JP2005328162A - Signal processing apparatus and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology capable of suppressing variations in contrast correction and reducing the circuit scale in a filtering processing for smoothing parts other than edges of image signals in a state that edges having steep changes in pixel values are accurately maintained. <P>SOLUTION: A delay section 81 generates 5 kinds of taps on the basis of an input signal. Control sections 91 to 98 respectively generate control signals C1 to C5 on the basis of each tap received from the delay section 81. An LPF composite section 82 processes the tap of interval 16 on the basis of the control signal C1. An LPF composite section 84 processes the tap of interval 8 on the basis of the control signal C2. An LPF composite section 86 processes the tap of interval 4 on the basis of the control signal C3. An LPF composite section 88 processes the tap of interval 2 on the basis of the control signal C4. An LPF composite section 90 processes the tap of interval 1 on the basis of the control signal C5. The technology above can be applied to television receivers or the like. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a signal processing apparatus and method, and more particularly to a signal processing apparatus and method suitable for use in enhancing an internal texture surrounded by an edge without over-emphasizing an edge in an image.

  Conventionally, in a video camera, the contrast (brightness / darkness difference) and sharpness (brightness of the boundary) of an image captured using an image sensor such as a CCD (Charge Coupled Device), CMOS (Complementary Metal-Oxide Semiconductor), etc. As an improvement method, a contrast enhancement method by gradation conversion, a high-frequency component enhancement method for enhancing the contrast of a high-frequency component in an image, and the like are considered.

  As a contrast enhancement method, tone curve adjustment for converting each pixel level of an image with a function having a predetermined input / output relationship (hereinafter referred to as a level conversion function), or frequency distribution of pixel levels A method called histogram equalization has been proposed in which the level conversion function is adaptively changed according to the above.

  As a high-frequency component emphasis method, a method called an unsharp mask that performs so-called contour emphasis in which the contour (edge) of a person or an object in an image is extracted and the extracted edge is emphasized has been proposed.

  However, the contrast enhancement method has a problem that the contrast can be improved only in a part of the luminance range in the entire dynamic range (difference between the maximum level and the minimum level) of the image. In addition, the tone curve adjustment is performed. In the case of (1), there is a problem that the contrast is lowered in the brightest and darkest parts of the image, and in the case of histogram equalization, in the vicinity of the luminance region where the frequency distribution is small.

  In the high-frequency component emphasis method, only the contrast of the high-frequency component of the image is emphasized, so that there is a problem that the vicinity of the edge of the image is unnaturally conspicuous and deterioration of the image quality cannot be avoided.

  As a method for solving such a problem, conventionally, an image signal processing apparatus configured as shown in FIG. 1 is used to store an edge other than the edge of the input image data in a state where a sharp change in pixel value is stored. There is a method for enhancing a portion other than an edge by amplifying the portion (for example, Patent Document 1).

  In the image signal processing apparatus shown in FIG. 1, the input image signal is input to the ε filter 1 and the subtraction unit 2. The ε filter 1 receives an image signal that slightly fluctuates across a steep edge as shown in FIG. 2A, and only the edge is extracted as shown in FIG. 2B (the part other than the edge is smoothed). E) converted into an image signal and output to the subtracting unit 2 and the adding unit 4. Here, in FIG. 2, the horizontal axis indicates the arrangement of pixels arranged continuously, and the vertical axis indicates each pixel value.

Specific processing of the ε filter 1 will be described with reference to FIGS. 3 and 4. The ε filter 1 sequentially determines each pixel of the input image as the target pixel C, and, as shown in FIG. 3, a plurality of neighboring pixels (in this case, six pixels L3) that are continuous in the horizontal direction with the target pixel C as the center. , L2, L1, R1, R2, R3), and a pixel value of the target pixel C and a plurality of neighboring pixels is set to a predetermined tap coefficient (for example, {1, 2, 3, 4, 3, 2, 1}) and the result is output as a conversion result C ′ corresponding to the target pixel C
C ′ = (1 · L3 + 2 · L2 + 3 · L1 + 4 · C
+ 3 · R1 + 2 · R2 + 1 · R3) / 16 (1)

However, as shown in FIG. 4, for neighboring pixels in which the difference in pixel value from the target pixel C is equal to or greater than a predetermined threshold ε (in the case of FIG. 4, neighboring pixels R2 and R3), the pixel value is set to the target pixel. Replace with the one of C and calculate. That is, in the case of FIG. 4, the following equation (2) is calculated.
C ′ = (1 · L3 + 2 · L2 + 3 · L1 + 4 · C
+ 3 · R1 + 2 · C + 1 · C) / 16 (2)

  Returning to FIG. The subtracting unit 2 subtracts the image signal input from the ε filter 1 from the image signal input from the previous stage (the same as the input to the ε filter 1), thereby slightly changing the image other than the edge. The signal is extracted and output to the amplifying unit 3. The amplification unit 3 amplifies the output of the subtraction unit 2 and outputs the amplified output to the addition unit 4. The adder 4 adds the image signal from which the part other than the edge output from the amplifier 3 is amplified and the image signal from which only the edge input from the ε filter 1 is extracted. Therefore, the addition result is an image signal in which a portion other than the edge is amplified in a state where a steep edge is held, and this is an output signal of the image processing apparatus.

JP 2001-298621 A

  By the way, in the ε filter 1 of the image signal processing apparatus shown in FIG. 1, the predetermined threshold value ε is 100. For example, as shown in FIG. 5, an edge image signal whose pixel value changes sharply is obtained. When input, if the height of the edge is 100 or more, the image signal output from the ε filter 1 retains the edge shape as shown in FIG. When it is slightly smaller than the threshold value ε (for example, when the height of the edge is 99), the image signal output from the ε filter 1 is dull because the shape of the edge is not maintained as shown in FIG. It becomes. In FIG. 5 and the like, the horizontal axis indicates the pixel arrangement, and the vertical axis indicates the pixel value.

  For example, when an image signal of an edge having a shape as shown in FIG. 8 is input, if the height of the edge is 100 or more, the image signal output from the ε filter 1 is shown in FIG. Thus, the shape of the edge is maintained, but when the height of the edge is slightly smaller than the threshold value ε (for example, when the edge height is 99), the image signal output from the ε filter 1 is As shown in FIG. 10, the shape of the edge is not maintained and is smoothed.

  Therefore, in the conventional ε filter 1, when the height of the edge of a continuously input image slightly fluctuates across the threshold ε, the contrast correction of the edge of the image after filtering greatly fluctuates. There was a problem that it visually deteriorated.

  Further, as another viewpoint, there has been a problem that it is desired to reduce the circuit scale of an apparatus that performs image processing as described above.

  The present invention has been made in view of such a situation. In a filtering process for smoothing a portion other than an edge in a state where an edge having a steep change in pixel value is accurately retained, fluctuations in edge contrast correction are achieved. An object of the present invention is to reduce the circuit scale.

