JP2005012641A - Block noise detecting device and block noise eliminating device using the same - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To get rid of block noises by detecting the block noises only from pixel data with a high degree of accuracy without using encoding information for the block noises that are prone to be generated in video pictures subjected to a compression process. <P>SOLUTION: An erroneous detection due to influence of random noises or the like that are prone to be generated when detecting the block noises only from the pixel data, is prevented by bringing a weighting process for spatial differences of video input signals by using a feature of the block noise to render the weighted results subjected to a cumulative process in a special direction and thereby detect the block noises, an additional cumulative process in a spatial direction (in a different direction) for the detected results, a process of adding hysteresis characteristics to the detected results in a frame (field) direction, and a process of performing a motion accommodation process to the detected results. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a video signal processing technique for compression distortion caused when a video signal is divided into a plurality of blocks and then encoded, transmitted, or recorded is decoded again.
[0002]
[Prior art]
In general, when data is transmitted or recorded, data compression is performed to reduce the amount of information. In recent years, with the increase in digital broadcasts featuring high-definition video and multi-channels in high-definition, the increase in long-term digital video recording due to the widespread use of DVD recorders and HDD recorders, and the increase in video distribution over the Internet, Its importance is increasing. Usually, a moving picture expert group (MPEG) method or the like is used for compressing a video signal. In this method, a video signal is divided into a plurality of blocks, discrete cosine transformation (hereinafter referred to as DCT) is performed, converted into DCT coefficients in a frequency domain, and then subjected to quantization processing and the like for encoding. . The larger the quantization step used in this quantization process, the greater the compression ratio. At this time, some values are discarded, resulting in quantization noise. Even if this is decoded, the original signal is completely reproduced. Will not be possible, causing image quality degradation. A typical example of such coding noise is block noise. This is because only the low frequency component is included in the block, and the boundary of the block appears in the video after decoding because the value of the low frequency component between adjacent blocks is different. As the compression rate increases, it becomes more conspicuous Become. In view of this, several methods have been proposed for detecting such block noise and removing (or making it inconspicuous) the block noise by filtering. For example, when decoding an encoded signal, the absolute value of the direct current component difference in the DCT coefficient and the distribution of each frequency component of the DCT coefficient, the quantization scale, the motion vector, etc. for the block to be processed and the adjacent block There is a method of detecting and removing block noise by using (see Patent Document 1). This is because the block noise is detected using coding information such as the difference in DC components between adjacent blocks and the quantization scale, which cause block noise in the decoding device, but high-precision detection can be expected. When the external input terminal is not used, the decoding apparatus is not provided and encoded information cannot be obtained, or the block size changes depending on the A / D or D / A frequency. Therefore, a method of detecting block noise from only pixel data without using the encoding information has been proposed. For example, there is a method of detecting block noise using a difference value between pixels across a block boundary and a difference value between pixels belonging to a block near the boundary (see Patent Document 2). This method makes it possible to detect block noise without using the encoded information, but in this case, the problem of false detection due to the influence of random noise or the like included in the signal becomes very large. Further, as another block noise detection method using only pixel data, for example, there is a method of extracting noise by extracting a block boundary in one entire frame (field). In this method, an HPF process is performed on a video signal, an absolute value thereof is accumulated, and a block boundary is extracted based on a periodicity as a result (see Patent Document 3). This method utilizes the fact that the encoded blocks are arranged uniformly in the spatial direction in the same source, and the boundary of the block can be extracted to some extent due to its periodicity, but the signal after the HPF processing is cumulatively processed. Accumulation of components due to pictures (edges) in the video results in problems with detection accuracy in a source that contains a lot of high-band video, and the signal itself after HPF processing (taken in absolute value) Accumulation increases the circuit scale.
