JP4650372B2 - Video processing circuit - Google Patents

Description

  The present invention relates to a video processing circuit, and in particular, protects textures due to minute amplitudes, reduces quantization noise caused by digital encoding while maintaining the texture of video, and improves quantization accuracy, thereby achieving smooth gradation. The present invention relates to a video processing circuit that can be obtained.

  With recent advances in digital encoding technology, digital video signals such as DVD (Digital Versatile Disc) and high-definition broadcasting have become widespread. This digital video signal is generally a video signal quantized with 8 bits, but it has also been pointed out that in recent television devices where high definition and a large screen are advanced, quantization accuracy is insufficient with 8 bits. Specifically, there is a case where 8-bit 1 gradation is recognized by the human eye and smooth gradation cannot be expressed. In particular, encoding is performed by MPEG2 (Moving Picture Experts Group 2), which is an encoding method that performs processing in blocks. The converted digital video signal has a problem that when the decoded video signal is displayed, a boundary between gradations may be seen in a block shape.

  In recent television apparatuses, the gamma curve of a video signal is often changed or contour enhancement is performed in order to improve the image quality. In this case, even an 8-bit signal may have a quantization accuracy of 7 bits or less depending on the gradation, and an improvement in the quantization accuracy is desired for smooth gradation expression.

  In order to solve such a problem, an input video signal having a quantization bit number of 8 bits is bit-extended to 10 bits, and the 10-bit video signal and a signal smoothed by a low-pass filter (low-pass filter) Only when the difference absolute value is less than or equal to a predetermined value, a video processing circuit that generates a smooth gradation by generating an intermediate gradation by adding and outputting the low-order 2 bits of the low-pass filter output to a 10-bit video signal. Conventionally disclosed (see, for example, Patent Document 1).

  In addition, while focusing mainly on reducing quantization noise, the introduction of a nonlinear low-pass filter and coring processing as a video processing circuit that can be diverted for the purpose of improving quantization accuracy from 8 bits. A technique is disclosed in which natural images without pseudo contours are obtained without edges being smoothed (see, for example, Patent Document 2).

JP-A-8-237669 JP 2005-354521 A

  However, in the conventional video processing circuit described in Patent Document 1, a low-pass filter (low-pass filter) is applied to the quantization boundary when the quantization accuracy of the input video signal is improved from 8 bits to 10 bits. As this low-pass filter, a 3 × 3 pixel two-dimensional low-pass filter is used, but this is not large enough to make the quantization boundary invisible. In order to make the quantization boundary invisible and obtain a smooth gradation, it is actually necessary to use a filter having a larger number of taps.

  By the way, in recent years, high definition and large screens of display devices have been advanced, and video signal processing for improving image quality has been advanced. In particular, the expression of fine video signals is said to lead to “texture” and “realism”. In a high-definition and large-screen display device, a subtle texture is expressed by the behavior of a video signal in units of 1 bit even in a flat region without an edge with a large amplitude. The conventional video processing circuit described in Patent Document 1 can solve the shortage of quantization accuracy by using a low-pass filter with a large number of taps. There is a problem that fine textures such as texture, sand grains, and hair are covered. In addition, when using a low-pass filter with a large number of taps, there is a problem in that a smooth filter output cannot be obtained due to the influence of a pixel value different from the surroundings such as a contour.

  With respect to this problem, the nonlinear low-pass filter used in the conventional video processing circuit described in Patent Document 2 eliminates the influence by not using the calculation if the difference from the processing target pixel is equal to or greater than a threshold value. Yes. However, since the switching is discontinuous before and after the threshold value, the accuracy of the low-pass filter may be lowered, and there is a problem that the circuit scale is large.

  The present invention has been made in view of the above points, and provides a video processing circuit capable of obtaining a smooth gradation with improved quantization accuracy without generating an unnatural portion in an image. Objective.

  Another object of the present invention is to provide a video processing circuit capable of improving the quantization accuracy more reliably with a small circuit scale.