  The signal processing device of the present invention sequentially specifies input signals that are continuously arranged as attention signals, and specifies a plurality of neighboring signals at predetermined intervals with reference to the attention signal from among the input signals, A first tap generating means for generating a plurality of different taps composed of a signal of interest and a neighboring signal; a plurality of calculating means for calculating a weighting coefficient based on the taps generated by the first tap generating means; A plurality of smoothed signals are calculated according to a weighting coefficient calculated by a calculation means by calculating a plurality of smoothed signals by weighted averaging the signal of interest and a plurality of neighboring signals using a plurality of different coefficient sets. A delay means for delaying the supply of the weighting coefficient calculated by the calculation means to the calculation means, and the output signal of the calculation means in order, the signal of interest And a second tap generation means for designating a plurality of neighboring signals at predetermined intervals from the output signal as a reference from the output signal and generating a tap composed of the noticed signal and the neighboring signal. To do.

  The input signal may be a pixel value of a pixel constituting the image.

  The first and second tap generating means may be constituted by flip-flop circuits.

  The signal processing method of the present invention sequentially specifies input signals that are continuously arranged as a target signal, specifies a plurality of neighboring signals at predetermined intervals from the input signal as a reference, and A first tap generation step for generating a plurality of different taps composed of a signal of interest and a neighborhood signal; a plurality of calculation steps for calculating a weighting coefficient based on the taps generated in the processing of the first tap generation step; A plurality of smoothed signals are calculated by performing weighted averaging of the signal of interest and a plurality of neighboring signals constituting a tap using a plurality of different coefficient sets, and the plurality of calculated signals are calculated according to the weighting coefficient calculated in the processing of the calculation step. A plurality of calculation steps for synthesizing the smoothed signal, a delay step for delaying the supply of the weighting coefficient calculated in the calculation step processing to the calculation step, and an operation The output signal in the step processing is sequentially designated as the attention signal, and multiple neighboring signals are designated at predetermined intervals based on the attention signal from the output signal, and a tap composed of the attention signal and the neighboring signal is generated. And a second tap generation step.

  In the signal processing apparatus and method of the present invention, a plurality of different taps are generated from continuously arranged input signals, a weighting coefficient is calculated based on the generated taps, and the attention signal constituting the taps A plurality of smoothed signals are calculated from a plurality of neighboring signals, and the calculated plurality of smoothed signals are synthesized according to the calculated weighting coefficient. Further, a tap is generated from the output signal in the calculation process.

  According to the present invention, it is possible to suppress fluctuations in edge contrast correction in a filtering process for smoothing a part other than an edge while accurately maintaining an edge with a sharp change in pixel value, and the circuit scale. Can be reduced.

  Embodiments of the present invention will be described below. Correspondences between constituent elements described in the claims and specific examples in the embodiments of the present invention are exemplified as follows. This description is to confirm that specific examples supporting the invention described in the claims are described in the embodiments of the invention. Therefore, even if there are specific examples that are described in the embodiment of the invention but are not described here as corresponding to the configuration requirements, the specific examples are not included in the configuration. It does not mean that it does not correspond to a requirement. On the contrary, even if a specific example is described here as corresponding to a configuration requirement, this means that the specific example does not correspond to a configuration requirement other than the configuration requirement. not.

  Further, this description does not mean that all the inventions corresponding to the specific examples described in the embodiments of the invention are described in the claims. In other words, this description is an invention corresponding to the specific example described in the embodiment of the invention, and the existence of an invention not described in the claims of this application, that is, in the future, a divisional application will be made. It does not deny the existence of an invention that is added by correction.

  The signal processing device according to claim 1 (for example, the non-linear filter 80 in FIG. 30) sequentially designates the input signals arranged continuously as the attention signal, and uses the attention signal as a reference from among the input signals. First tap generation for designating a plurality of neighboring signals at predetermined intervals and generating a plurality of different taps (five types of taps of intervals 16, 8, 4, 2, 1) composed of the target signal and the neighboring signal 30 (for example, the delay unit 81 in FIG. 30) and a plurality of calculation units (for example, the control units 91, 92, 94, FIG. 30 in FIG. 30) that calculate weighting coefficients based on the taps generated by the first tap generation unit. 96, 98) and a weighted average calculated by calculating means by calculating a plurality of smoothed signals by performing weighted averaging of a signal of interest and a plurality of neighboring signals constituting a tap using different coefficient sets. A plurality of arithmetic means (for example, LPF synthesis units 82, 84, 86, 88, and 90 in FIG. 30) that synthesizes the plurality of smoothed signals calculated according to the number, and a weighting coefficient calculation means calculated by the calculation means The delay means for delaying the supply (for example, the delay unit 93 in FIG. 30) and the output signal of the calculation means are sequentially designated as the attention signal, and a plurality of output signals are set at predetermined intervals with reference to the attention signal. It includes a second tap generation means (for example, a delay unit 83 in FIG. 30) that designates a proximity signal and generates a tap (interval 8 tap) composed of the signal of interest and the proximity signal.

  The signal processing method according to claim 4 sequentially specifies input signals that are continuously arranged as a signal of interest, and outputs a plurality of neighboring signals at predetermined intervals from the input signal with reference to the signal of interest. A first tap generation step (for example, step S11 in FIG. 31) for generating a plurality of different taps (5 types of taps of intervals 16, 8, 4, 2, 1, and 1) that are designated and made up of a signal of interest and neighboring signals And a plurality of calculation steps (for example, step S13 in FIG. 31) for calculating a weighting coefficient based on the tap generated in the first tap generation step processing, and the attention signal and the plurality of neighboring signals constituting the tap. Calculating a plurality of smoothed signals by performing weighted averaging using a plurality of different coefficient sets, and calculating the calculated plurality of smoothed signals according to the weighting coefficients calculated in the processing of the calculation step. A plurality of calculation steps (for example, step S14 in FIG. 31), a delay step (for example, step S13 in FIG. 31) for delaying the supply of the weighting coefficient calculated in the calculation step processing to the calculation step, and a calculation step In this process, the output signal is sequentially designated as a signal of interest, a plurality of neighboring signals are designated at predetermined intervals from the output signal as a reference, and a tap (interval 8) comprising the signal of interest and the neighboring signal is designated. And a second tap generation step (for example, step S15 in FIG. 31).

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 11 shows a configuration example of the nonlinear filter. This non-linear filter 11 is used in place of the ε filter 1 of the image signal processing apparatus shown in FIG. 1, and the height of the edge included in the image signal of the continuously input image has a predetermined value. In the case of slight fluctuations across the threshold value, edge correction is also made to fluctuate gently. More specifically, even if the height of the edge of the pixel value included in the image signal is smaller than the predetermined threshold value ε1, the converted image is not reduced until it becomes smaller than the threshold value ε2 (ε2 <ε1). This is to moderate the fluctuation of the edge included in the signal.