[0003]
[Patent Document 1]
JP 2001-204029 (paragraph 0032)
[Patent Document 2]
JP 2002-199403 A (paragraph 0012)
[Patent Document 3]
JP 2000-50275 (paragraph 0011)
[0004]
[Problems to be solved by the invention]
If block noise detection is performed using coding information such as the difference in DC components between adjacent blocks that causes block noise and quantization scale, high-precision detection can be expected. For example, when using an external input terminal of a TV As described above, this method cannot be applied when the decoding apparatus is not provided and encoded information cannot be obtained, or when the block size changes depending on the frequency of A / D or D / A. Therefore, a method of detecting block noise only from pixel data without using encoded information can be considered, but in this case, the problem of erroneous detection due to the influence of random noise included in the signal becomes very large. In addition, when the signal after HPF processing is accumulated using only pixel data and block noise is detected based on the periodicity, the component due to the pattern (edge) in the original image is also accumulated, so that the high-band image can be obtained. In many sources, there is a problem in detection accuracy, and the signal scale after HPF processing (accumulated absolute value) is accumulated, so that the circuit scale becomes large.
[0005]
The present invention solves these problems, and performs weighting using the feature of block noise on the spatial difference for the video input signal and cumulatively processes the weighted result in the spatial direction, thereby causing errors due to the effects of random noise. While preventing detection, an increase in circuit scale is suppressed, and highly accurate block noise detection can be performed only from pixel data without using encoded information. Further, by constructing a block noise removal device incorporating this block noise detection device, highly accurate block noise removal can be performed without using encoded information.
[0006]
[Means for Solving the Problems]
For the video input signal obtained by dividing the video signal into blocks consisting of a plurality of pixels, encoding and decoding again, a horizontal difference processing unit for taking a difference from a pixel adjacent in the horizontal direction for each pixel and each pixel A weighting processing unit that performs block boundary determination using the difference signal from the horizontal difference processing unit for the reference pixel and a plurality of peripheral pixels in the horizontal direction as a reference pixel, and weights and outputs the result according to the level And a vertical accumulation processing section for accumulating and outputting the output signals from the weighting processing section in the vertical direction, a vertical accumulation window generating section for defining an area for accumulation processing in the vertical accumulation processing section, and the vertical accumulation processing section A horizontal block noise having a block boundary discriminating unit for determining a block boundary position based on the accumulated result in the block and outputting a horizontal block boundary signal Detector allows the horizontal direction of the block noise detection influence from only the pixel data to suppress the high-precision, such as random noise included in the video input signal by.
[0007]
Further, a horizontal accumulation processing unit that performs cumulative addition processing of horizontal block boundary signals output from the block boundary determination unit in a horizontal direction, a horizontal accumulation window generation unit that defines an area in which accumulation processing is performed by the horizontal accumulation processing unit, and the horizontal A pixel is detected by a horizontal block noise detection device to which a horizontal block boundary signal output from the block boundary determination unit is corrected using an accumulation result in the accumulation processing unit and an erroneous detection and removal unit that outputs a new horizontal block boundary signal is added. It is possible to prevent erroneous detection that is likely to occur in block noise detection using only data.
[0008]
In addition, for a video input signal obtained by dividing a video signal into blocks composed of a plurality of pixels, and encoding and decoding again, a vertical difference processing unit that takes a difference from a pixel adjacent in the vertical direction for each pixel and each pixel Weighting processing for determining a boundary of a block using difference signals from the vertical difference processing unit for the reference pixel and a plurality of peripheral pixels in the vertical direction of the reference pixel and weighting the result according to the level And a horizontal accumulation processing unit for accumulating and outputting output signals from the weighting processing unit in the horizontal direction, a horizontal accumulation window generating unit for defining an area to be accumulated in the horizontal accumulation processing unit, and the horizontal accumulation process Block having a block boundary discriminating unit that determines a block boundary position based on a cumulative result in the block and outputs a vertical block boundary signal To allow accurate vertical block noise detected from only the pixel data to suppress the influence of random noise included in the video input signal by noise detector.
[0009]
In addition, a vertical accumulation processing section for accumulating the vertical block boundary signal output from the block boundary determination section in the vertical direction in vertical block noise detection, and a vertical accumulation window for defining an area to be accumulated by the vertical accumulation processing section A vertical block including a generation unit and a false detection removal unit that corrects the vertical block boundary signal output from the block boundary determination unit using the accumulation result in the vertical accumulation processing unit and outputs the corrected signal as a new vertical block boundary signal The noise detection device can prevent erroneous detection that is likely to occur in block noise detection using only pixel data.