  In order to achieve the above object, the present invention provides an input video signal composed of pixels having a first quantization bit number with a second quantization bit number with a bit accuracy higher than the first quantization bit number. A video processing circuit for converting an output video signal composed of pixels to attenuate a high-frequency component of an input video signal and to set a quantization bit accuracy of each pixel corresponding to the input video signal to a first quantization bit Filter means for generating a first video signal converted from a number to a second quantization bit number, a first video signal, and an input video signal that is a basis for generating the first video signal. Subtracting means for obtaining a second video signal by subtracting with the accuracy of the second number of quantization bits, and a horizontal direction and a vertical direction of a screen, which are predetermined for the first predetermined number of pixels and configured by the input video signal Move in units of pixels in at least one of the directions In the first detection area to be generated, a flatness evaluation value that is larger as the degree of luminance change between adjacent pixels in the first detection area is smaller is calculated, and the calculated flatness evaluation value is Flatness evaluation value detecting means for performing the process of setting the flatness evaluation value of the first processing target pixel, which is a pixel located at the center of one detection area, each time the first detection area is moved in units of pixels; The second predetermined number of pixels is set in advance, and is moved in units of pixels in synchronization with the first detection region in at least one of the horizontal direction and the vertical direction of the screen formed by the input video signal. In the detection region, an interval between pixels in which the degree of luminance change between adjacent pixels in the second detection region is larger than a predetermined first threshold is detected as an edge portion, and the detected edge portion and the second edge portion are detected. Located in the center of the detection area Every time the second detection region is moved in units of pixels, the distance to the second processing target pixel, which is a pixel, is calculated and the calculated distance is used as the edge distance evaluation value of the second processing target pixel. The edge distance evaluation value detection means to be performed and either one of the flatness evaluation value and the edge distance evaluation value are referred to, and the second threshold value is variably controlled to be smaller as the referred value is smaller. In addition, when the absolute value of the second video signal obtained by the subtracting means is greater than or equal to the second threshold value, a value obtained by subtracting the second threshold value from the second video signal is output, and the absolute value of the second video signal is output. When the value is smaller than the second threshold, at least a process of outputting a predetermined value smaller than a value obtained by subtracting the second threshold from the second video signal, the first detection area and the second detection area as a pixel unit Coring means and filter that are used every time And adding means for adding the first video signal output from the means and the signal output from the coring means with the accuracy of the second quantization bit number to obtain an output video signal. To do.

  In the present invention, in the video signal portion having a large texture value with a small flatness evaluation value or edge distance evaluation value or a video signal portion near the edge, the second threshold value in the coring means is variably controlled to a small value. Therefore, since the second video signal as the input signal is output as it is or without being attenuated from the coring means, it is finally composed of pixels of the second number of quantization bits. In the adding means for generating the output video signal, the high-frequency component attenuated by the filter means is added as it is to the first video signal output from the filter means, and on the other hand, an image in a region far from the edge Since the second threshold value is not reduced in the signal part or the video signal part having a small texture value, the coring means is included in the first video signal output from the filter means. Et signals or not at all added, or not added little.

  Here, the coring means selects an evaluation value having a smaller value out of the flatness evaluation value and the edge distance evaluation value, and variably controls the second threshold value using the selected evaluation value. It is characterized by doing.

  In the present invention, the filter means in the above invention is preset to the third predetermined number of pixels, and is moved in units of pixels in at least one of the horizontal direction and the vertical direction of the screen formed by the input video signal. In the processing target area to be processed, a third processing target pixel that is a pixel to be processed and is located at the center of the processing target area, and a plurality of tap pixels that are pixels other than the third processing target pixel For each of the plurality of calculated difference values, if the difference value is less than or equal to the third threshold value, the difference value is used as an output value, and the difference value is greater than the third threshold value. If the difference value is less than or equal to the fourth threshold value and less than or equal to the fourth threshold value, a value obtained by subtracting the third threshold value from the difference value is used as an output value. Threshold A comparison process using a predetermined value smaller than the value obtained by subtracting as an output value is performed, and a plurality of output values obtained as a result of the comparison process are added and averaged with the accuracy of the second quantization bit number. By performing the addition process of the value obtained by the averaging and the third processing target pixel for each pixel constituting the input video signal, the pixel is configured with pixels of the second quantization bit number and has a high frequency. The filter means outputs the first video signal with the component attenuated.

  According to the present invention, the threshold value of the coring means is varied based on the pixel texture value and the evaluation value corresponding to the distance from the edge, and the signal output from the coring means, that is, attenuated by the filter means. By varying the amount of high-frequency components, the filter effect by the filter means is almost zero in the video signal part with a large texture value and a minute amplitude, and the video signal part near the edge. The intensity of the filter effect, which is removed by the filter effect by the filter means, can be switched adaptively, protecting the video expression of minute amplitudes and edges, and improving the quantization accuracy without generating an unnatural part in the image An improved smooth gradation can be obtained.

  According to the present invention, more accurate filter output can be obtained by the filter means, and the improvement in quantization accuracy can be made more reliable, and the flatness evaluation value or the edge distance evaluation value is the texture amount or the edge distance. Therefore, there is no discontinuity, no unnatural part is generated in the image, and the mounting with a circuit scale smaller than that of the conventional circuit can be realized.

  Next, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows a block diagram of an embodiment of a video processing circuit according to the present invention. In this figure, an input video signal having a quantization bit number of 8 bits is supplied to a non-linear low pass filter (LPF) 1 and a high frequency component having a quantization bit number of 10 bits is attenuated (band-limited). The video signal is converted into a video signal and supplied to the delay circuit 2, the flatness evaluation circuit 4, and the edge distance circuit 5, respectively.