  The nonlinear filter 11 includes a control signal generator 12, an LPF (Low Pass Filter) 13, and a synthesizer 14.

  The control signal generation unit 12 is adjacent to the image signal that is input from the upper stage and includes pixel values of the pixels in the raster order, for example, the five pixels L2 that are adjacent to each other with no space left and right around the target pixel C as shown in FIG. , L1, C, R1, and R2 (that is, taps of interval 1) are set, weighting coefficients w1 and w3 are generated based on the taps, and output to the synthesizer 14 as control signals. Hereinafter, the pixel values of the pixels L2, L1, C, R1, and R2 constituting the tap are also described as pixel values L2, L1, C, R1, and R2 as appropriate. In addition, among the pixels constituting the tap, pixels other than the target pixel are described as neighboring pixels.

  The LPF 13 sets the same tap as that of the control signal generator 12, performs a weighted average of five pixel values included in the tap using three types of tap coefficients, and outputs three types of smoothed signals F1, F3, and F5. Calculate. The synthesizer 14 synthesizes the smoothed signals F1, F3, F5 according to the control signal generated by the control signal generator 12.

FIG. 13 shows an example of tap coefficients used in the LPF 13. When the smoothed signal F1 is calculated, tap coefficients {0, 0, 1, 0, 0} are used as shown in the following equation (3).
F1 = (0 * L2 + 0 * L1 + 1 * C + 0 * R1 + 0 * R2) / 1 (3)

When the smoothed signal F3 is calculated, tap coefficients {0, 1, 2, 1, 0} are used as shown in the following equation (4).
F3 = (0 · L2 + 1 · L1 + 2 · C + 1 · R1 + 0 · R2) / 4 (4)

When the smoothed signal F5 is calculated, tap coefficients {1, 2, 2, 2, 1} are used as shown in the following equation (5).
F5 = (1 · L2 + 2 · L1 + 2 · C + 2 · R1 + 1 · R2) / 8 (5)

  Next, the operation of the nonlinear filter 11 will be described with reference to the flowchart of FIG.

  In step S <b> 1, the control signal generation unit 12 sequentially determines the input pixels in the raster order as the target pixel C one by one. In step S2, as shown in FIG. 12, the control signal generator 12 sets a tap composed of pixels L2, L1, C, R1, and R2 adjacent to each other with no interval left and right around the target pixel C. Hereinafter, the pixels L2, L1, R1, and R2 included in the tap are described as neighboring pixels L2, L1, R1, and R2.

  In step S3, the LPF 13 sets taps in the same manner as the control signal generator 12, performs weighted averaging of the tap pixels L2, L1, C, R1, and R2 using the equations (1) to (3), The resulting smoothed signals F1, F3, F5 are output to the synthesis unit 14.

  In step S4, the control signal generator 12 determines the pixel value differences | L2-C |, | L1-C |, | R1-C |, | of the pixel value of the target pixel C and the neighboring pixels L2, L1, R1, and R2. R2-C | is calculated.

In step S5, the control signal generation unit 12 calculates weighting coefficients w1 and w3 used in the synthesis unit 14 based on the difference calculated in the process of step S4. Specifically, the larger of the difference | L1-C |, | R1-C | between the pixel of interest C and the neighboring pixels L1 and R1 located symmetrically about the pixel of interest C is the variable d1. And the weighting coefficient w1 is calculated according to the variable d1, as shown in FIG.
d1 = MAX [| L1-C |, | R1-C |]
If d1 <ε2, w1 = 0
When ε2 ≦ d1 <ε1, w1 = (d1−ε2) / (ε1−ε2)
When ε1 ≦ d1, w1 = 1

Similarly, the larger of the pixel value differences | L2-C |, | R2-C | between the target pixel C and neighboring pixels L2 and R2 that are symmetric with respect to the target pixel C as the variable d2. Substituting and calculating the weighting coefficient w3 according to the variable d2.
d2 = MAX [| L2-C |, | R2-C |]
When d2 <ε2, w3 = 0
When ε2 ≦ d2 <ε1, w3 = (d2−ε2) / (ε1−ε2)
When ε1 ≦ d2, w3 = 1

The weighting coefficients w1 and w3 calculated in this way are output to the synthesis unit 14 as control signals. In step S6, the synthesizer 14 synthesizes the smoothed signals F1, F3, F5 calculated by the LPF 13 according to the following equation (6) using the weighting coefficients w1, w3 calculated by the control signal generator 12. , The value C ′ after filtering of the target pixel C is output.
C ′ = w1 · F1 + (1−w1) · w3 · F3
+ (1-w1). (1-w3) .F5 (6)
This is the end of the description of the operation of the nonlinear filter 11.

  According to this non-linear filter 11, for example, when the threshold value ε1 is 100 and the threshold value ε2 is 70, and an edge image signal whose pixel value changes sharply as shown in FIG. If the edge height is 100 which is equal to or greater than the threshold value ε1, the image signal output from the nonlinear filter 11 has an edge shape as shown in FIG. 16, and the height of the edge is the threshold value ε1. Even if it is smaller than 90 and 80, the image signal output from the non-linear filter 11 is held with only a slight change, as shown in FIGS. Furthermore, as the edge height gradually decreases, the image signal output from the non-linear filter 11 gradually changes in edge shape, and the edge height reaches 70, which is equal to the threshold value ε2. The shape of the edge of the image signal output from the nonlinear filter 11 is as shown in FIG.

  As is clear from comparison between FIG. 19 and FIG. 7 showing the output of the conventional ε filter 1, in the conventional ε filter 1, the height of the edge of the input image signal is higher than the threshold ε = 100. Even if it is slightly small 99, the shape of the edge after filtering has changed greatly as shown in FIG. 7, but in the nonlinear filter 11 to which the present invention is applied, the edge of the input image signal is changed. Even if the height of is smaller than the threshold value ε1, the shape of the edge after filtering changes gently until the threshold value ε2.

  Next, FIG. 20 shows another configuration example of the nonlinear filter. This non-linear filter 21 is used in place of the ε filter 1 of the image signal processing apparatus shown in FIG. 1, and a narrowband processing unit 22 that performs a filtering process by setting a tap consisting of 5 pixels at an interval 4. An intermediate band processing unit 26 that performs filtering by setting a tap of 5 pixels in interval 2 for the output of the narrow band processing unit 22, and a tap of 5 pixels in interval 1 for the output of the intermediate band processing unit 26 A wideband processing unit 30 that performs filtering processing by setting, a delay unit 34 that delays the input signal by the processing time in the narrowband processing unit 22, and a delay unit 35 that delays the input signal by the processing time in the mediumband processing unit 26. Is done.

  As shown in FIG. 21, in the tap of the interval 4, neighboring pixels are determined at intervals of 3 pixels with the target pixel C as a reference in the pixel array arranged in the horizontal direction. That is, the eighth pixel l8 on the left side of the target pixel C, the fourth pixel l4, the fourth pixel r4 on the right side, and the eighth pixel r8 are set as neighboring pixels L2, L1, R1, and R2, respectively.