[0010]
Further, the horizontal block boundary signal or the vertical block boundary signal finally output in the horizontal block noise detection device and the vertical block noise detection device is held for a plurality of frames (fields), and the direction of the frame (field) is changed according to the change. By using a block noise detection device with a hysteresis characteristic addition unit that adds hysteresis characteristics to the frame, it is possible to prevent small fluctuations for each frame (field) due to false detection that is likely to occur in block noise detection using only pixel data. Is possible.
[0011]
Also, in each of the block noise detection devices, the absolute value of the frame difference output from the frame difference processing unit that takes a frame difference with respect to the video input signal and the frame difference processing unit are compared with an arbitrary threshold value. The binarization processing unit that performs binarization on a pixel-by-pixel basis, the horizontal / vertical accumulation processing unit that performs accumulation processing in the horizontal / vertical direction on the signals from the binarization processing unit, and the accumulation processing in the horizontal / vertical accumulation processing unit The block noise detection device based on the result of the motion determination unit and the motion determination unit that performs motion determination for each frame (field) using the cumulative window generation unit that defines the area to perform Motion adaptive processing for performing correction processing on the horizontal block boundary signal or the vertical block boundary signal finally output from Desired motion adaptive processing of the frame (field) rate for moving and still images of mixed video input signal by the noise detection apparatus obtained by adding the part becomes possible.
[0012]
In addition, the screen is divided into a plurality of blocks with respect to the video input signal obtained by dividing the video signal into blocks composed of a plurality of pixels, and encoding and decoding again, and each of the blocks is blocked by any one of the block noise detection devices. By performing noise detection, block noise detection for each screen block can be performed on a video input signal in which different video sources such as multi-screen synthesized video are mixed.
[0013]
Further, block noise detection is performed by dividing any one of the block noise detection devices or a plurality of the block noise detection devices into a video input signal obtained by dividing the video signal into blocks composed of a plurality of pixels, and encoding and decoding again. A buffer for holding the horizontal block boundary signal and the vertical block boundary signal output from the noise detection device for one frame (field), and the horizontal block boundary signal and the vertical block boundary signal held in the buffer for each pixel. By using a block noise removing device to which a filter processing unit that performs filter processing and removes block noise is added, block noise can be detected and removed only from pixel data.
[0014]
Further, in the block noise removing apparatus, a filter is obtained by adding a filter coefficient generating unit that generates a filter coefficient to be used in the filter processing unit based on information obtained from a block noise detecting process in the block noise detecting apparatus. The processing can be adaptively changed according to the block noise level.
[0015]
Further, in any one of the block noise removing apparatuses, the picture detecting unit adds a picture detecting unit that determines whether each pixel is a high-band part from the video input signal. By controlling the filter processing based on the detection result of the above, it is possible to prevent portions other than block noise located near the block boundary from being blurred by the filter processing.
[0016]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0017]
(Embodiment 1)
First, the details of the invention according to claim 1 will be described.
[0018]
FIG. 1 shows a configuration diagram of the present invention.
[0019]
In the figure, reference numeral 101 denotes a video input signal obtained by dividing a video signal into blocks composed of a plurality of pixels, encoding and decoding again, and reference numeral 102 in the figure denotes a horizontal that takes a difference from a pixel adjacent in the horizontal direction for each pixel. A difference processing unit 103 in the figure is based on the characteristics of block noise using horizontal difference signals from the horizontal difference processing unit 102 for the reference pixel and a plurality of peripheral pixels in the horizontal direction with each pixel as a reference pixel. A weighting processing unit that performs block boundary determination and weights and outputs the result according to the level. In the figure, reference numeral 104 denotes a vertical accumulation that outputs the output signal from the weighting processing unit 103 by accumulating the output signal in the vertical direction. In the figure, reference numeral 105 denotes a vertical accumulation window generator for defining an area in which accumulation processing is performed by the vertical accumulation processing section 104. 6 is a block boundary determination unit for outputting a horizontal block boundary signal Zuchu 107 determines the boundary position of the block by the cumulative result of the vertical cumulative section 104.
[0020]
The operation will be described below with a specific example.
