JPH10178560A - Noise elimination circuit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate efficiently only noise while suppressing deterioration in an image. SOLUTION: Two field memories 41, 42 separate a current signal S41 of an object for filtering, a signal Si corresponding to a succeeding field signal and a preceding field signal S42. A still image is detected from the signals S41, S42. A field memory 43 outputs a signal S43 preceding by one frame to the signal S41. The signal S43 is corrected through a motion correction circuit 45 with a motion vector MV. Edges of the image is detected by vertical and horizontal edge detection circuits 50, 51 based on the signal S48 whose the motion is corrected and a discrimination circuit 52 outputs a selection signal S52 corresponding to the image. A selection circuit 59 selects and outputs an output signal of a vertical median filter 54 of a noise elimination filter, a horizontal median filter 55, an spatial median filter 56 or a delay circuit 57 without filter processing based on the selection signal S52.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and the like for removing noise (noise) from an input image using a median filter (order filter), which is a kind of nonlinear digital filter, to improve image quality. Related to a noise removal circuit.

[0002]

2. Description of the Related Art Conventionally, as a technique relating to this type of median filter, there has been a technique described in the following literature, for example. Literature; NHK Giken R & D, [7] (1989-11), Kobayashi et al., "Nonlinear Digital Filters for Images-LSI", page 27-36. As described in the above literature, television signals (hereinafter referred to as TV signals). ), Etc., when various impulsive noises such as glare-like noise seen in a BS reception image at the time of rainfall and burst-like noise in a VTR (video) playback image are mixed in the image passing through the transmission path. There is. In particular, since a TV signal requires real-time processing, it is difficult to remove noise without deteriorating the resolution of an original image by only linear operation using a digital filter. Therefore, the above-mentioned document proposes a noise elimination technique using a median filter, which is a kind of nonlinear digital filter, as an optimal method for removing impulse noise.

[0003] In general, speaking of a digital filter,
Refers to a linear digital filter. The linear digital filter delays the input by a plurality of stages, multiplies the input by a certain coefficient, and outputs the sum (linear operation). However, in TV signal processing, there are many processes in which the input / output relationship is not represented only by a linear relationship, such as square, logarithm, and threshold value processing. An arithmetic unit including these nonlinear elements is called a nonlinear digital filter. A median filter is hardware that implements a function similar to a non-linear filter for expanding and contracting an image in a non-linear processing algorithm.
FIG. 2 is a block diagram showing a schematic configuration of a conventional median filter.

The median filter determines a sequence relationship between neighboring pixels and a kernel forming circuit 11 for extracting neighboring pixels from an input image signal based on a kernel selection signal a (ie, determines a sequence relationship between a plurality of input data). ), And a selector 13 for selecting and outputting a pixel of a specific order based on the order selection signal b based on the determined order relationship. Median filter 10
Then, the kernel forming circuit 11 extracts a pixel existing in the vicinity of the target pixel (hereinafter referred to as “kernel”) in the sampled image, and uses the parallel sorter 12 for the extracted pixel to determine the signal level. Perform the ordering. Then, the selector 13 selects and outputs a pixel corresponding to a specific rank (rank). This operation
This is a non-linear operation. Such a median filter 1
When 0 is used to remove impulse noise, for example, noise removal such as filtering of 3 pixels × 3 lines in a field of the image signal or filtering of 3 pixels × 3 lines between fields is performed. A circuit is conceivable.

[0005]

However, the conventional noise elimination circuit has the following problems. Due to the filter processing for outputting the intermediate point of 3 pixels × 3 lines (9 samples), a line of one line or a point within four pixels in the image is judged as noise and removed, and there is a problem that the image quality deteriorates. Was. Accordingly, the applicant of the present application has previously filed Japanese Patent Application No. 7-163792 (hereinafter referred to as the earlier proposal).
Has proposed a noise elimination circuit that can minimize image quality degradation and effectively remove only noise. A further proposal is that a noise removal circuit using a median filter inputs a current signal to be subjected to an image signal filtering process and a previous signal before the current signal, and detects whether the current signal is a still image or a moving image. Detecting means, a vertical edge detecting means for detecting the presence or absence of a vertical edge of the image using the previous signal, a horizontal edge detecting means for detecting the presence or absence of the horizontal edge of the image using the previous signal, and adjacent using the previous image. A vertical median filter that selects and outputs a predetermined pixel signal near a pixel of interest from pixel signals of a vertical x (x is a natural number) line and a pixel signal of an adjacent horizontal y (y is a natural number) pixel A horizontal median filter that selects and outputs a predetermined pixel signal in the vicinity of the pixel of interest, and a processing delay time in the vertical median filter and the horizontal median filter, Delay means for delaying the image signal, and a selection means.