  The delay circuit 2 performs a delay process on the input video signal for phase matching with the pixel to be processed of the output video signal of the nonlinear LPF 1. The subtracter 3 performs a process of subtracting a video signal with a high frequency component attenuated with a quantization bit number of 10 bits output from the nonlinear LPF 1 from a video signal with a quantization bit number of 8 bits output from the delay circuit 2. A difference signal is obtained and supplied to the coring circuit 7. The above difference signal corresponds to the high frequency component of the input video signal attenuated by the nonlinear LPF 1.

  In the coring circuit 7, one of the flatness evaluation value obtained by the flatness evaluation circuit 4 described later and the edge distance evaluation value obtained by the edge distance evaluation circuit 5 is selected by the switch 6 and inputted. Then, the size of the coring for the differential signal is adaptively changed according to the evaluation value. The coring result by the coring circuit 7 and the band limited video signal output from the nonlinear LPF 1 are added by the adder 8 and output as a final video signal having a quantization bit number of 10 bits.

Next, the nonlinear LPF 1 will be described with reference to FIGS. FIG. 2 shows a block diagram of an embodiment of the nonlinear LPF 1 in FIG. As shown in FIG. 2, the nonlinear LPF 1 has a pair of _n different (n is a natural number of 2 or more) differentiators 11 _{1 to} 11 _n provided in parallel, and one pair on each output side of the differentiators 11 _{1 to} 11 _n . 1, the non-linear operation units 12 _{1 to} 12 _n provided corresponding to 1, the addition averager 13 that adds and averages the output signals of the non-linear operation units 12 _{1 to} 12 _n , the output signal of the addition averager 13 and the processing target An adder 14 that outputs the processing target image signal to which the low-pass filter characteristic is added by adding the pixels.

Next, the operation of this nonlinear LPF 1 will be described. In the input video signal having the quantization bit number of 8 bits, the processing target pixel located at the center of the detection area is supplied to the differentiators 11 _{1 to} 11 _n in units of pixels, while each of the detection area other than the processing target pixels in the detection area. After the pixels are separately supplied to 11 _{1 to} 11 _n as tap pixels and the difference is taken, the difference value is supplied to the nonlinear arithmetic units 12 _{1 to} 12 _n to perform the nonlinear calculation. As shown in FIG. 3, when the input differential value (absolute value) is less than or equal to the first threshold value | lmt |, the non-linear operation is used as the output value as it is, and the input differential value (absolute value) is When the value is larger than the first threshold value | lmt | and equal to or smaller than the second threshold value 2 × | lmt |, an output value is obtained by subtracting | lmt | from the input difference value, and the input difference value (absolute value) is 2. When x is larger than | lmt |, the output value is “0” or at least a predetermined value set smaller than the value obtained by subtracting | lmt | from the input difference value.

An 8-bit value obtained by nonlinear calculation by each of the nonlinear calculation units 12 _{1 to} 12 _n is within a certain area by the addition averager 13 (for example, a predetermined number of pixels of 10 pixels or more in the vertical and horizontal directions). Are averaged. The 10-bit average value obtained by the adder 13 is added to the pixel to be processed by the adder 14 and output as a video signal having a quantization bit number of 10 bits to which a nonlinear low-pass filter characteristic is added. .

With the configuration shown in FIG. 2, the value of | lmt | is the upper limit for the input of the adder / averager 13, and the numerical value to be handled is smaller than that of adding and averaging the pixel values themselves, so that the circuit scale can be reduced. Further, the non-linear operation units 12 _{1 to} 12 _n that perform non-linear operation on the difference value between the pixel to be processed and the tap pixel have symmetrical input / output characteristics with respect to the threshold value | lmt | In the case of | lmt | or more, it is easy to obtain an average value that changes more smoothly than a processing method that does not use for calculation.

  This will be described with reference to FIG. FIG. 4A shows the output of the nonlinear filter used in the invention described in Patent Document 1, and FIG. 4B shows the output of the nonlinear LPF 1 of the present embodiment. The left side of the pixel to be processed is a case where the difference from the pixel to be processed is within the threshold value, and the right side slightly exceeds the threshold value, and shows a pattern having a large step near the quantization boundary.

  As shown in FIG. 4A, in the conventional nonlinear filter used in the invention described in Patent Document 1, a pixel whose pixel value differs from the processing target pixel by 1 mt or more is not used for the filter output at all. Is not symmetric about the quantization boundary.

  On the other hand, in the nonlinear filter according to the present embodiment, as shown in FIG. 4B, since a pixel whose pixel value is different from the processing target pixel by 1 mt or more is also considered, the filter output is symmetric with respect to the quantization boundary. It will be described with reference to FIG. 13 that the object of the present invention is easily achieved when an output symmetric with respect to the quantization boundary is obtained. FIG. 13A is an extraction of the quantization boundary portion of FIG. FIG. 13B shows an output waveform obtained by applying the nonlinear filter of the present embodiment that can obtain an output symmetric with respect to the quantization boundary. In FIG. 13, the difference between (A) and (B) is the high frequency component (C), and coring is performed on (C). When coring within the range of the dotted line in (C), the high frequency component becomes zero as shown in FIG. FIG. 13E shows the final output obtained by adding FIG. 13D to the waveform of FIG. As is apparent from the figure, the quantization boundary is smooth.