[0021]
For the video input signal 101, the horizontal difference processing unit 102 obtains a difference from the adjacent pixel in the horizontal direction for each pixel. For example, the difference from the signal delayed by 1T by the circuit of FIG. The weighting processing unit 103 uses the horizontal difference signal from the horizontal difference processing unit 102 to perform block boundary determination based on the characteristics of block noise, and weights and outputs the result according to the level. That is, the output from the weighting processing unit 103 is a value weighted at an arbitrary stage (m stage) for each pixel. In this case, for example, weighting is performed in two stages, that is, block noise characteristics, that is, the possibility of block noise (1 is output) and no block noise characteristics, that is, there is no possibility of block noise (output 0). think of. As a feature of block noise, only low frequency components are included in the block as shown in FIG. 3A, and the values of the low frequency components between adjacent blocks are different. That is, the level between pixels in both block 1 and block 2 That is, there is little fluctuation and there is a difference in the level (DC level) between block 1 and block 2. When horizontal difference processing is performed on such a signal, a distribution of difference levels as shown in FIG. 3B is obtained (here, absolute values are taken for the differences). With respect to such a distribution of difference levels, each pixel is used as a reference pixel, and a block boundary is determined based on block noise characteristics using a horizontal difference signal for a reference pixel and a plurality of peripheral pixels in the horizontal direction. In response to this, the results are weighted and output, but here, for example, two-stage weighting is performed by a method as shown in FIG. When the reference pixel is p20, the absolute value of the difference with respect to p20 (here, | p20−p17 |) is larger than threshold value 1 and smaller than threshold value 2, and each of the five left and right reference pixels as peripheral pixels. (In this case, p13 to p17 and p21 to p25), and it is determined that there is a possibility of block noise when the absolute value of the difference (p17 is | p17−p16 |) is less than or equal to the threshold value 1 1 is output as the determination result for p20, and 0 is output otherwise. Such weighting processing is performed for each pixel, and 0 or 1 is output. Of course, several other weighting methods are conceivable. For example, in the case of weighting in four stages, depending on the above conditions, the difference value between the maximum difference level for the peripheral pixels and the difference level for the reference pixel is also added. Weighting from 0 to 3 can also be performed. In the vertical accumulation processing unit 104, the weighting processing result in the weighting processing unit 103 is cumulatively added in the vertical direction. Here, since each pixel is weighted to 0 or 1, this is cumulatively added. The area where the accumulation process is performed is defined by the vertical accumulation window generation unit 105. For example, as shown in FIG. 4A, the outer square is the entire screen, and the part other than the hatched line is the accumulation process window. In this range, the weighted values are accumulated in the vertical direction. The block boundary determination unit 106 determines the position of block noise, that is, the block boundary position based on the accumulation result in the vertical accumulation processing unit 104. Here, for example, the threshold value for the accumulation result as shown in FIG. It is determined whether it is a block boundary by comparing the sizes of. That is, if the accumulated result is larger than the threshold value, the horizontal block boundary signal 107 is obtained by determining that the pixel position is a block boundary. As shown in FIG. 4C, the block boundary signals 107 are all set to the same position in the vertical direction at the horizontal pixel position, but there is no particular problem because the encoding processing blocks are uniformly arranged in the spatial direction. In this case, the detection result is used (accumulated in a buffer or the like) in the next frame (field) subjected to weighting processing to perform accumulation processing, but block noise is continuously generated in the frame (field) direction. There is no problem in practical use because it is often done. By performing the cumulative addition process on the weighted signals in this way, it becomes possible to detect horizontal block noise while suppressing an increase in circuit scale from only pixel data and suppressing the influence of random noise.
[0022]
(Embodiment 2)
Next, the details of the invention according to claim 2 will be described.
[0023]
FIG. 5 shows a configuration diagram thereof. 201 to 206 in the figure are the same as 101 to 106 in FIG. 1 in (Embodiment 1). In the figure, reference numeral 207 denotes a horizontal accumulation processing unit for accumulating the horizontal block boundary signal output from the block boundary determination unit 206 in the horizontal direction, and 208 in the figure defines an area where the horizontal accumulation processing unit 207 performs accumulation processing. A horizontal accumulation window generation unit 209 in the figure corrects the horizontal block boundary signal output from the block boundary determination unit 206 using the accumulation result in the horizontal accumulation processing unit 207 and outputs it as a new horizontal block boundary signal 210. This is a false detection removal unit.
[0024]
The operation will be described below with a specific example.