  As shown in FIG. 22, in the tap of interval 2, neighboring pixels are determined at intervals of one pixel with reference to the target pixel C in the pixel array arranged in the horizontal direction. That is, the left fourth pixel l4, the second pixel l2, the right second pixel r2, and the fourth pixel r4 are set as neighboring pixels L2, L1, R1, and R2, respectively.

  As shown in FIG. 23, in the tap of interval 1, neighboring pixels are determined without any interval on the basis of the target pixel C in the pixel array arranged in the horizontal direction. That is, the second pixel l2, the first pixel l1, the first pixel r1, the right pixel r1, and the second pixel r2 on the left side are the neighboring pixels L2, L1, R1, and R2, respectively.

  Returning to FIG. The narrow band processing unit 22 is configured in the same manner as the nonlinear filter 11 shown in FIG. That is, the control signal generation unit 23, LPF 24, and synthesis unit 25 of the narrowband processing unit 22 correspond to the control signal generation unit 12, LPF 13, and synthesis unit 14 of the nonlinear filter 11, respectively. The difference between the narrowband processing unit 22 and the non-linear filter 11 is that the non-linear filter 11 processes the tap of the interval 1 shown in FIG. 12, whereas the narrowband processing unit 22 uses the interval shown in FIG. 4 taps and the control signal generator 23 generates a control signal using pixels other than taps (details will be described later).

  The middle band processing unit 26 is also configured similarly to the nonlinear filter 11 shown in FIG. That is, the control signal generation unit 27, the LPF 28, and the synthesis unit 29 of the intermediate band processing unit 26 correspond to the control signal generation unit 12, the LPF 13, and the synthesis unit 14 of the nonlinear filter 11, respectively. The difference between the intermediate band processing unit 26 and the non-linear filter 11 is that the tap of the interval 2 shown in FIG. 22 is processed and the control signal generating unit 27 generates a control signal using pixels other than the tap. (Details will be described later).

  Further, the broadband processing unit 30 is configured in the same manner as the nonlinear filter 11 shown in FIG. That is, the control signal generator 31, LPF 32, and synthesizer 33 of the intermediate band processor 26 correspond to the control signal generator 12, LPF 13, and synthesizer 14 of the nonlinear filter 11, respectively. The broadband processing unit 30 processes the tap of the interval 1 shown in FIG.

  FIG. 24 shows an example of tap coefficients used in the LPF 24 of the narrowband processing unit 22, the LPF 28 of the medium band processing unit 26, and the LPF 32 of the wideband processing unit 30.

  In the LPF 24, when the tap of the interval 4 is set and the smoothed signal F11 is calculated, the tap coefficients {0, 0, 1, 0, 0} are used as in the equation (3). In this calculation, in the conventional ε filter 1, taps composed of 17 pixels adjacent to each other with no interval are set as the center, and tap coefficients {0, 0, 0, 0, 0, 0, 0, 0, 1 are set. , 0, 0, 0, 0, 0, 0, 0, 0}.

  In the LPF 24, when the tap of the interval 4 is set and the smoothed signal F13 is calculated, the tap coefficients {0, 1, 2, 1, 0} are used as in the equation (4). In this calculation, in the conventional ε filter 1, taps of 17 pixels adjacent to each other with no interval are set around the target pixel, and tap coefficients {0, 0, 0, 0, 1, 0, 0, 0, 2 are set. , 0, 0, 0, 1, 0, 0, 0, 0}.

  In the LPF 24, when the tap of the interval 4 is set and the smoothed signal F15 is calculated, the tap coefficients {1, 2, 2, 2, 1} are used as in the equation (5). In this calculation, in the conventional ε filter 1, taps of 17 pixels adjacent to each other with no interval are set around the target pixel, and tap coefficients {1, 0, 0, 0, 2, 0, 0, 0, 2 are set. , 0, 0, 0, 2, 0, 0, 0, 1}.

  In the LPF 28, when the tap of the interval 2 is set and the smoothed signal F21 is calculated, the tap coefficients {0, 0, 1, 0, 0} are used as in the equation (3). In this calculation, in the conventional ε filter 1, taps of 9 pixels adjacent to each other with no interval are set around the target pixel, and tap coefficients {0, 0, 0, 0, 1, 0, 0, 0, 0 are set. } Is equivalent to the weighted average.

  In the LPF 28, when the tap of the interval 2 is set and the smoothed signal F23 is calculated, the tap coefficients {0, 1, 2, 1, 0} are used as in the equation (4). In this calculation, in the conventional ε filter 1, taps of 9 pixels adjacent to each other with no interval are set with the pixel of interest at the center, and tap coefficients {0, 0, 1, 0, 2, 0, 1, 0, 0 are set. } Is equivalent to the weighted average.

  In the LPF 28, when the tap of the interval 2 is set and the smoothed signal F25 is calculated, the tap coefficients {1, 2, 2, 2, 1} are used as in the equation (5). In this calculation, in the conventional ε filter 1, taps of 9 pixels adjacent to each other with no interval are set with the pixel of interest at the center, and tap coefficients {1, 0, 2, 0, 2, 0, 2, 0, 1 } Is equivalent to the weighted average.

  In the LPF 32, when the tap of the interval 1 is set and the smoothed signal F31 is calculated, the tap coefficients {0, 0, 1, 0, 0} are used as in the equation (3).

  When the tap of interval 1 is set in the LPF 32 and the smoothed signal F33 is calculated, the tap coefficients {0, 1, 2, 1, 0} are used as in the equation (4).

  In the LPF 32, when the tap of the interval 1 is set and the smoothed signal F35 is calculated, the tap coefficients {1, 2, 2, 2, 1} are used as in the equation (5).

  Next, the operation of the nonlinear filter 21 will be described. First, the filtering process of the narrowband processing unit 22 will be described with reference to the flowchart of FIG.

  In step S <b> 1, the control signal generation unit 23 sequentially determines the pixels in the raster order constituting the input image signal as the target pixel C one by one. In step S <b> 2, the control signal generation unit 23 sets the tap of the interval 4 with the target pixel C as the center, as illustrated in FIG. 21.

  In step S3, the LPF 24 sets a tap in the same manner as the control signal generator 23, performs weighted averaging of the tap pixels L2, L1, C, R1, and R2 using equations (1) to (3), The resulting smoothed signals F11, F13, F15 are output to the synthesizer 25.

  In step S4, the control signal generator 23 compares the pixel values | L2-C |, | L1-C |, | R1-C |, | of the pixel value between the target pixel C and the neighboring pixels L2, L1, R1, and R2. R2-C | is calculated. Further, the control signal generating unit 23 is located on the target pixel C and the target pixel C side with respect to the neighboring pixels L2 and R2, and is not included in the tap. The pixels l7, l6, l5, l3, l2, l1, r1 , R2, r3, r5, r6, r7, pixel value differences | l7-C |, | l6-C |, | l5-C |, | l3-C |, | l2-C |, | l1-C |, | R1-C |, | r2-C |, | r3-C |, | r5-C |, | r6-C |, | r7-C |

  In step S5, the control signal generation unit 23 calculates weighting coefficients w1 and w3 used in the synthesis unit 25 based on the difference calculated in the process of step S4.