  On the other hand, FIG. 13F shows an output waveform when a conventional nonlinear filter that cannot obtain an output symmetrical with respect to the quantization boundary is applied. The high-frequency component (G), which is the difference between FIG. 13A and FIG. 13F, has a waveform that is distorted positive and negative, as is apparent from FIG. If coring is performed on the waveform (G) within the same range as (C), the positive side cannot be fully cored as shown in FIG. Accordingly, the final output obtained by adding the waveform shown in FIG. 13H to FIG. 13F is as shown in FIG. As is clear from FIG. 6I, when a conventional nonlinear filter that cannot obtain an output symmetric with respect to the quantization boundary is used, the quantization boundary portion cannot be completely smoothed.

  As described above, according to the nonlinear filter of the present embodiment, the quantization boundary can be made smoother than that of the conventional nonlinear filter.

  Returning again to FIG. In this embodiment, the difference between the input video signal and the nonlinear LPF 1, that is, the high frequency component is input from the differentiator 3 to the coring circuit 7, and the output high frequency component is added again by the adder 8 with the output video signal of the nonlinear LPF 1. The final output video signal. That is, if the output is 0 as a result of the calculation in the coring circuit 7, the input video signal is output as it is.

  FIG. 5 is a characteristic diagram of an example of the coring circuit 7. As indicated by I in FIG. 5, "0" is input if the absolute value of the input is equal to or smaller than the coring value | cor |, and from the input if the absolute value of the input is greater than the coring value | cor |. Returns the value minus. In this input / output characteristic I, there is no large discontinuity and no unnatural portion is generated in the image, but the coring value | cor | is subtracted from all input signals exceeding the coring value | cor |. If this is the case, the high frequency components will be attenuated throughout the image and will be lost.

  Therefore, in this embodiment, in areas where it is not desired to smooth an image such as a fine texture portion or an edge, the coring value + cor and −cor of the coring characteristic in FIG. 5 are set to “0”, and the input and the output are equal. As an input / output characteristic indicated by II, an image is held as it is.

Next, the configuration of the flatness evaluation value circuit 4 in FIG. 1 will be described. FIG. 6 shows a block diagram of an embodiment of the flatness evaluation value circuit 4. 6, a horizontal direction of two adjacent pixels tap (x, y) fixed in the detection area of the input video signal and tap the (x + 1, y) are supplied to a subtracter 40 _1, or even its horizontal two adjacent pixels tap (x + 1, y) and tap the (x + 2, y) is supplied to the subtractor 40 ₂ is supplied to the subtracter two adjacent pixels in the horizontal direction is not shown Similarly , Subtracters 40 ₁ , 40 ₂ ,... Extract the difference value between two adjacent pixels in the horizontal direction within a predetermined horizontal pixel range.

On the other hand, the longitudinal direction of two adjacent pixels tap (x, y) of the above the detection area of the input video signal and tap (x, y + 1) and are supplied to the subtracter 41 _1, or even adjacent in the vertical direction 2 pixel tap (x, y + 1) and tap (x, y + 2) and are supplied to the subtracter 41 ₂ is supplied similarly to the longitudinal direction of two adjacent pixels is not shown subtractor following the subtraction The difference values between two adjacent pixels in the vertical direction within a plurality of predetermined vertical pixel ranges are extracted from the devices 41 ₁ , 41 ₂ ,. Subtractor ₄₀ 1, ₄₀ 2, a difference value between two adjacent pixels taken from ..., the threshold comparator ₄₂ 1, ₄₂ 2 is supplied to the ... subtracter ₄₁ 1, 41 ₂ ,..., And the difference value between two adjacent pixels in the vertical direction is supplied to the threshold comparators 43 ₁ , 43 ₂ ,.

The threshold comparators 42 ₁ , 42 ₂ ,..., 43 ₁ , 43 ₂ ,... Compare a preset threshold with an input difference value, and “1” if the input difference value is equal to or greater than the threshold. If it is smaller than the threshold, “0” is output. The threshold used here is for eliminating the influence of noise, and is set to a very small value such as “4” corresponding to “1” of 8 bits in 10-bit processing. In other words, when 2 bits are added to the lower 8 bits to make 10 bits, when only 8 bits LSB is “1”, “0” is added to the lower 2 bits to make 10 bits. As for the value, only the third bit from the lower side is “1”, which corresponds to a value of “4” in 10 bits.

The output signals of the threshold comparators 42 ₁ , 42 ₂ ,... Are supplied to a detection circuit 44 composed of a plurality of one-pixel transmission period delay circuits T, where the amount of texture, that is, the complexity of the image pattern is determined. After being detected, it is supplied to the cumulative sum circuit 46 and added to obtain the cumulative sum of a certain area. Similarly, the output signals of the threshold comparators 43 ₁ , 43 ₂ ,... Are also accumulated after the complexity of the picture pattern of the image is detected by the detection circuit 45 including a plurality of one-pixel transmission period delay circuits T. 47 to be the cumulative sum of a certain detection area. The above-mentioned fixed detection area shifts (moves) in a fixed direction in units of pixels.