[0006] The selection means is a motion detection means,
Based on the detection results of the vertical edge detection means and the horizontal edge detection means, the output signal of the vertical median filter when there is no vertical edge in the still image, the output signal of the horizontal median filter when there is a vertical edge in the still image and there is no horizontal edge, Alternatively, when there is a vertical edge and a horizontal edge in a still image and in a moving image, the output signal of the delay unit is selected. That is, when the current signal and the previous signal in the image signal are given to the motion detecting means, whether the image is a still image or a moving image is detected. The presence / absence of a vertical edge and the presence / absence of a horizontal edge in the previous signal are detected by the vertical edge detection means and the horizontal edge detection means, respectively. When the image signal is input to the vertical median filter, the horizontal median filter, and the delay unit, a predetermined pixel signal near the pixel of interest is selected by each median filter, and the pixel signal is further delayed by the delay unit for a predetermined time. You. The selection means selects an output signal of a vertical median filter when there is no vertical edge in a still image based on the detection result of each of the detection means, and outputs a horizontal signal when there is a vertical edge in the still image and there is no horizontal edge. Select the output signal of the median filter. The selector selects the output signal of the delayer when there are a vertical edge and a horizontal edge in a still image and in a moving image.

In the previously proposed noise elimination circuit, the vertical median filter, the horizontal median filter, and the filter through (filter bypass) through the delay means are adaptively switched based on the detection result of each detection means. The edges are not mistaken for noise. Further, since the edge detection is performed using the previous signal of the still image detected by the motion detecting means, the impulse noise can be effectively removed. However, in the previously proposed noise elimination device, a signal one frame before is used for switching between a horizontal median filter, a vertical median filter, a filter through, and the like. Therefore, after determining whether the image is a still image by the motion detecting means, only the still image can be processed by the vertical median filter and the horizontal median filter.

[0008]

Means for Solving the Problems In order to solve the above-mentioned problems, a first invention of the present invention has the following configuration in a noise elimination circuit. That is, the noise elimination circuit of the first invention performs a delay process on an image signal which is composed of a plurality of fields and is sequentially input, and performs a filtering process on a current signal to be filtered and a previous field which does not include the current signal. A signal separating unit that separately outputs a previous field signal and a subsequent field signal that does not include the current signal and outputs the separated signal in parallel, and determines whether an image is a still image or a moving image using the previous field signal and the subsequent field signal. The apparatus comprises a still image detecting means for detecting and a previous signal generating means for obtaining a previous signal of a frame preceding the current signal. The noise elimination circuit inputs a motion vector representing a motion amount between the current signal and the previous signal and a zero vector, and selects the zero vector when the result of detection by the still image detecting means is a still image. When the detection result is a moving image, a selection unit that selects the motion vector, a motion correction unit that corrects the previous signal with the vector selected by the selection unit, and a determination unit that determines whether the motion vector is accurate. A motion-compensated inter-frame calculating means for extracting a difference between the corrected previous signal and the current signal; a vertical-edge detecting means for detecting presence or absence of a vertical edge in an image using the motion-corrected previous signal; Horizontal edge detecting means for detecting the presence or absence of a horizontal edge in an image using a signal.

Further, an image signal is input to the noise elimination circuit, and a pixel signal of an x-line (x is a natural number) pixel adjacent to the target pixel in the vertical direction is obtained from the pixel signal in the vicinity of the target pixel. A vertical median filter that selects and outputs a predetermined pixel signal; and an y pixel (where y is a natural number) that receives an image signal and is horizontally adjacent to the pixel of interest
A horizontal median filter that selects and outputs a predetermined pixel signal in the vicinity of the pixel of interest from the pixel signal of the pixel of interest, and inputs an image signal and uses pixel signals of a plurality of pixels in the fields before and after adjacent to the pixel of interest. A spatial median filter that selects and outputs a predetermined pixel signal near the pixel of interest, a vertical median filter, a horizontal median filter, and a delay unit that delays the image signal by a filter processing time in the spatial median filter, and a determination unit.
Selection output means is provided. The determination means determines the detection result of the still image detection means, the difference obtained by the motion compensation inter-frame calculation means, the detection result of the vertical edge detection means, and the detection result of the horizontal edge detection means, and determines a vertical median filter, a horizontal median filter. , A selection signal for selecting a spatial median filter or a delay means. The selection output means is configured to select and output an output signal of one of a vertical median filter, a horizontal median filter, a spatial median filter, and a delay means based on a selection signal of the determination means.