Specifically, the difference between the pixel values of the pixel of interest C and pixels located closer to the pixel of interest C than the neighboring pixels L1 and R1 that are symmetric with respect to the pixel of interest C | L1-C |, | l3-C |, | L2-C |, | l1-C |, | r1-C |, | r2-C |, | r3-C |, | R1-C | As shown in FIG. 15, the weighting coefficient w1 is calculated according to the variable d1.
d1 = MAX [| L1-C |, | l3-C |, | l2-C |, | l1-C |
, | R1-C |, | r2-C |, | r3-C |, | R1-C |]
If d1 <ε2, w1 = 0
When ε2 ≦ d1 <ε1, w1 = (d1−ε2) / (ε1−ε2)
When ε1 ≦ d1, w1 = 1

Similarly, the difference between the pixel values of the pixel of interest C and pixels located closer to the pixel of interest C than the neighboring pixels L2 and R2 that are symmetric with respect to the pixel of interest C | L2-C |, | l7-C | , | L6-C |, | l5-C |, | r5-C |, | r6-C |, | r7-C |, | R2-C | The weighting coefficient w3 is calculated according to d2.
d2 = MAX [| L2-C |, | l7-C |, | l6-C |, | l5-C |
, | R5-C |, | r6-C |, | r7-C |, | R2-C |]
When d2 <ε2, w3 = 0
When ε2 ≦ d2 <ε1, w3 = (d2−ε2) / (ε1−ε2)
When ε1 ≦ d2, w3 = 1

  The weighting coefficients w1 and w3 calculated in this way are output to the synthesis unit 25 as control signals. In step S6, the synthesizer 25 synthesizes the smoothed signals F11, F13, F15 calculated by the LPF 24 according to the equation (6) using the weighting coefficients w1, w3 calculated by the control signal generator 23, The value C ′ after filtering of the target pixel C is output to the middle band processing unit 26. Above, description of the filtering process by the narrowband process part 22 is complete | finished.

  Next, the image signal output from the narrowband processing unit 22 is subjected to filtering processing by the intermediate band processing unit 26 and further subjected to filtering processing by the wideband processing unit 30. The filtering process in the medium band processing unit 26 and the wide band processing unit 30 is the same as the filtering process performed by the narrow band processing unit 22 except that the tap interval is different, and thus the description thereof is omitted.

  According to this non-linear filter 21, for example, an image signal having a threshold value ε1 of 100 and a threshold value ε2 of 70 and having a steep edge width within the interval 4 is input as shown in FIG. In this case, if the height of the edge is 100 that is equal to or greater than the threshold ε1, the image signal output from the nonlinear filter 21 has the shape of the edge substantially retained as shown in FIG. When the value is 90 or 80, which is smaller than the threshold value ε1, the image signal output from the nonlinear filter 21 has an edge shape that gradually changes as shown in FIGS. Further, as the edge height gradually decreases, the image signal output from the non-linear filter 21 has a shape in which the edge shape gradually changes and the edge height reaches 70 which is equal to the threshold value ε2. The shape of the edge of the image signal output from the nonlinear filter 21 is as shown in FIG.

  As is apparent from a comparison between FIG. 28 and FIG. 10 showing the output of the conventional ε filter 1, in the conventional ε filter 1, the height of the edge of the input image signal is higher than the threshold ε = 100. Even if it is slightly small 99, the shape of the edge after filtering has changed greatly as shown in FIG. 10. However, in the nonlinear filter 21 to which the present invention is applied, the edge of the input image signal is changed. Even if the height of is smaller than the threshold value ε1, the shape of the edge after filtering changes gently until the threshold value ε2.

  Next, FIG. 29 expands the nonlinear filter 21 composed of the three band processing units (the narrow band processing unit 22, the middle band processing unit 26, and the wide band processing unit 30) shown in FIG. The structural example of the nonlinear filter comprised from a band process part is shown.

  The non-linear filter 51 in FIG. 29 has a narrowest band processing unit 52 that performs a filtering process by setting a tap at an interval 16 for an input signal, and a narrow band that performs a filtering process by setting a tap at an interval 8 for an input signal. A band processing unit 55, a medium band processing unit 56 that performs a filtering process by setting a tap of an interval 4 for an input signal, a broadband processing unit 57 that performs a filtering process by setting a tap of an interval 2 for an input signal, And the widest bandwidth processing unit 58 that sets the tap of the interval 1 for the input signal and performs the filtering process. The narrowest band processing unit 52, the narrow band processing unit 55, the middle band processing unit 56, the wide band processing unit 57, and the widest band processing unit 58 are the same as the narrow band processing unit 22 of the nonlinear filter 21 shown in FIG. Perform the process.

  Further, the non-linear filter 51 includes a delay unit 59 that delays the input signal by the processing time in the narrowest band processing unit 52, a delay unit 60 that delays the input signal by the processing time in the narrowest band processing unit 55, and the narrowest input signal. The delay unit 61 delays the processing time in the band processing unit 56 and the delay unit 62 delays the input signal by the processing time in the narrowest band processing unit 527.

  Note that the middle band processing unit 56, the wide band processing unit 57, and the widest band processing unit 58 of the nonlinear filter 51 are the narrow band processing unit 22, the middle band processing unit 26, and the wide band processing of the nonlinear filter 21 shown in FIG. It corresponds to the section 30. Therefore, as compared with the nonlinear filter 21, the nonlinear filter 51 can filter a narrower band signal component by the amount provided with the narrowest band processing unit 52 and the narrow band processing unit 55.

  The narrowest band processing unit 52 includes an LPF synthesis unit 53 and a control signal generation unit 54. The LPF synthesis unit 53 sets a tap of an interval 16 based on an input signal composed of pixel values of pixels in raster order, performs a weighted average of pixel values of 5 pixels included in the tap using three types of tap coefficients, Three types of smoothed signals are calculated. Further, the LPF synthesis unit 53 synthesizes three types of smoothed signals in accordance with the control signal generated by the control signal generation unit 54. Similar to the LPF synthesis unit 53, the control signal generation unit 54 sets a tap for the interval 16, generates a weighting coefficient based on the tap, and outputs the weighting coefficient to the LPF synthesis unit 53 as a control signal.

  Similar to the narrowest band processing unit 52, the narrow band processing unit 55, the middle band processing unit 56, the wide band processing unit 57, and the widest band processing unit 58 are each configured by an LPF synthesis unit and a control signal generation unit. However, the illustration thereof is omitted.