Here, for example, if the fixed detection area is 10 pixels vertically and 10 pixels horizontally for the sake of simplicity, the threshold comparators 42 ₁ , 42 ₂ ,. Is output by the detection circuit 44 and 10 comparison results of the nine horizontal direction adjacent pixel difference values are output in the vertical direction and are cumulatively added by the cumulative sum circuit 46. Similarly, the threshold comparators 43 ₁ , 43 ₂ ,... Output the comparison results of nine vertical adjacent pixel difference values, and the detection circuit 45 compares the nine vertical adjacent pixel difference values. Are output in the horizontal direction and are cumulatively added by the cumulative sum circuit 47. The cumulative sums obtained by the cumulative sum circuits 46 and 47 are added by an adder 48 and supplied to the nonlinear arithmetic circuit 49 as a texture evaluation value for nonlinear calculation.

  For example, the nonlinear arithmetic circuit 49 is set to have a characteristic in which the horizontal axis texture evaluation value shown in FIG. 7 is input and the vertical axis flatness evaluation value is output. If the input texture evaluation value is small, the nonlinear arithmetic circuit 49 is flat. Assuming that the flatness evaluation is high, and the texture evaluation value is large, it is determined that the flatness evaluation is small.

  The value of the texture evaluation value on the horizontal axis in FIG. 7 varies depending on the size of the detection area. An actual example is shown in FIG. In a video signal with a small level change as shown in FIG. 8A, the texture amount is small and the flatness evaluation value is large. In such an image region having a large flatness evaluation value, as will be described later with respect to the input video signal, coring processing having a large coring absolute value | cor | shown in FIG. The effect of the filter of the nonlinear LPF 1 is obtained by reducing the difference signal level added by the adder 8.

  In addition, in the video signal in which the level change as shown in FIG. 8B is small and the level change appears at short intervals, the texture amount is large, and the flatness evaluation is low. In such an image region having a small flatness evaluation value, the correlating process is not performed on the input video signal, and the difference signal level output from the coring circuit 7 and added by the adder 8 is reversed. By increasing the value, the filter action by the filter of the nonlinear LPF 1 is substantially stopped.

  Even when the quality of the original input video signal is poor and noise of a large amplitude is included, a signal as shown in FIG. 8B is obtained, and the filter effect cannot be obtained. However, in general, when there is a lot of noise, the gradation step is not noticeable. For this reason, the object of the present invention is to obtain a smooth gradation. Conversely, it is preferable to provide some kind of noise reduction circuit before the video processing circuit of the present invention. When the noise is eliminated, the video processing circuit of the present invention in which the quantization accuracy is increased and smoothed in the gradation portion and the texture portion is preserved works most effectively.

  However, only with the flatness evaluation, the edge portion of the video signal as shown in FIG. 8C has a small amount of texture and is judged to be flat. Based on the judgment result, coring is performed on the input video signal. If processing is performed to obtain the effect of the filter of the nonlinear LPF 1, the edge is lost. Therefore, in order to solve the above problem, in this embodiment, an edge distance evaluation circuit 5 is provided as shown in FIG.

FIG. 9 shows a block diagram of an example of the edge distance evaluation circuit 5. 9, two pixels adjacent in the horizontal direction in a predetermined two-dimensional detection area of the input video signal are supplied to the subtracters 50 ₁ , 50 ₂ ,..., And are input in the vertical direction in the two-dimensional detection area. Two adjacent pixels are supplied to the subtracters 51 ₁ , 51 ₂ ,. Thereby, the difference value between the adjacent pixels in the horizontal direction in the detection area is extracted from the subtracters 50 ₁ , 50 ₂ ,..., And the vertical direction in the detection area from the subtractors 51 ₁ , 51 ₂ ,. A difference value between adjacent pixels is extracted. Subtractor ₅₀ 1, ₅₀ 2, lateral adjacent-pixel difference value retrieved from ..., the threshold comparator ₅₂ 1, ₅₂ 2 is supplied to the ... subtracter ₅₁ 1, ₅₁ 2, The difference value between adjacent pixels in the vertical direction extracted from... Is supplied to the threshold value comparators 53 ₁ , 53 ₂ ,.

The threshold comparators 52 ₁ , 52 ₂ ,..., 53 ₁ , 53 ₂ ,... Compare a preset threshold value with the input difference value, and “1” if the input difference value is equal to or greater than the threshold value. If it is smaller than the threshold, “0” is output. The threshold used here is for detecting a signal that is clearly a contour component that is not caused by insufficient quantization or noise. For example, in 10-bit processing, the threshold is about “32” corresponding to “8” of 8 bits. Set to.