[0025]
A horizontal block boundary signal is output from the block boundary discriminating unit 206 through the same process as in the first embodiment. The horizontal accumulation processing unit 207 accumulates this boundary signal in the horizontal direction. Specifically, for example, as shown in FIG. 4B, the number of pixels in which the accumulation result of the vertical accumulation processing unit 204 is larger than the threshold value. In this case, the cumulative result is 8. At this time, accumulation processing is performed in the area of the horizontal accumulation window defined by the horizontal accumulation window generator 208. Then, the erroneous detection removal unit 209 corrects the horizontal block boundary signal output from the block boundary determination unit 206 based on the accumulation result of the horizontal accumulation processing unit 207 and outputs it as a new horizontal block boundary signal. For example, if the erroneous detection removal unit 209 makes a determination based on the magnitude relationship between the cumulative result of the horizontal cumulative processing unit 207 and an arbitrary threshold value (if the cumulative result is smaller than the threshold value, it is determined that a false detection occurs), FIG. In the case of b), when the threshold value in the false detection removal unit 209 is set to 4, the accumulation result of the horizontal accumulation processing unit 207 is 8, which is larger than the threshold value, so the horizontal block output from the block boundary determination unit 206 The boundary signal is output as it is as a new horizontal block boundary signal 210. When the accumulation result of the horizontal accumulation processing unit 207 is 2, for example, it is smaller than the threshold value 4, so it is determined that there is a possibility of erroneous detection, and the horizontal block boundary signal output from the block boundary determination unit 206 is Output after being corrected that there is no block boundary. The reason why such processing is performed is that block noise is generally generated in a certain amount in the horizontal direction to some extent, so if the number of horizontal block boundaries is about 1 or 2, it is a false detection. This is because there is a high possibility. As described above, by adding the accumulation process in the horizontal direction to the horizontal block noise detection apparatus in the first embodiment, it is possible to reduce erroneous detection in block noise detection.
[0026]
(Embodiment 3)
Next, the details of the invention according to claim 3 will be described.
[0027]
FIG. 6 shows a configuration diagram thereof. In the figure, 301 is a video input signal obtained by dividing the video signal into blocks composed of a plurality of pixels, and encoding and decoding again. 302 in the figure is a vertical that takes the difference between adjacent pixels in the vertical direction for each pixel. Reference numeral 303 in the figure designates each pixel as a reference pixel, based on the characteristics of block noise using vertical difference signals from the vertical difference processing unit 302 for the reference pixel and a plurality of peripheral pixels in the vertical direction thereof. A weighting processing unit that performs block boundary determination and weights and outputs the result according to the level. In the figure, reference numeral 304 denotes a horizontal accumulation that outputs the output signal from the weighting processing unit 303 by accumulating in the horizontal direction. 305 in the figure is a horizontal accumulation window generator for defining an area in which accumulation processing is performed by the horizontal accumulation processor 304, and 3 in the figure 6 is a block boundary determination unit for outputting a vertical block boundary signal Zuchu 307 determines the boundary position of the block by the cumulative result of the horizontal cumulative section 304. The specific operation is omitted here because it is an extension of the horizontal block noise detection method in (Embodiment 1) in the vertical direction. As a result, the vertical block boundary signals 307 are all set to the same position in the horizontal direction at the vertical pixel position, but there is no particular problem because the encoding processing blocks are uniformly arranged in the spatial direction. By performing the cumulative addition process on the weighted signals in this way, it becomes possible to detect vertical block noise while suppressing an increase in circuit scale from only pixel data and suppressing the influence of random noise.
[0028]
(Embodiment 4)
Next, the details of the invention according to claim 4 will be described.
[0029]
FIG. 7 shows a configuration diagram thereof. 401 to 406 in the figure are the same as 301 to 306 in FIG. 6 in (Embodiment 3). In the figure, reference numeral 407 denotes a vertical accumulation processing unit that performs vertical addition processing on the vertical block boundary signal output from the block boundary determination unit 406 in the vertical direction, and reference numeral 408 in the figure defines an area in which the vertical accumulation processing unit 407 performs accumulation processing. A vertical accumulation window generation unit 409 in the figure corrects the vertical block boundary signal output from the block boundary determination unit 406 using the accumulation result in the vertical accumulation processing unit 407 and outputs it as a new vertical block boundary signal 410. This is a false detection removal unit. The specific operation is omitted here because it is an extension of the horizontal block noise detection method in (Embodiment 2) in the vertical direction. In this way, it is possible to reduce false detection in block noise detection by adding the accumulation processing in the vertical direction to the vertical block noise detection apparatus in (Embodiment 3).