  Next, regarding the electronic circuit for realizing the nonlinear filter 51, the number of flip-flop circuits (hereinafter simply referred to as FF) to be used will be considered. Hereinafter, it is assumed that an input signal is input for each pixel in synchronization with the operation clock of the non-linear filter 51 and the pixel value is 8 bits.

  In general, a circuit that generates a tap composed of P pixels (b bits) at an interval I requires (I × (P−1) × b) FFs.

  Therefore, in the narrowest band processing unit 52 in which the taps of the interval 16 are independently generated at two places, 1024 (= 16 × 4 × 8 × 2) FFs are required. Similarly, 512 (= 8 × 4 × 8 × 2) FFs are required in the narrowband processing unit 55 in which the taps of the interval 8 are independently generated at two locations. In the middle band processing unit 56 in which the taps of the interval 4 are independently generated at two places, 256 (= 4 × 4 × 8 × 2) FFs are required. In the broadband processing unit 57 in which the taps of the interval 2 are independently generated at two places, 128 (= 2 × 4 × 8 × 2) FFs are required. In the widest bandwidth processing unit 58 in which the taps of the interval 1 are independently generated at two locations, 64 (= 1 × 4 × 8 × 2) FFs are required.

  The delay units 59 to 62 respectively input signals corresponding to the number of clocks corresponding to the processing time in the narrowest band processing unit 52, the narrow band processing unit 55, the middle band processing unit 56, the wide band processing unit 57, and the widest band processing unit 58. Therefore, if the time required for processing (processing for calculating and synthesizing the smoothed signal) other than tap generation by the narrowest band processing unit 52 or the like is M clocks, ((I × 2 + M) × b) clock.

Therefore, if M = 3, the delay amount of the delay unit 59 is 280 (= (16 × 2 + 3) × 8) clocks. Therefore, the delay unit 59 requires 280 FFs. Similarly, the delay amount of the delay unit 60 is 152 (= (8 × 2 + 3) × 8) clocks. Therefore, the delay unit 60 requires 152 FFs. The delay amount of the delay unit 61 is 88 (= (4 × 2 + 3) × 8) clocks. Therefore, the delay unit 61 requires 88 FFs.
The delay amount of the delay unit 62 is 56 (= (2 × 2 + 3) × 8) clocks. Therefore, the delay unit 62 requires 56 FFs.

  From the above results, when the nonlinear filter 51 is realized as an electronic circuit under the above assumption, at least 2600 (= 1024 + 512 + 256 + 128 + 64 + 280 + 152 + 88 + 56 + 40) FFs are required.

  Next, FIG. 30 is an embodiment of the present invention, which is a non-linear filter that can obtain an output signal similar to that of the non-linear filter 51, and can be realized with a smaller number of FFs than the non-linear filter 51. An example of the configuration of a filter is shown.

  In this non-linear filter 80, the delay unit (DL) 81 generates taps (signals S1 to S5) consisting of five pixels of interval 16 based on the input signal, and outputs them to the LPF synthesis unit 82 and the control unit 91. Taps (signals S6 to S10) consisting of 5 pixels at interval 8 are generated and output to the controller 92, taps (signals S11 to S15) consisting of 5 pixels at interval 4 are generated and output to the controller 94, Taps (signals S16 to S20) consisting of 5 pixels in interval 2 are generated and output to the control unit 96, and taps (signals S21 to S25) consisting of 5 pixels in interval 1 are generated and output to the control unit 98. Therefore, since the delay unit 81 generates a tap composed of 5 pixels at the interval 16 at the maximum, 512 (= 16 × 4 × 8) FFs are required.

  The signals S1 to S5 are delayed by 0, 16, 32, 48, and 64 clocks, respectively, with respect to the input signal. The signals S6 to S10 are delayed by 32, 40, 48, 56, and 64 clocks, respectively, with respect to the input signal. The signals S11 to S15 are delayed by 48, 52, 56, 60, and 64 clocks, respectively, with respect to the input signal. The signals S16 to S20 are delayed by 56, 58, 60, 62, and 64 clocks, respectively, with respect to the input signal. The signals S21 to S25 are delayed by 60, 61, 62, 63, and 64 clocks, respectively, with respect to the input signal.

  The LPF synthesizing unit 82 calculates the three types of smoothed signals by weighted average of the pixel values of the taps composed of the five pixels in the interval 16 input from the delay unit 81, and generates the control signal C 1 generated by the control unit 91. Based on this, three types of smoothed signals are synthesized and output to the delay unit 83.

  Based on the output of the LPF synthesis unit 82, the delay unit 83 generates a tap composed of 5 pixels in interval 8 (the signal S 26 is the pixel value of the target pixel) and outputs the tap to the LPF synthesis unit 84. Therefore, the delay unit 83 requires 256 (= 8 × 4 × 8) FFs.

  The LPF synthesizing unit 84 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of the five pixels of interval 8 input from the delay unit 83, and is generated by the control unit 92. 3 types of smoothed signals are synthesized based on the control signal C <b> 2 input via the control signal C <b> 2 and output to the delay unit 85.

  Based on the output of the LPF synthesis unit 84, the delay unit 85 generates a tap composed of 5 pixels in interval 4 (the signal S27 is the pixel value of the pixel of interest) and outputs it to the LPF synthesis unit 86. Therefore, the delay unit 85 requires 128 (= 4 × 4 × 8) FFs.

  The LPF synthesis unit 86 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of 5 pixels in interval 4 input from the delay unit 85, and is generated by the control unit 94. 3 types of smoothed signals are synthesized based on the control signal C <b> 3 input via the control signal C <b> 3 and output to the delay unit 87.

  Based on the output of the LPF synthesis unit 86, the delay unit 87 generates a tap composed of 5 pixels in interval 2 (the signal S28 is the pixel value of the pixel of interest) and outputs it to the LPF synthesis unit 88. Therefore, the delay unit 87 requires 64 (= 2 × 4 × 8) FFs.

  The LPF synthesizing unit 88 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of the five pixels in interval 2 input from the delay unit 87, and is generated by the control unit 96. 3 types of smoothed signals are synthesized based on the control signal C 4 input via the control signal C 4 and output to the delay unit 89.

  Based on the output of the LPF synthesis unit 88, the delay unit 89 generates a tap composed of 5 pixels in interval 1 (the signal S29 is the pixel value of the target pixel) and outputs the tap to the LPF synthesis unit 90. Therefore, the delay unit 89 requires 32 (= 1 × 4 × 8) FFs.

  The LPF synthesizing unit 90 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of the five pixels in interval 1 input from the delay unit 89, and is generated by the control unit 98. The three types of smoothed signals are synthesized based on the control signal C5 input via the control signal C5 and output to the subsequent stage of the nonlinear filter 80 as an output signal.

  The control unit 91 calculates a weighting coefficient based on the pixel value of the tap composed of 5 pixels in the interval 16 input from the delay unit 81, and outputs the weighting coefficient to the LPF synthesis unit 82 as a control signal C 1 for the LPF synthesis unit 82.