The output signals of the threshold comparators 52 ₁ , 52 ₂ ,... Are supplied to a detection circuit 54 composed of a plurality of one-pixel transmission period delay circuits T. Here, after the presence or absence of an edge in a certain area is detected, This is supplied to the OR circuit 56. Similarly, after each of the output signals of the threshold comparators 53 ₁ , 53 ₂ ,... Is detected by the detection circuit 55 including a plurality of one-pixel transmission period delay circuits T, the presence or absence of an edge in a certain area is detected. 57. The OR circuits 56 and 57 are each composed of eight OR circuits.

  That is, from the detection circuit 54, there are a total of eight horizontal directions, including three pixels in the horizontal direction around the pixel to be processed, one pixel in the horizontal direction outside the three pixels, and one pixel in the horizontal direction outside the pixel. Edges are detected independently from each other within the pixel range, and "1" is output in parallel from the OR circuit corresponding to the pixel range in which the edge is detected, and "0" is output in parallel from the OR circuit corresponding to the pixel range in which no edge is detected. Is output. Similarly, from the detection circuit 55, a total of eight vertical pixels, including three pixels in the vertical direction centering on the pixel to be processed, one pixel in the vertical direction outside the above three pixels, and one pixel in the vertical direction outside the pixel, are provided. Edges are detected independently from each other in the direction pixel range, and a signal of “1” is output from the OR circuit corresponding to the pixel range in which the edge is detected, and “0” is output from the OR circuit corresponding to the pixel range in which no edge is detected. Output in parallel.

  The 8 parallel output signals respectively taken out from the OR circuits 56 and 57 are supplied to an OR circuit 58 composed of 8 OR circuits, where the output signals of the OR circuits in the corresponding pixel range are ORed. Here, the 8 parallel output signals respectively extracted from the OR circuits 56 and 57 are signals indicating whether or not there are edges in the eight horizontal pixel ranges and the eight vertical pixel ranges centering on the pixel to be processed. Therefore, as shown in FIG. 10, the eight output signals of the OR circuit 58 are signals indicating whether or not edges exist in eight pixel ranges that are equidistant from the processing target pixel indicated by the black circle. When there is an edge, the logical value of the logical sum output signal indicating the pixel range is “1”.

  Here, eight pixel ranges that are equidistant from the processing target pixel are defined as edge distances, and the edge distance value is incremented by 1 from “1” in order from the closest to the processing target pixel. The value of the edge distance of the pixel range farthest from the target pixel is “8”. The eight logical sum signals output in parallel from the OR circuit 58 and indicating whether or not there is an edge in each of the eight edge distance regions shown in FIG. 10 are supplied to the nonlinear arithmetic circuit 59.

  The non-linear operation circuit 59 determines the logical sum signal having the smallest edge distance among the logical sum signals having the value “1” among the output logical sum signals from the OR circuit 58 as the final edge distance. Further, for example, based on the characteristic that the edge distance on the horizontal axis shown in FIG. 11 is input and the evaluation value on the vertical axis is output, the evaluation value with a smaller value is obtained when the edge distance is closer, and the evaluation value with a larger value is obtained with longer distance. Is output as an edge distance evaluation value.

Returning to FIG. 1 again, the flatness evaluation value from the flatness evaluation circuit 4 and the edge distance evaluation value from the edge distance evaluation circuit 5 obtained as described above are held in phase. The evaluation value with the smaller value is selected by the switch circuit 6 and input to the coring circuit 7 as “coring strength”. In the coring circuit 7, the following formula: coring value = coring initial value × (coring strength / 4)
After the coring value is obtained by the above, the coring characteristic is given to the high frequency component from the subtracter 3 according to the coring characteristic shown in FIG. Here, the coring values calculated by the above equation are + cor and -cor in FIG.

  Therefore, when the pixel to be processed is in the vicinity of the edge and the edge distance is “1”, the edge distance evaluation value is the minimum value “0” from the characteristics of FIG. Is supplied to the coring circuit 7, the coring value (= ± cor) is “0” from the above equation, and the coring circuit 7 has a characteristic that the input indicated by II in FIG. It becomes.

  As a result, the adder 8 of FIG. 1 adds the high-frequency component attenuated by the nonlinear LPF 1 to the 10-bit band-limited video signal output from the nonlinear LPF 1 as it is through the coring circuit 7. A video signal having a quantization bit number of 10 bits that is not subjected to the filtering operation is extracted.

  Further, since the edge distance evaluation value increases as the processing target pixel moves away from the edge, the value of the coring strength supplied to the coring circuit 7 increases, and as a result, the coring value (| cor | Since the difference signal from the subtractor 3 corresponding to the high frequency component of the input video signal attenuated by the nonlinear LPF 1 is output as “0” by the coring circuit 7, the input level range is widened. As a result, the high frequency component added to the 10-bit band-limited video signal output from the nonlinear LPF 1 in the adder 8 of FIG. 1 decreases as the processing target pixel moves away from the edge, and the filtering action by the nonlinear LPF 1 filter is performed. Becomes stronger as the processing target pixel moves away from the edge.