According to the first aspect of the present invention, since the noise removal circuit is configured as described above, the current signal, the previous field signal, and the subsequent field signal are separately output by the signal separating means. The still image detecting means detects whether the image is a still image or a moving image based on the previous field signal and the subsequent field signal that do not include the current signal. In the case of a moving image, the previous signal of the previous frame is corrected by the motion vector by the motion correction unit.
The vertical edge detecting means and the horizontal edge detecting means detect the edge of the image based on the corrected previous signal. That is,
The current signal is not used for edge detection. In addition, since edge detection is performed on a signal in which the previous signal is motion-compensated, the vertical median filter and the horizontal median filter can be applied even in the case of a moving image. Whether or not the motion vector is accurate is detected by the motion compensation inter-frame difference calculating means. By the determination means, from the detection result of the still image detection circuit, the vertical edge detection means, the horizontal edge detection means and the difference from the motion correction inter-frame calculation means,
A selection signal corresponding to the image is output. Based on this selection signal, the output signals of the vertical median filter, horizontal median filter, spatial median filter, and delay means are adaptively selected. Here, since the spatial median filter performs filtering using pixels in a field before and after the target pixel, even if line noise of a plurality of lines occurs in the target pixel and the field to which the target pixel belongs, the pixel signal of the target pixel is generated. Can be selected appropriately.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Here, a noise elimination circuit (hereinafter referred to as a moving image type noise elimination circuit) (I) based on the above proposal and improved so that median filter processing can be performed even on a moving image, After describing the previous proposal and the problem (II) of the moving image type noise elimination circuit, a noise elimination circuit that solves the problem (II) will be described. (I) Moving Picture Noise Elimination Circuit FIG. 3 is a block diagram showing the configuration of the moving picture noise elimination circuit. The moving image type noise elimination circuit includes a frame memory 21 that outputs a signal S21 of a previous frame delayed by one frame by storing an input image signal (that is, a current frame) Si of one frame, and a signal S21 of a previous frame. And a motion correction circuit 22 for inputting the motion vector MV and the motion vector MV. The motion vector MV is obtained by an iterative gradient method or a block matching method, and the motion correction circuit 22 uses the motion vector MV to generate a signal S of a previous frame.
21 and a function of outputting a corrected signal S22.

The signal S22 is output to the output side of the motion compensation circuit 22.
, A vertical edge detection circuit 24, and a horizontal edge detection circuit 25, to which a common input is input. The image signal S
i is also a connection to be input. The motion vector MV is input to the delay circuit 26 and is connected. The delay circuit 26 delays the motion vector MV and transmits it to the determination circuit 27. The determination circuit 27 includes the detection results S24 and S25 of the edge detection circuits 24 and 25 and the inter-frame difference calculation circuit 2
3 is the connection to which the operation result S23 is also input. The current frame, that is, the image signal Si, is further input to a vertical median filter 28, a horizontal median filter 29, a spatial median filter 30, and a delay circuit 31. The vertical median filter 28 selects and outputs a predetermined pixel signal in the vicinity of the pixel of interest from pixel signals of three vertically adjacent lines in a field of the current frame Si. Horizontal median filter 2
Numeral 9 is for selecting and outputting a predetermined pixel signal in the vicinity of the pixel of interest from pixel signals of nine horizontally adjacent pixels in the field of the current frame signal Si. The spatial median filter 30 applies a median filter process to three points of the target pixel and pixels adjacent to the target pixel on the upper and lower lines. The delay circuit 31 delays the current frame signal Si.

On the other hand, a delay circuit 32 is connected to the output side of the determination circuit 27. Output signal S32 of delay circuit 32
Is a connection input to the selection circuit 33. Selection circuit 33
Are input to the median filters 28 to 30 and the signals S28, S29, S30,
S31 is a connection to be input. The selection circuit 33 outputs signals S28 and S2 based on the output signal S32 of the delay circuit 32.
9, S30, or S31, and selects the video signal So.
Output. In this moving image type noise elimination circuit, the frame memory 21 outputs a signal S21 of the previous frame delayed by one frame to the motion correction circuit 22 with respect to the input image signal Si. The motion correction circuit 22 corrects the signal S21 using the motion vector MV, and outputs a signal S22 as a result of the correction. If the motion vector MV is accurate, the signal Si and the signal S22 should match.
In order to perform this match determination, the inter-frame difference calculation circuit 2
3 indicates that both signals Si,
The sum of the absolute values of the differences in S22 is obtained, and a signal S23 of the sum is output to the determination circuit 27. The determination circuit 27 determines the signal S23 with a predetermined threshold α, and if the value of the signal S23 is equal to or less than α, determines that the signal Si and the signal S22 match.

At the same time, the motion-compensated signal S22 is input to a vertical edge detection circuit 24 and a horizontal edge detection circuit 25, and a motion vector MV is input to a delay circuit 26. The vertical edge detection circuit 24 detects the presence or absence of a vertical edge in the signal S22 in which the motion of the previous frame has been corrected,
The edge detection information is output as a signal S24. Similarly,
The horizontal edge detection circuit 25 detects the presence or absence of a horizontal edge, and outputs edge detection information as a signal S25. The output signals S24 and S25 of the edge detection circuits 24 and 25 are output to the determination circuit 27. The delay circuit 26 delays the motion vector MV by the time required for processing in each edge detection circuit 24, 25, and outputs the result to the determination circuit 27. The judgment circuit 27 outputs the signal S2 from each of the edge detection circuits 24 and 25.
4, S25 and the output signal S26 of the delay circuit 26 are compared with the prepared threshold values β, γ, and δ, respectively, and a filter selection signal S27 corresponding to the comparison result is output.