  The control unit 92 calculates a weighting coefficient based on the pixel value of the tap composed of 5 pixels of interval 8 input from the delay unit 81 and outputs the weighting coefficient to the delay unit 93 as the control signal C2 for the LPF synthesis unit 84. The delay unit 93 delays the control signal C2 from the control unit 92 so that the signal S26 from the delay unit 83 and the control signal C2 from the control unit 92 are input to the LPF synthesis unit 84 in synchronization. Here, the delay amount with respect to the input signal of the signal S26 is 48 + M (= 16 × 2 + M + 8 × 2) clocks when the processing time in the LPF synthesis unit 82 is M clocks, and the delay amount with respect to the input signal of the control signal C2 is As with the signal S8, there are 48 clocks. Therefore, the delay unit 93 delays the control signal C2 by M clocks.

  The control unit 94 calculates a weighting coefficient based on the pixel value of the tap composed of 5 pixels in the interval 4 input from the delay unit 81 and outputs the weighting coefficient to the delay unit 95 as the control signal C3 for the LPF synthesis unit 86. The delay unit 95 delays the control signal C3 from the control unit 94 so that the signal S27 from the delay unit 85 and the control signal C3 from the control unit 94 are input to the LPF synthesis unit 86 in synchronization. Here, the delay amount with respect to the input signal of the signal S27 is 56 + 2M (= 16 × 2 + M + 8 × 2 + M + 4 × 2) clock if the processing time in the LPF synthesizing unit 84 is M clock, and the delay with respect to the input signal of the control signal C3. The amount is 56 clocks, similar to signal S13. Therefore, the delay unit 95 delays the control signal C3 by 2M clocks.

  The control unit 96 calculates a weighting coefficient based on the pixel value of the tap composed of 5 pixels in interval 2 input from the delay unit 81 and outputs the weighting coefficient to the delay unit 97 as a control signal C4 for the LPF synthesis unit 88. The delay unit 97 delays the control signal C4 from the control unit 96 so that the signal S28 from the delay unit 87 and the control signal C4 from the control unit 96 are input to the LPF synthesis unit 88 in synchronization. Here, the delay amount with respect to the input signal of the signal S28 is 60 + 3M (= 16 × 2 + M + 8 × 2 + M + 4 × 2 + M + 2 × 2) clock, assuming that the processing time in the LPF synthesis unit 86 is M clock, and the input signal of the control signal C4 The delay amount with respect to is 60 clocks as in the signal S18. Therefore, the delay unit 96 delays the control signal C4 by 3M clocks.

  The control unit 98 calculates a weighting coefficient based on the pixel value of the tap composed of five pixels of interval 1 input from the delay unit 81 and outputs the weighting coefficient to the delay unit 99 as a control signal C5 for the LPF synthesis unit 90. The delay unit 99 delays the control signal C5 from the control unit 98 so that the signal S29 from the delay unit 89 and the control signal C5 from the control unit 98 are input to the LPF synthesis unit 90 in synchronization. Here, the delay amount of the signal S29 with respect to the input signal is 62 + 4M (= 16 × 2 + M + 8 × 2 + M + 4 × 2 + M + 2 × 2 + M + 1 × 2) clock if the processing time in the LPF synthesis unit 88 is M clocks, and the control signal C5 The delay amount with respect to the input signal is 62 clocks like the signal S23. Therefore, the delay unit 99 delays the control signal C5 by 4M clocks.

  Next, the operation of the nonlinear filter 80 will be described with reference to the flowchart of FIG. In step S11, based on the input signal, the delay unit 81 taps consisting of 5 pixels in interval 16, taps consisting of 5 pixels in interval 8, taps consisting of 5 pixels in interval 4, taps consisting of 5 pixels in interval 2 , And generation of taps composed of 5 pixels in interval 1 are started, and are output to the LPF synthesis unit 82 and the control unit 91, the control unit 92, the control unit 94, the control unit 96, or the control unit 98, respectively.

  In step S12, the control unit 91, the control unit 92, the control unit 94, the control unit 96, and the control unit 98 control the control signal C1 for the LPF synthesis unit 82 and the control signal C2 for the LPF synthesis unit 84 based on the input tap. The control signal C3 for the LPF synthesis unit 86, the control signal C4 for the LPF synthesis unit 88, or the control signal C5 for the LPF synthesis unit 90 is generated and output to the subsequent stage. In step S13, the delay unit 93, the delay unit 95, the delay unit 97, and the delay unit 99 start delaying the control signals C1 to C5 input from the previous stage.

  In step S <b> 14, the LPF synthesis unit 82 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of 5 pixels in the interval 16 input from the delay unit 81, and is generated by the control unit 91. Based on the control signal C <b> 1, three types of smoothed signals are combined and output to the delay unit 83. In step S <b> 15, the delay unit 83 generates a tap composed of 5 pixels at interval 8 based on the output of the LPF synthesis unit 82 and outputs the tap to the LPF synthesis unit 84.

  In step S <b> 16, the LPF synthesis unit 84 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of 5 pixels in the interval 8 input from the delay unit 83, and is generated by the control unit 92. Based on the control signal C <b> 2 input via the delay unit 93, three types of smoothed signals are synthesized and output to the delay unit 85. In step S <b> 17, the delay unit 85 generates a tap composed of 5 pixels in interval 4 based on the output of the LPF synthesis unit 84 and outputs the tap to the LPF synthesis unit 86.

  In step S <b> 18, the LPF synthesis unit 86 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of 5 pixels in the interval 4 input from the delay unit 85, and is generated by the control unit 94. Based on the control signal C <b> 3 input via the delay unit 95, three types of smoothed signals are synthesized and output to the delay unit 87. In step S <b> 19, the delay unit 87 generates a tap composed of 5 pixels in interval 2 based on the output of the LPF synthesis unit 86 and outputs the tap to the LPF synthesis unit 88.

  In step S20, the LPF synthesis unit 88 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of the five pixels in interval 2 input from the delay unit 87, and is generated by the control unit 96. Based on the control signal C <b> 4 input via the delay unit 97, three types of smoothed signals are synthesized and output to the delay unit 89. In step S <b> 21, the delay unit 89 generates a tap composed of 5 pixels in interval 1 based on the output of the LPF synthesis unit 88 and outputs the tap to the LPF synthesis unit 90.

  In step S22, the LPF synthesis unit 90 calculates the three types of smoothed signals by performing weighted averaging of the pixel values of the taps composed of the five pixels in interval 1 input from the delay unit 89, and is generated by the control unit 98. Based on the control signal C5 input through the delay unit 99, three types of smoothed signals are synthesized and output to the subsequent stage of the nonlinear filter 80 as an output signal. Above, description of operation | movement of the nonlinear filter 80 is complete | finished.