  Next, a processing example by the video processing circuit of this embodiment will be described with reference to FIG. FIG. 12A shows an input video signal having a quantization bit number of 8 input to the nonlinear LPF 1 or the like, and has a texture portion a1 having a minute amplitude, a signal portion a2 having a quantization step, and an edge portion a3. . FIG. 12B shows a video signal waveform having a quantization bit number of 10 bits output from the nonlinear LPF 1 and a high frequency component is attenuated.

  FIG. 12C shows the waveform of the differential signal output from the subtractor 3, which is a signal corresponding to the high frequency component attenuated by the nonlinear LPF 1. FIG. 12D shows the flatness evaluation value, which is a small value in the texture portion a1 of the input video signal shown in FIG. FIG. 12E shows the edge distance evaluation value, which is small near the edge portion a3 of the input video signal shown in FIG. 12A and increases as the distance from the edge increases.

  FIG. 12F shows the coring strength in which the smaller one of the flatness evaluation value shown in FIG. 12D and the edge distance evaluation value shown in FIG. As can be seen from FIGS. 12D to 12F, the texture portion a1 of the input video signal shows the coring strength according to the flatness evaluation value, and the core according to the edge distance evaluation value at the edge portion a3 and its vicinity. Indicates ring strength.

  The difference signal (signal corresponding to the high-frequency component attenuated by the nonlinear LPF 1) in FIG. 12C is obtained by the coring circuit 7 having the characteristics shown in FIG. 5 that changes based on the coring value calculated according to the coring intensity. A signal waveform obtained by applying a coring characteristic to the waveform is a signal waveform shown in FIG.

  FIG. 12H shows the final output video obtained by adding the difference signal shown in FIG. 12G and the output band limited video signal of the nonlinear LPF 1 shown in FIG. Signals are shown. As can be seen from FIG. 12H, the flatness evaluation value is “0” in the texture portion a1 of the input video signal shown in FIG. 12A, and the edge distance evaluation value is “0” in the edge portion a3. Therefore, the coring strength is “0”, the coring operation of the coring circuit 7 is not performed, and the filtering action of the non-linear filter 1 is not performed. Therefore, the final output video signal is a texture portion of the input video signal. While the a1 and the edge portion a3 are held, only the flat step where the insufficient quantization accuracy is easily visible is smoothed as indicated by a2 ′.

  As described above, according to the present embodiment, the clear contour component is excluded from the filter processing target by “edge distance evaluation”, and the texture having a minute amplitude is excluded from the filter processing target by “flatness evaluation”. Therefore, it is possible to improve the quantization accuracy and obtain a smooth gradation without deteriorating the image quality by eliminating the side effects such as the outline being distorted and the texture being crushed.

  In addition, since each of the evaluation values changes stepwise according to the edge distance / texture amount, there is no discontinuity, and an unnatural portion or the like is not generated in the image. The fact that the edges and texture of the output video signal are not different from the input video signal means that the quantization accuracy will not improve, but it is the flat part that is easy to see the coarseness of the quantization accuracy due to human visual characteristics. The roughness of quantization accuracy is difficult to detect at edges and textures. If the purpose is to obtain a smooth gradation, there is no problem even if the edges and textures are not processed, and conversely, sharpness reduction and texture attenuation caused by filtering can be prevented. In order to improve the quantization accuracy of edges and textures, for example, a method such as disposing a contour emphasis circuit at the subsequent stage of the circuit of the present invention and expanding the quantization bit width of the output may be suitable.

  In addition, this invention is not limited to said embodiment, For example, although the coring value showed the case where it was set as 4 steps | paragraphs of evaluation, it may be possible to provide finer steps, such as 8 steps | paragraphs. Needless to say. Further, in the above embodiment, as shown in FIG. 11, the edge detection range covers the eight pixels around the pixel to be processed, but this detection range may be widened, in which case the nonlinearity is non-linear. Although the filter effect of the LPF 1 is reduced, a signal in the vicinity of the edge is stored, and the texture of the image is easily maintained. On the other hand, the detection range may be narrower than that of the above-described embodiment. In this case, the texture of the image is easily lost, but the filter effect of the nonlinear LPF 1 is easily obtained.

It is a block diagram of one embodiment of a video processing circuit of the present invention. It is a block diagram of an example of the nonlinear LPF in FIG. It is a figure which shows an example of the input-output characteristic of the nonlinear calculating part in FIG. It is a figure which compares and shows each example of the nonlinear filter output of a conventional circuit, and the output of the nonlinear LPF of the circuit of this invention. It is a figure which shows an example of the input-output characteristic of the coring circuit in FIG. It is a block diagram of an example of the flatness evaluation circuit in FIG. It is an input-output characteristic of an example of the nonlinear arithmetic circuit which converts a texture evaluation value into a flatness evaluation value. It is a figure which shows the pattern of an input signal. It is a block diagram of an example of an edge distance evaluation circuit. It is a figure which shows the definition of edge distance. It is an input-output characteristic of an example of the nonlinear arithmetic circuit which converts an edge distance into an edge distance evaluation value. It is a signal waveform diagram of each part of one embodiment of the video processing circuit of the present invention. It is a wave form diagram explaining that the quantization boundary can be smoothed compared with the conventional nonlinear filter by the nonlinear filter used by the circuit of the present invention.