[0030]
(Embodiment 5)
Next, the details of the invention according to claim 5 will be described.
[0031]
FIG. 8 shows a configuration diagram thereof. In the figure, reference numeral 501 denotes a horizontal block boundary signal or vertical block boundary signal that is finally output in the block noise detection apparatus according to claim 1, claim 2, claim 3, and claim 4. A block boundary signal or vertical block boundary signal 501 is held for a plurality of frames (fields), and a hysteresis characteristic is added in the frame (field) direction according to the change to output a new horizontal block boundary signal or vertical block boundary signal 503. This is a hysteresis characteristic adding section.
[0032]
The operation will be described below with a specific example.
[0033]
Here, a case is considered in which a hysteresis characteristic adding unit 502 is added to the block noise detection device in (Embodiment 1). For example, it is assumed that the horizontal block boundary signal finally output from the block noise detection apparatus in (Embodiment 1) is 1 (is a horizontal block boundary) or 0 (not a horizontal block boundary) for each pixel position. It is assumed that a detection result such as The left side of FIG. 9 is the horizontal block boundary signal 501 that is finally output from the horizontal block noise detection device, and the right side of FIG. 9 is the new horizontal block boundary signal that is output with the hysteresis characteristic added by the hysteresis characteristic addition unit 502. 503. Consider the case where the horizontal block boundary signal is detected from the nth frame (field) as n + 1, n + 2,... The hysteresis characteristic is added for the purpose of preventing the boundary signal from fluctuating every frame (field) due to erroneous detection (in this case, 1 and 0 are changed). For example, there is a block boundary from a state without a block boundary. Is changed on condition that the horizontal block boundary signal 501 is 1 in 3 frames (fields) continuously. Further, in the case of changing from a state with a block boundary to a state without a block boundary, it is a condition that the horizontal block boundary signal 501 is 0 continuously for two frames (fields). That is, in this case, a buffer that holds the horizontal block boundary signal 501 for two lines is used, and the determination is made based on the boundary signal of 2 and 1 frame (field) delay in this buffer and the boundary signal for the current frame (field). In the case of FIG. 9, block boundary signals are detected in order starting from the nth frame (field). However, in the n + 1 and n + 2th frames (fields), since the above condition is not three consecutive 1s, the right side of FIG. As a final block boundary signal 503, all pixels 0 are output. In the left side of FIG. 9, the pixels p1 and p8 for the first time in the (n + 3) th frame (field) become 1 continuously for 3 frames (fields), so p1 and p8 in the final block boundary signal 503 as indicated by (1) on the right side of FIG. 1 is output for. Further, on the left side of FIG. 9, only one frame (field) is 0 in the (n + 5) th frame (field) with respect to the pixel p1, but in this case, it does not become 0 continuously for the two frames (field) of the above condition. As in 2 ▼, the value of the previous frame (field) is held. On the other hand, for the pixel p8 on the left side of FIG. 9, it becomes 0 continuously in the n + 5 and n + 6th frames (fields), so in the n + 6th frame (field), the final result is shown in (3) on the right side of FIG. 0 is output as the block boundary signal 503. When such a hysteresis characteristic is added, the detection result is delayed by several frames (field) from the actual, but block noise often occurs continuously in the frame (field) direction, so there is no problem in practical use. . Of course, when the state with a block boundary changes to the state without a boundary, it is possible to change the boundary signal immediately to prevent image quality deterioration due to filter processing. By adding such hysteresis characteristics, it is possible to prevent small fluctuations for each frame (field) due to erroneous detection that is likely to occur in block noise detection using only pixel data.
[0034]
(Embodiment 6)
Next, the details of the invention according to claim 6 will be described.
