is determined by the determination result that the target pixel constitutes an edge, the adjusted motion difference signal is selected,
A selection means for selecting the averaged motion difference signal when it is not determined that the target pixel constitutes an edge;
The video signal processing apparatus according to claim 5, wherein an output of the selection unit is output as a motion difference signal after edge portion adjustment. The non-cyclic noise removal processing means, as the signal after the non-cyclic noise removal, an actual scanning line signal after the non-cyclic noise removal and an interpolation scanning line signal after the non-cyclic noise removal. Output,
The three-dimensional scanning line conversion processing means includes:
Intra-field interpolation processing means for generating an interpolated scanning line signal based on signals in the interpolated field of the interlaced scanning video signal;
Interpolation motion detecting means for receiving the video signal of the current field and the delayed video signal, detecting a motion of the signal from a difference between fields of the video signal, and generating an interpolation switching signal based on the detection result;
Based on the interpolation switching signal output from the interpolation motion detection means, the interpolation scanning line signal after the non-cyclic noise removal processing means from the non-cyclic noise removal processing means and the interpolation scanning line from the intra-field interpolation processing means Interpolation signal generating means for performing three-dimensional scanning line processing on the signal and generating an interpolated scanning line signal subjected to intra-field interpolation processing and inter-field interpolation processing;
The interpolated scanning line signal output from the interpolation signal generating means is fitted between the corresponding scanning lines of the real scanning line signal after the non-cyclic noise removal from the non-cyclic noise removing processing means, and double-speed converted, A time axis converting means for converting to a progressive signal and outputting it,
3. The video signal processing apparatus according to claim 2, wherein the interlaced scanning video signal is converted into a progressive scanning video signal, and the progressive scanning video signal from which noise components are removed is output. The interpolated motion detection means includes:
The difference information between one frame of the video signal and the difference information between two frames are obtained, the movement of the video signal between one frame and two frames is detected based on the obtained difference information, and the movement of the video signal is detected. Motion detection means for outputting a motion detection signal indicating the degree;
The difference between the frames of the video signal and the difference between the fields are obtained, and whether the video signal of the current field or frame is a telecine video signal is detected from the obtained difference, and telecine interpolation is performed based on the detection result. Telecine detection means for generating and outputting a telecine detection signal indicating an interpolation phase and detection of a telecine video signal for performing
The video signal processing apparatus according to claim 7, wherein a motion detection signal from the motion detection means and a telecine detection signal from the telecine detection means are output as the interpolation switching signal. It further comprises a cyclic noise removal processing means for receiving an input video signal and performing cyclic noise removal processing,
The field delay means delays the input video signal by the cyclic noise removal processing means by the cyclic noise removal processing means to delay the signal obtained in units of the field period,
The cyclic noise removal processing means performs the noise removal processing upon receiving the delayed video signal in addition to the input video signal,
The non-cyclic noise removal processing means receives the cyclic noise-removed signal output from the cyclic noise removal processing means as a signal of the current field or frame and uses it together with the delayed video signal to Type noise removal processing,
The three-dimensional scanning line conversion processing means sequentially converts the interlaced scanning video signal into a scanning video signal using the cyclic noise removed signal output from the cyclic noise removal processing means and the delayed video signal. The video signal processing apparatus according to any one of claims 1 to 8, wherein the video signal processing apparatus performs the process of: The video signal processing device according to any one of claims 1 to 9,
Display means;
A video signal display device comprising: display processing means for causing the display means to display a video based on the progressive scanning video signal from which the noise component output from the video signal processing device has been removed. A video signal processing method for removing a noise component having no correlation between fields or frames of a video signal, and obtaining a sequentially scanned video signal from which the noise component has been removed,
A plurality of delays by delaying an input video signal or a video signal of a current field or frame composed of a video signal obtained by cyclic noise removal processing in units of field periods, which are periods divided by a vertical synchronization signal A field delay step for outputting a video signal;
A video signal of a plurality of fields or frames composed of the video signal of the current field or frame and the plurality of delayed video signals is input, and a motion is obtained from a difference between signals of the plurality of fields or frames. A non-circular noise removal processing step for performing a filtering process controlled by movement to remove noise and generating a signal after acyclic noise removal;
When the input video signal is an interlaced scanning video signal,
A scanning line interpolation signal is generated from one or a combination of intra-field interpolation processing and inter-field interpolation processing from the video signals of the plurality of fields or frames and the signal after the non-cyclic noise removal, and interlaced scanning is performed. A three-dimensional scanning line conversion processing step for converting a video signal into a progressive scanning video signal and generating a progressive scanning video signal from which noise components have been removed;
Depending on whether the input video signal is an interlaced scanning video signal or a progressive scanning video signal, either the signal after the non-cyclic noise removal or the progressive scanning video signal from the three-dimensional scanning line interpolation processing step is selected. And a switching step of outputting the sequentially scanned video signal from which the noise component has been removed.

Images