FIG. 4 is a view for explaining the threshold value judgment in the judgment circuit 27 in FIG. 3 and the contents of the filter selection signal S27. The selection circuit 33 selects each of the median filters 28 to 30 and each of the output signals S28 to S31 of the delay circuit 31 based on the filter selection signal S27, and the determination circuit 27 specifies the selection by the signal S27. For example,
When the signal S23 is equal to or larger than the threshold value α, the motion vector MV is not reliable.
2 are determined to be different, and the output signal S31 of the delay circuit 31 is designated. Signal S23 <threshold α (accurate motion vector MV), signal S26 <threshold δ (still image) and signal S24
When <threshold β (no vertical edge), the output signal S30 of the spatial median filter 30 is specified. Signal S23
<Threshold α (accurate motion vector MV), signal S26> When threshold δ (moving image) and signal S24 <threshold β (no vertical edge), output signal S of vertical median filter 28
28 is specified. Similarly, when signal S23 <threshold α (accurate motion vector MV), signal S25> threshold γ (no horizontal edge) and signal S24> threshold β (with vertical edge), output signal S29 of horizontal median filter 29 is specified. Is done. Even if signal S23 <threshold α (accurate motion vector MV), when signal S25 <threshold γ (with horizontal edge) and signal S24> threshold β (with vertical edge), output signal S31 of delay circuit 31 is selected. Is done. When the output signal S31 of the delay circuit 31 is selected, the median filter processing is not performed, that is, the through mode is specified.

The vertical median filter 28 selects a predetermined pixel signal in the vicinity of the pixel of interest from pixel signals of three lines vertically adjacent to each other in the same field as the current frame Si, and outputs the selected signal to the selection circuit 33. The horizontal median filter 29 selects a predetermined pixel signal near the pixel of interest from nine horizontally adjacent pixel signals in the field of the current frame signal Si, and outputs the selected pixel signal to the selection circuit 33. The spatial median filter 30 applies a median filter process to pixels adjacent to the target pixel among the upper and lower lines sandwiching the target pixel, that is, pixels in a field different from the target pixel, to obtain an image signal of the target pixel. And outputs the result to the selection circuit 33. The delay circuit 31 delays the current frame signal Si by the processing time of each of the median filters 28 to 30 and outputs the signal to the selection circuit 33. Judgment circuit 27
Is output by the delay circuit 32 and is supplied to the selection circuit 33 as the signal S32. The selection circuit 33 selects one of the output signals S28 to S30 of each of the median filters 28 to 30 and the output signal S31 of the delay circuit 31 based on the signal S32, and outputs the selection result signal as the video signal So. .

As described above, the edge detection is performed on the signal S22 obtained by correcting the signal S21 of the previous frame by the motion vector MV, and the optimum processing of the median filters 28 to 30 for the current frame Si is performed using the detection result. , So that noise can be effectively removed from moving images. (II) Problems with the moving picture noise elimination circuit proposed in (I) and the moving picture noise elimination circuit described in (I). , The noise cannot be removed.
The reason is that line noise having a width of two or more lines cannot be removed by filtering with the 3-tap vertical median filter 28 or horizontal median filter 29 in the field. For example, vertical median filter 29
It may be possible to remove line noise by increasing the number of taps such as..., But increasing the number of taps of the vertical median filter 29 results in applying a filter to an image with less correlation between lines. May cause malfunction. Also, in the case of line noise over 10 lines, the number of taps needs to be at least 21 or more, which is not practical.

If the noisy image is limited to a still image without vertical edges, the use of the spatial median filter 30 for two or more line noises is effective for noise removal. However, in the moving image type noise elimination circuit of FIG. 4, one of the conditions for using the spatial median filter 30 is that “an inter-frame difference value using its own field (filtered field) is equal to or less than a certain threshold α”. Contains conditions. Therefore, in the case of two or more line noises, an inter-frame difference occurs even in a still image, and as a result, there is a problem that the processing of the spatial median filter 30 is not used.

(III) Noise Elimination Circuit Solving Problem (II) FIG. 1 is a block diagram showing a configuration of a noise elimination circuit according to an embodiment of the present invention. This noise elimination circuit stores the input image signal Si for one field and outputs the signal S41 by delaying the input image signal Si by one field, similarly to the first field memory 41 that outputs the signal S41 for one field.
A signal S4 stored for one field and delayed by one field
2 and a second field memory 42 for outputting a signal S4
2 is stored for one field, and a third field memory 43 that outputs a signal S43 delayed by one field is provided. The input image signal Si is composed of, for example, two fields. The field memories 41 and 42 constitute signal separation means, and the input image signals Si
Are output in parallel with the subsequent field signal, the signal S41 as the current signal to be filtered, and the signal S42 as the previous field signal. The field memory 43 is a previous signal generating means, and the signal S43 is a signal of a previous frame with respect to the current signal S41. The output side of the field memory 42 is connected to an inter-frame difference calculation circuit 44 which receives the signal Si and the signal S42 as inputs. The inter-frame difference calculation circuit 44 constitutes a still image detecting means for detecting whether the image is a still image or a moving image.