[0035]
FIG. 10 shows a configuration diagram thereof. In the figure, reference numeral 601 denotes a video input signal to be input to the block noise detecting device according to claim 1, claim 2, claim 3, claim 4 and claim 5, and reference numeral 602 denotes a frame of the video input signal 601. A frame difference processing unit that takes a difference. In the figure, reference numeral 603 denotes a binarization that outputs a 1 or 0 depending on the size of the pixel by comparing the frame difference value in the frame difference processing unit 602 and an arbitrary threshold value for each pixel. A processing unit 604 in the figure is a horizontal / vertical accumulation processing unit that performs accumulation processing in the horizontal / vertical direction on a signal output from the binarization processing unit 603, and 605 in the figure is a horizontal / vertical accumulation processing unit 604. Numeral 606 in the figure is a motion determination unit that performs a motion determination using the result of accumulation in the horizontal / vertical accumulation processing unit 604. In the figure, reference numeral 607 denotes a horizontal block boundary signal or a vertical block boundary signal that is finally output in the block noise detection apparatus according to claim 1, claim 2, claim 3, claim 4, and claim 5. In the figure, reference numeral 608 denotes a new horizontal block boundary signal or vertical block which is subjected to correction processing on the horizontal block boundary signal or vertical block boundary signal 607 depending on whether it is a moving image or a still image based on the motion determination result in the motion determination unit 606. It is a motion adaptive processing unit that outputs a boundary signal 609.
[0036]
The operation will be described below with a specific example.
[0037]
The frame difference processing unit 602 takes the frame difference of the video input signal 601 and inputs it to the binarization processing unit 603. Here, an arbitrary threshold value is compared with the absolute value of the frame difference for each pixel, and 1 is output if the absolute value is greater than the threshold value, and 0 is output if it is smaller. Then, in the horizontal / vertical accumulation processing unit 604, the binarized signal is cumulatively added to pixels in the range of the accumulation window defined by the accumulation window generation unit 605. The motion determination unit 606 performs motion determination using the accumulation result in the horizontal / vertical accumulation processing unit 604. Here, for example, the motion determination is performed by comparing with the accumulation result using an arbitrary threshold value. If the accumulated result is larger than the threshold value, it is determined that the movement is made, and if it is smaller, 1 is outputted, and if it is smaller, 0 is outputted. The motion adaptation processing unit 608 corrects the horizontal block boundary signal or the vertical block boundary signal 607 based on the motion determination result of 1 or 0. For example, if the motion determination result is 0, that is, a still image, a horizontal block boundary signal or a vertical block boundary signal in the previous frame (field) is output (holds the boundary signal) or output as no block boundary, and the motion determination result is 1, that is, a moving image If so, the input horizontal block boundary signal or vertical block boundary signal 607 is output as it is. The reason for this processing is that block noise is likely to occur when there is a lot of motion as a video, that is, when there is a large amount of information in compression. This is to prevent false detection. By performing motion determination in this manner, desired motion adaptation processing at a frame (field) rate can be performed on a video input signal in which a moving image and a still image are mixed.
[0038]
(Embodiment 7)
Next, the details of the invention according to claim 7 will be described.
[0039]
According to the present invention, a screen is divided into a plurality of blocks, and block noise is detected in each block using the block noise detection device according to claim 1, claim 2, claim 3, claim 4, claim 5, and claim 6. Detection is performed. For example, as shown in FIG. 11A, it is conceivable to divide the screen into two blocks in the vertical direction and apply the block noise detection device according to claim 1 respectively. In order to reduce the circuit scale, it is possible to share some circuits in a plurality of block noise detection devices. For example, when two block noise detection devices according to the first aspect are used, the configuration shown in FIG. In the figure, reference numeral 701 denotes a video input signal obtained by dividing the video signal into blocks composed of a plurality of pixels, encoding and decoding again, and reference numeral 702 in the figure indicates a horizontal that takes a difference between adjacent pixels in the horizontal direction for each pixel. A difference processing unit 703 in the figure is based on block noise characteristics using horizontal difference signals from the horizontal difference processing unit 702 for the reference pixel and a plurality of peripheral pixels in the horizontal direction with each pixel as a reference pixel. A weighting processing unit that performs block boundary determination and weights and outputs the result according to the level, and reference numeral 704 in the figure denotes a first that outputs the output signal from the weighting processing unit 703 by performing cumulative addition processing in the vertical direction. 705 in the figure is a first vertical accumulation window that defines an area in which accumulation processing is performed by the first vertical accumulation processing unit 704. 706 in the figure is a first block boundary discriminating section that determines the boundary position of the block based on the accumulation result in the first vertical accumulation processing section and outputs the horizontal block boundary signal 1 of 707 in the figure. A middle 708 is a second vertical accumulation processing unit that outputs an output signal from the weighting processing unit 703 by accumulating in the vertical direction, and 709 in the figure performs accumulation processing by the second vertical accumulation processing unit 708. Reference numeral 710 in the figure determines the boundary position of the block based on the accumulation result in the second vertical accumulation processing section, and outputs the horizontal block boundary signal 2 of 711 in the figure. A second block boundary determination unit. These operations are the same as those in (Embodiment 1), but the horizontal difference processing unit 702 in the figure and the weighting processing unit 703 in the figure are shared, and the vertical accumulation process and the boundary determination process are performed separately, respectively. We are reducing the scale. By dividing the screen into a plurality of blocks and performing block noise detection separately for each block, block noise detection for each screen block can be performed for video input signals mixed with different video sources such as multi-screen composite video. It becomes possible.