  Here, it is assumed that the M clocks of the processing time in each of the LPF synthesis units 82, 84, 85, 86, and 88 are 3 clocks, and that the control signals C2, C3, C4, and C5 are 4 bits. Then, in order to realize the delay units 93, 95, 97, and 99, 12 (= 3 × 4), 24 (= 6 × 4), 36 (= 9 × 4), and 48 (= 12 × 4), respectively. ) FFs are required.

  Therefore, when the nonlinear filter 80 of FIG. 30 is realized as an electronic circuit under the above-described assumption, 1112 (= 512 + 256 + 128 + 64 + 32 + 12 + 24 + 36 + 48) FFs are required.

  This means that 1488 FFs can be reduced compared to the case where the nonlinear filter 51 capable of obtaining an output signal equivalent to the nonlinear filter 80 requires 2600 FFs. Therefore, the circuit scale can be reduced, and cost saving and space saving can be realized.

  By the way, the present invention can also be applied to a nonlinear filter that processes an input signal by dividing it into X band components. In other words, the nonlinear filter 80 in FIG. 30 is a configuration example when X = 5.

  When the input signal is processed by dividing it into X band components, when the nonlinear filter is realized as an electronic circuit, (2X-1 + 2X-2 +... + 512 + 256 + 128 + 64 + 32 + 12 + 24 + 36 + 48 +... +12 (X-2) +12 (X-1)) FF is required.

  The present invention can be applied to any apparatus that handles image signals, such as a video camera, a digital still camera, a printer, a display, and a computer.

  In the present specification, the steps describing each flowchart include not only processes performed in time series according to the described order, but also processes executed in parallel or individually even if not necessarily performed in time series. Is also included.

It is a block diagram which shows the structural example of the image signal processing apparatus which emphasizes parts other than an edge in the state which preserve | saved the steep edge in an image. It is a figure which shows the image signal input into the epsilon filter of FIG. 1, and the output image signal. It is a figure which shows an example of the tap used with the epsilon filter of FIG. It is a figure for demonstrating operation | movement of the epsilon filter of FIG. It is a figure which shows an example of the image signal before a filter process. It is a figure which shows the image signal output from an epsilon filter corresponding to the image signal shown by FIG. It is a figure which shows the image signal output from an epsilon filter corresponding to the image signal shown by FIG. It is a figure which shows an example of the image signal before a filter process. FIG. 9 is a diagram illustrating an image signal output from an ε filter with the image signal illustrated in FIG. 8 as an input. FIG. 9 is a diagram illustrating an image signal output from an ε filter with the image signal illustrated in FIG. 8 as an input. It is a block diagram which shows the structural example of a nonlinear filter. It is a figure which shows the tap set with the nonlinear filter of FIG. It is a figure which shows the tap coefficient used with the nonlinear filter of FIG. It is a flowchart explaining the filtering process by the nonlinear filter of FIG. It is a figure for demonstrating the determination method of a weighting coefficient. It is a figure which shows the image signal output from the nonlinear filter of FIG. 11 by using the image signal shown in FIG. 5 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 11 by using the image signal shown in FIG. 5 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 11 by using the image signal shown in FIG. 5 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 11 by using the image signal shown in FIG. 5 as an input. It is a block diagram which shows the other structural example of a nonlinear filter. It is a figure which shows the tap of the interval 4 set by the narrowband process part of FIG. It is a figure which shows the tap of the interval 2 set by the middle band process part of FIG. It is a figure which shows the tap of the interval 1 set by the wideband process part of FIG. It is a figure which shows the tap coefficient used with the nonlinear filter of FIG. It is a figure which shows the image signal output from the nonlinear filter of FIG. 20 by using the image signal shown in FIG. 8 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 20 by using the image signal shown in FIG. 8 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 20 by using the image signal shown in FIG. 8 as an input. It is a figure which shows the image signal output from the nonlinear filter of FIG. 20 by using the image signal shown in FIG. 8 as an input. It is a block diagram which shows the structural example of the nonlinear filter which expanded the nonlinear filter of FIG. FIG. 30 is a block diagram illustrating a configuration example of a nonlinear filter to which the present invention is applied, which obtains an output signal similar to that of the nonlinear filter of FIG. FIG. 31 is a flowchart for explaining the operation of the nonlinear filter of FIG. 30. FIG.

Explanation of symbols

  80 nonlinear filter, 81 delay unit, 82 LPF synthesis unit, 83 delay unit, 84 LPF synthesis unit, 85 delay unit, 86 LPF synthesis unit, 87 delay unit, 88 LPF synthesis unit, 89 delay unit, 90 LPF synthesis unit, 91 , 92 control unit, 93 delay unit, 94 control unit, 95 delay unit, 96 control unit, 97 delay unit, 98 control unit, 99 delay unit

Claims (4)

In a signal processing device that adjusts the level of an input signal that is continuously arranged,
The input signals that are continuously arranged are sequentially designated as attention signals, and a plurality of neighboring signals are designated at predetermined intervals with reference to the attention signal from among the input signals. First tap generation means for generating a plurality of different taps composed of the neighborhood signals;
A plurality of calculating means for calculating a weighting coefficient based on the taps generated by the first tap generating means;
A plurality of smoothed signals are calculated by performing weighted averaging of a target signal and a plurality of neighboring signals constituting a tap using different coefficient sets, and the calculation is performed according to the weighting coefficient calculated by the calculation unit. A plurality of arithmetic means for combining a plurality of smoothed signals;
Delay means for delaying supply of the weighting coefficient calculated by the calculation means to the calculation means;
The output signal of the arithmetic means is sequentially designated as the attention signal, and a plurality of neighboring signals are designated at predetermined intervals from the output signal as a reference, and consists of the attention signal and the neighboring signal. And a second tap generation means for generating a tap. The signal processing apparatus according to claim 1, wherein the input signal is a pixel value of a pixel constituting an image. The signal processing apparatus according to claim 1, wherein the first and second tap generation units include a plurality of flip-flop circuits. In a signal processing method for adjusting the level of an input signal arranged continuously,
The input signals that are continuously arranged are sequentially designated as attention signals, and a plurality of neighboring signals are designated at predetermined intervals with reference to the attention signal from among the input signals. A first tap generation step of generating a plurality of different taps composed of neighboring signals;
A plurality of calculation steps for calculating a weighting coefficient based on the taps generated in the processing of the first tap generation step;
A plurality of smoothed signals are calculated by performing weighted averaging of the signal of interest and a plurality of neighboring signals constituting the tap using a plurality of different coefficient sets, and calculation is performed according to the weighting coefficient calculated in the processing of the calculation step. A plurality of calculation steps for combining the plurality of smoothed signals,
A delay step of delaying the supply of the weighting coefficient calculated in the calculation step to the calculation step;
The output signal in the processing of the calculation step is sequentially designated as a signal of interest, and a plurality of neighborhood signals are designated at predetermined intervals from the output signal with reference to the signal of interest, and the signal of interest and the neighborhood are designated. A signal processing method comprising: a second tap generation step of generating a tap composed of a signal.

Images