The noise elimination circuit is further provided with a field memory 45, to which the motion vector MV is inputted. On the output side of the field memory 45, a selection circuit 4 as selection means is provided.
6 are connected. A zero vector is also input to the selection circuit 46. A delay circuit 47 is connected to the output side of the field memory 43, and a motion compensation circuit 48 as a motion compensation means is connected to the output sides of the selection circuit 46 and the delay circuit 47. On the output side of the motion correction circuit 48, a motion correction inter-frame difference calculation circuit 49 as a motion correction inter-frame difference calculation means, a vertical edge detection circuit 50 as a vertical edge detection means, and a horizontal edge detection means as a horizontal edge detection means. The circuit 51 is connected. Vertical edge detection circuit 50, horizontal edge detection circuit 5
1, and a determination circuit 52, which is a determination unit that receives the signals S49 to S51 output from them and the output signal S44 of the inter-frame difference calculation circuit 44 on the output side of the motion-correction inter-frame difference calculation circuit 49, It is connected.

On the other hand, a delay circuit 53 is connected to an output terminal of the field memory 41. The delay circuit 53
A signal S53 obtained by delaying the signal S43 is output to the motion vector correction inter-frame difference calculating circuit 49. The vertical median filter 54, the horizontal median filter 55, and the spatial Median filter 56 and delay circuit 5
7 is a connection that is output in common. The determination circuit 52 includes a vertical median filter 54, a horizontal median filter 5
5. A determination means for outputting a filter selection signal S52 for selecting the spatial median filter 56 and the delay circuit 57. The output side of the determination circuit 52 outputs the signal S52 to the vertical median filter 54, the horizontal median filter 55,
A delay circuit 58 that delays by the processing time in the spatial median filter 56 is connected. The output side of the delay circuit 58 is connected to a selection circuit 59 as selection output means. Signals S54 to S57 output from each of the vertical median filter 54, the horizontal median filter 55, the spatial median filter 56, and the delay circuit 57 are input to the input terminal of the selection circuit 59.

Next, the operation of the noise elimination circuit of FIG. 1 will be described with reference to FIGS. The frame of the input image signal Si is composed of two fields forming odd lines and even lines. The image signal Si is delayed by one field by the field memory 41, and further delayed by one field by the field memory. Therefore, the output signal S41 of the field memory 41 serving as the current signal and the image signal Si serving as the subsequent field signal and the field memory 42 serving as the previous field signal are provided.
Output signal S42 is shifted by one frame. The inter-frame difference calculation circuit 44 calculates and accumulates the absolute value of the difference between the subsequent field signal Si and the signal S42 in units of pixels or blocks of a predetermined size. This cumulative result indicates whether the image is a still image or a moving image.
The signal 44 is output to the selection circuit 46 and the determination circuit 52. The selection circuit 46 and the determination circuit 52 determine the signal S44 with a threshold, and determine that the image is a still image when the signal S44 is equal to or less than the threshold. The motion vector MV input to the noise elimination circuit is a motion vector detected by an iterative gradient method, a block matching method, or the like.
Delays the motion vector MV by an amount corresponding to the delay of the signal Si in the field memory 41, and outputs it to the selection circuit 46 as a signal S45.

FIG. 5 is a diagram for explaining selection of the selection circuit 46 in FIG. The selection circuit 46, based on the signal S44,
The signal 45 or 0 vector corresponding to the motion vector MV is selected. The result of the selection is output to the motion vector correction circuit 48 as a signal S46. For example, when the signal S44 is smaller than the threshold a, the selection circuit 46 determines that the image is a still image, and sets a zero vector to the signal S46. Conversely, the signal S
When 44 is larger than the threshold a, the selection circuit 46 sets the value of the motion vector MV in the signal 46. On the other hand, a signal obtained by delaying the previous field signal S42 by the field memory 43 is the signal S43. The current signal S41 and the signal S43 output from the field memory 41 have a relationship of one frame apart. That is, the signal S43 is a previous signal corresponding to the image signal of the previous frame with respect to the current signal S41. Previous signal S4
3 is supplied to the delay circuit 47. The delay circuit 47 outputs to the motion correction circuit 48 a signal S47 obtained by delaying the previous signal S43 by the delay in the inter-frame difference calculation circuit 44. Therefore, the signal S46 and the signal S47 have the same phase. The motion correction circuit 48 corrects the signal S47 using the signal S46 representing the motion vector. This correction result is a signal S48. Output signal S of field memory 41
41 is given to a delay circuit 53. Delay circuit 53
Delays the signal S41 by the processing time of the delay circuit 47 and the motion compensation circuit 48 that process the signal signal S43. That is, the phase of the signal S48 and the phase of the signal S53 output from the delay circuit 43 are matched.