Explanation of symbols

1 Nonlinear LPF (low pass filter)
2 Delay circuit 3, 40 ₁ , 40 ₂ , 41 ₁ , 41 _2, 50 ₁ , 50 ₂ , 51 ₁ , 51 ₂ Subtractor 4 Flatness evaluation circuit 5 Edge distance evaluation circuit 6 Switch circuit 7 Coring circuit 8, 14 adder 11 ₁ to 11 _n differentiator 12 ₁ to 12 _n nonlinear operator 13 averager 42 _{1, 42 2,} 43 _{1, 43 2,} 52 1, 52 _2, 53 1, 53 ₂ threshold comparator 44, 45 , 54, 55 Detection circuit 46, 47 Cumulative sum circuit 48, 59 Nonlinear operation circuit 56-58 OR circuit

Claims (3)

An input video signal composed of pixels of the first quantization bit number is converted into an output video signal composed of pixels of the second quantization bit number having a bit accuracy higher than the first quantization bit number. A video processing circuit,
A high frequency component of the input video signal is attenuated, and a quantization bit precision of each pixel corresponding to the input video signal is converted from the first quantization bit number to the second quantization bit number. Filter means for generating one video signal;
Subtracting means for subtracting the first video signal and the input video signal from which the first video signal was generated with accuracy of the second quantization bit number to obtain a second video signal When,
The first predetermined number of pixels is set in advance, and the first detection region is moved in units of pixels in at least one of the horizontal direction and the vertical direction of the screen formed by the input video signal. A flatness evaluation value that is larger as the degree of luminance change between adjacent pixels in the detection region is smaller is calculated, and the calculated flatness evaluation value is calculated by a pixel located at the center of the first detection region. Flatness evaluation value detection means for performing processing for setting the flatness evaluation value of a certain first processing target pixel each time the first detection region is moved in units of pixels;
A second predetermined number of pixels is set in advance, and is moved in units of pixels in synchronization with the first detection area in at least one of the horizontal direction and the vertical direction of the screen formed by the input video signal. In the two detection regions, pixels between which the degree of luminance change between adjacent pixels in the second detection region is larger than a predetermined first threshold are detected as edge portions, and the detected edge portions and A process of calculating a distance from a second processing target pixel that is a pixel located in the center of the second detection region, and setting the calculated distance as an edge distance evaluation value of the second processing target pixel, An edge distance evaluation value detection means that is performed each time the second detection region is moved in units of pixels;
One of the flatness evaluation value and the edge distance evaluation value is referred to, and the second threshold is variably controlled to be smaller as the referred value is smaller. When the absolute value of the obtained second video signal is greater than or equal to the second threshold value, a value obtained by subtracting the second threshold value from the second video signal is output, and the absolute value of the second video signal is output. When the value is smaller than the second threshold value, a process of outputting a predetermined value that is at least smaller than a value obtained by subtracting the second threshold value from the second video signal is performed using the first detection region and the second detection value. Coring means to be performed each time the region is moved in units of pixels;
Adding means for adding the first video signal output from the filter means and the signal output from the coring means with accuracy of the second quantization bit number to obtain the output video signal; A video processing circuit comprising:   The coring means selects an evaluation value having a smaller value out of the flatness evaluation value and the edge distance evaluation value, and variably controls the second threshold value using the selected evaluation value. The video processing circuit according to claim 1. The filter means includes
In the processing target area that is determined in advance as the third predetermined number of pixels and moves in units of pixels in at least one of the horizontal direction and the vertical direction of the screen formed by the input video signal, the center of this processing target area Difference values between a third processing target pixel that is a pixel to be processed and a plurality of tap pixels that are pixels other than the third processing target pixel, respectively, are calculated. For each difference value, if the difference value is less than or equal to the third threshold value, the difference value is used as an output value, and the difference value is greater than the third threshold value and greater than the third threshold value. When the difference value is less than or equal to the fourth threshold value, a value obtained by subtracting the third threshold value from the difference value is set as an output value. When the difference value is larger than the fourth threshold value, at least the third threshold value is calculated from the difference value. Deduction A comparison process using a predetermined value smaller than the output value as an output value is performed, and the plurality of output values obtained as a result of the comparison process are added and averaged with the accuracy of the second quantization bit number. By performing the addition process of the value obtained by the averaging and the third processing target pixel for each pixel constituting the input video signal, the pixel is configured with pixels of the second quantization bit number, And it is a filter means for outputting the first video signal in which the high frequency component is attenuated,
The video processing circuit according to claim 1, wherein the video processing circuit is a video processing circuit.

Images