FIG. 6 is a diagram for explaining the motion correction by the motion correction circuit 48 in FIG. In a moving image, as shown in FIG.
Moving by V matches the image of the current frame 61. If the signal S46 given from the selection circuit 46, that is, the motion vector MV is accurate, the signal S48 and the signal S53 should match. The signal 48 subjected to the motion correction is supplied to a vertical edge detection circuit 50, a horizontal edge detection circuit 51, and a motion correction inter-frame difference calculation circuit 49. Signal S
53 is also supplied to the motion-compensated inter-frame difference calculation circuit 49 and output to the vertical median filter 54, the horizontal median filter 55, the spatial median filter 56, and the delay circuit 57.

The motion compensation inter-frame difference calculation circuit 49
In order to determine the coincidence between the signal S48 and the signal S53, the cumulative sum of the absolute value of the difference between the signal S48 and the signal S53 is obtained for each block of a certain size. The cumulative result is signal S49, which is output to determination circuit 52. The determination circuit 52 determines the signal S49 with a predetermined threshold α, and if the value of the signal S49 is equal to or less than α, determines that the signal S48 and the signal S53 match. The vertical edge detection circuit 50 detects the presence or absence of a vertical edge of the image using the corrected signal S48 as a previous signal, and outputs a signal S50 corresponding to edge detection information. The horizontal edge / vertical edge detection circuit 51 detects the presence or absence of a horizontal edge in an image and outputs a signal S51 corresponding to edge detection information. As a method of detecting a horizontal edge, for example, there is a method of obtaining a difference between adjacent pixels, comparing the difference value with a threshold value, and determining a horizontal edge when the comparison result is equal to or larger than the threshold value. Each signal S50, S51
Is given to the judgment circuit 52.

FIG. 7 is a diagram showing the relationship between the threshold value judgment of the judgment circuit 57 in FIG. 1 and the filter selection signal S52. The determination circuit 52 includes an inter-frame difference operation circuit 44, a motion compensation inter-frame difference operation circuit 49, a vertical edge detection circuit 50,
And the output signals S44 and S4 of the horizontal edge detection circuit 51.
7 to determine one of the median filters 54 to 56 and the delay circuit 57 based on the determination results of the absolute values of 9 to S51 based on the threshold values α, β, γ, and δ. A filter selection signal S52 as shown in FIG. For example, when the signal S44 is smaller than the threshold α (still image) and the signal S50 is smaller than the threshold β (there is no vertical edge), the determination circuit 52 instructs to select the spatial median filter 56 with the signal S52. Similarly, when the signal S44 is larger than the threshold value α (moving image), the signal S44
When it is determined that 49 is smaller than the threshold δ (the motion vector MV is accurate) and the signal S50 is smaller than the threshold β (no vertical edge), the signal S52 instructs to select the vertical median filter 50. Is done.

The signal S44 is larger than the threshold α (moving picture), the signal S49 is smaller than the threshold δ (accurate motion vector MV), the signal S50 is larger than the threshold β (there is a vertical edge), and the signal S51 is When it is determined that the horizontal median filter 51 is smaller than the threshold γ (there is no horizontal edge), the signal S52 instructs to select the horizontal median filter 51. The signal S44 is smaller than the threshold α (still image), the signal S50 is larger than the threshold β (there is a vertical edge),
When it is determined that the signal S51 is smaller than the threshold value γ (there is no horizontal edge), the signal S52 instructs to select the horizontal median filter 50. Signal S50
Is larger than the threshold β (there is a vertical edge), the signal S51
Is larger than the threshold value γ (there is a horizontal edge), the signal S52 instructs to select the delay circuit 57. That is, the through mode in which the median filter processing is not performed is selected. That is, as a result of the threshold determination,
If the signal S44 is equal to or larger than the threshold α, the determination circuit 52 determines that the image is a moving image, and does not select the spatial median filter 56. In addition to being a moving image, the signal S4
When 9 is equal to or greater than the threshold value δ, the reliability of the motion vector MV is not reliable, and it is determined that the signal S48 does not match the signal S53, and the delay circuit 57 is selected. In other words, if the signal S48 and the signal S53 do not match, the position where the edge is detected is shifted from the position where the actual filtering is performed, so that the through mode in which the median filtering is not performed is selected.

FIG. 8 is a diagram for explaining the processing of the vertical median filter in FIG. Vertical median filter 5
4 receives the signal S53 and receives an adjacent x line (x in FIG. 8) belonging to the same field as the target pixel to be filtered.
= 3), a pixel signal of a pixel adjacent to the target pixel is selected as a pixel signal of the target pixel to be obtained. For example, median filter processing is performed using the pixel signal of the pixel of interest on the nth line, the pixel signal of the pixel above the pixel of interest on the (n−1) th line, and the pixel signal of the pixel below the pixel of interest on the (n + 1) th line. To select the pixel signal of the target pixel.
FIG. 9 is a view for explaining the processing of the horizontal median filter 55 in FIG. The horizontal median filter 55 receives the signal S53 and receives y pixels (y = 9 in FIG. 9) in the same field as the target pixel and on the same line as the target pixel.
And selects the selected pixel signal as the pixel signal of the target pixel. For example, the pixel signal of the nth pixel of interest is subjected to median filtering using the pixel signals of n−4 to n + 4 pixels, and the pixel signal of the pixel of interest is selected.

FIG. 10 is a diagram for explaining the processing of the spatial median filter 56 in FIG. Processing in the vertical median filter 54 and the horizontal median filter 55 is as follows.
Although the filtering process is performed in the field, the spatial median filter 56 performs the filtering process on the target pixel using the pixel data of the pixels belonging to the preceding and succeeding fields. As shown in FIG. 10, when viewed on the image, the N-th field is located around the target pixel a in the M line, and N−1
The pixel of the (M + 1) th line of the Nth field, the pixel of the (M−1) th line of the (N−1) th field, the pixel of the (M + 1) th line of the (N + 1) th field, and the
There is one line of pixels. Spatial median filter 5
6 is, for example, the M of the (N-1) th field indicated by a black dot.
A median filter process is performed using the pixel signals of the +1 line pixel and the M-1 line pixel of the (N + 1) th field to select the pixel signal of the target pixel. By doing so, M + belonging to the same field as the pixel of interest a
Even if there is line noise on the M-2 line, the noise of the target pixel a can be removed. Further, since the pixels in the preceding and succeeding fields are closer to the target pixel than the pixels in the same field, a high resolution can be obtained even as a result of the filtering process. The delay circuit 57 includes the vertical median filter 5
4. The signal S53 is delayed by the filter processing time of the horizontal median filter 55 and the spatial median filter 56, and the delayed signal S57 is output to the selection circuit 59.

On the other hand, the filter selection signal S52 output from the determination circuit 52 is output to the delay circuit 58. The delay circuit 56 generates a signal S <b> 58 obtained by delaying the signal S <b> 52 by the filter processing time of the vertical median filter 54, the horizontal median filter 55, and the spatial median filter 56, and outputs the signal S <b> 58 to the selection circuit 59. The selection circuit 59 selects one of the output signals S54 to S57 of the vertical median filter 54, the horizontal median filter 55, the spatial median filter 56, and the delay circuit 57 based on the signal S58, and outputs it as the video signal So. I do. FIG.
FIG. 11 is a diagram illustrating line noise of an image. The effect of the present embodiment will be described with reference to FIG.

In the noise elimination circuit of the present embodiment, the edge detection is performed by performing the motion correction on the previous signal S43 of the previous frame using the motion vector MV, so that the optimum median filter processing can be applied even in the case of a moving image. it can.
Further, it has field memories 41 and 42 for delaying the input image signal Si, and outputs a rear field signal, a current signal field, and a previous field signal separately with a simple configuration, and outputs the signal S41 with the current signal S41 to be filtered. In this case, the signal Si is a field signal after the next field and the signal S42 is a field signal before the previous field. The output signal S44 of the inter-frame difference calculation circuit 44, which is one of the parameters for selecting the spatial median filter 56, is calculated in the previous and next fields that do not include the pixel to be filtered.
Therefore, erroneous detection is eliminated at the stage of detecting a still image, and line noise can be effectively removed. That is, even if line noise having a plurality of line widths is generated as shown in FIG. 11, the noise can be removed by applying the spatial median filter 56. Note that the present invention is not limited to the above embodiment, and various modifications are possible. For example,
The sampling numbers of the vertical median filter 54, the horizontal median filter 55, and the spatial emmedian filter are as follows.
The number of lines and the number of pixels are not limited to the pixels shown in FIGS.

[0032]

As described in detail above, the first to third embodiments
According to the present invention, a signal separating unit that outputs a current signal to be filtered, a front field signal not including the current signal, and a subsequent field signal separately and in parallel, and a front field signal and a rear field signal are used. Still image detecting means for detecting whether a still image or a moving image, a previous signal generating means for obtaining a previous signal of a frame before the current signal, a motion correcting means,
The apparatus includes horizontal and vertical edge detection means, motion compensation frame calculation means, and determination means. Therefore, a selection signal suitable for an image can be output without erroneous detection, and output signals of a vertical median filter, a horizontal median filter, a spatial median filter, and a delay unit can be adaptively switched, and noise can be removed even in a moving image. I have. Furthermore, since the still image is detected using the front field signal and the rear field signal, even if there is Latin noise of two or more lines, the output signal of the spatial median filter can be selected, and line noise generated in a plurality of widths is also removed. Becomes possible.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a noise removing circuit according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a conventional median filter.

FIG. 3 is a block diagram showing a configuration of a moving image type noise elimination circuit.

FIG. 4 is a diagram illustrating threshold value determination in a determination circuit 27 in FIG. 3 and contents of a filter selection signal S27.

FIG. 5 is a diagram illustrating selection of a selection circuit 46 in FIG. 1;

FIG. 6 is a diagram for explaining motion correction by a motion correction circuit 48 in FIG. 1;

FIG. 7 is a diagram showing a relationship between a threshold value determination of a determination circuit 57 in FIG. 1 and a filter selection signal S52.

FIG. 8 is a view for explaining processing of a vertical median filter in FIG. 1;

FIG. 9 is a diagram illustrating processing of a horizontal median filter 55 in FIG. 1;

FIG. 10 is a view for explaining processing of a spatial median filter 56 in FIG. 1;

FIG. 11 is a diagram illustrating line noise of an image.

[Explanation of symbols]

41 to 43 First to third field memories 44 Inter-frame difference calculation circuit (still image detection means) 46 Selection circuit (selection means) 48 Motion correction circuit 49 Motion correction inter-frame difference calculation circuit 50 Vertical edge detection circuit 51 Horizontal edge Detection circuit 52 Judgment circuit 54 Vertical median filter 55 Horizontal median filter 56 Spatial median filter 57 Delay circuit (Delay means) 59 Selection circuit (Selection output means) Si Input image signal (Next field signal) S41 Current signal S42 Previous field signal S43 Previous Signal MV motion vector

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Yoshiharu Hoshino 2-2-1 Jinnan, Shibuya-ku, Tokyo Inside the Japan Broadcasting Corporation Broadcasting Center

Claims (3)

[Claims] 1. A delay process is performed on an image signal composed of a plurality of fields and sequentially input, and a current signal to be filtered, a previous field signal of a previous field not including the current signal, and the current signal A signal separation unit that outputs in parallel apart from the after-field signal of the after-field that does not include the still-field detection unit that detects whether an image is a still image or a moving image using the before-field signal and the after-field signal. A previous signal generating means for obtaining a previous signal of a frame preceding the current signal; a motion vector representing a motion amount between the current signal and the previous signal and a zero vector; If the detection result by is a still image, select the 0 vector,
Selecting means for selecting a motion vector when the detection result is a moving image; motion correcting means for correcting the previous signal with a vector selected by the selecting means; and determining whether the motion vector is accurate. Motion-compensated inter-frame calculating means for extracting a difference between the corrected previous signal and the current signal; vertical edge detecting means for detecting presence or absence of a vertical edge in an image using the motion-corrected previous signal; Horizontal edge detection means for detecting the presence or absence of a horizontal edge in an image by using the corrected previous signal; and inputting the image signal, the x-line (where x is a natural number) vertically adjacent to the pixel of interest A vertical median filter that selects and outputs a predetermined pixel signal in the vicinity of the pixel of interest from the pixel signal of the pixel; A horizontal median filter for selecting and outputting a predetermined pixel signal in the vicinity of the pixel of interest from pixel signals of y pixels (where y is a natural number) adjacent in the horizontal direction; A spatial median filter for selecting and outputting a predetermined pixel signal in the vicinity of the pixel of interest using pixel signals of a plurality of pixels in the preceding and succeeding fields adjacent to the pixel of interest; the vertical median filter, the horizontal median filter, and the space A delay unit for delaying the image signal by a filter processing time in a median filter; a detection result of the still image detection unit, a difference obtained by the motion compensation inter-frame calculation unit, a detection result of the vertical edge detection unit, and the horizontal edge detection. The vertical median filter, the horizontal median filter, Determining means for generating a selection signal for selecting the spatial median filter or the delay means; and any one of the vertical median filter, the horizontal median filter, the spatial median filter or the delay means based on the selection signal of the determination means And a selection output means for selecting and outputting the output signal of (i). 2. The first field memory for outputting the current signal by storing and delaying the image signal as the post-field signal for one field and delaying the current signal, A second field memory for storing the current signal output from the field memory for one field, delaying the current signal by one field and outputting the previous field signal, wherein the previous signal generation means comprises: 3. The noise elimination circuit according to claim 1, comprising a third field memory for storing one or more fields and delaying the previous field signal and outputting the previous signal. 3. The determination means based on a detection result of the still image detection means, a difference obtained by the motion compensation inter-frame calculation means, a detection result of the vertical edge detection means, and a detection result of the horizontal edge detection means. When it is determined that the image is a still image and the image has no vertical edge, the spatial median filter is selected, the image is a moving image, the motion vector is accurate, and the image is perpendicular to the image. When it is determined that there is no edge, the vertical median filter is selected, and it is determined that the image is a moving image, the motion vector is accurate, the image has a vertical edge, and the image has no horizontal edge. When selecting the horizontal median filter, the image is a still image, the image has a vertical edge,
And when it is determined that there is no horizontal edge in the image, the horizontal median filter is selected.When it is determined that the image has a vertical edge and the image has horizontal edges, the delay means is selected. 3. The noise elimination circuit according to claim 1, wherein the delay unit is selected when it is determined that the image is a moving image and the motion vector is incorrect.