[0040]
(Embodiment 8)
Next, the details of the invention according to claim 8 will be described.
[0041]
In the present invention, block noise detection is carried out by detecting block noise by combining a single block noise detecting device or a plurality of block noise detecting devices according to claims 1, 2, 3, 4, 5, 6, and 7. It is a block noise removing apparatus that can detect and remove block noise from only pixel data by performing filter processing according to the result. FIG. 12 shows an example of the configuration. In the figure, reference numeral 801 denotes a video input signal obtained by dividing a video signal into blocks composed of a plurality of pixels, encoding and decoding again, and reference numeral 802 in the figure indicates a horizontal that takes a difference between adjacent pixels in the horizontal direction for each pixel. A difference processing unit 803 in the figure is based on the characteristics of block noise using horizontal difference signals from the horizontal difference processing unit 802 for the reference pixel and a plurality of peripheral pixels in the horizontal direction with each pixel as a reference pixel. 1 is a first weighting processing unit that performs block boundary determination and weights and outputs the result according to the level, and 804 in the figure cumulatively adds the output signal from the first weighting processing unit 803 in the vertical direction. Reference numeral 805 in the figure denotes a first vertical accumulation processing unit that defines an area in which accumulation processing is performed by the first vertical accumulation processing unit 804. In the figure, reference numeral 806 denotes a first block boundary determination unit that determines a block boundary position based on the accumulation result in the first vertical accumulation processing unit 804 and outputs a horizontal block boundary signal. Is a first horizontal accumulation processing unit that performs horizontal addition processing on the horizontal block boundary signal output from the first block boundary determination unit 806 in the horizontal direction, and reference numeral 808 in the figure denotes accumulation processing performed by the first horizontal accumulation processing unit 807. Is a first horizontal accumulation window generation unit that defines an area for performing the horizontal block, and reference numeral 809 in the figure denotes a horizontal block output from the first block boundary determination unit 806 using the accumulation result in the first horizontal accumulation processing unit 807 This is a first erroneous detection and removal unit that corrects the boundary signal and outputs a new horizontal block boundary signal. 810 in the figure is output from the first erroneous detection and removal unit 809. A first hysteresis characteristic adding unit that holds a flat block boundary signal for a plurality of frames (fields), adds a hysteresis characteristic in the frame (field) direction according to the change, and outputs a new horizontal block boundary signal. Perform horizontal block noise detection. In the figure, reference numeral 811 denotes a vertical difference processing unit that takes a difference from a pixel adjacent in the vertical direction for each pixel, and reference numeral 812 in the figure denotes each of the pixels as a reference pixel for the reference pixel and a plurality of peripheral pixels in the vertical direction. 813 is a second weighting processing unit that performs block boundary determination based on block noise characteristics using the vertical difference signal from the vertical difference processing unit 811 and weights and outputs the result according to the level. Is a second horizontal accumulation processing unit that outputs the output signal from the second weighting processing unit 812 in a cumulative direction in the horizontal direction, and 814 in the figure is the second horizontal accumulation processing unit 813. A second horizontal accumulation window generator defining a region to be performed. In the figure, reference numeral 815 denotes a block boundary position based on an accumulation result in the second horizontal accumulation processor 813. 2 is a second block boundary discriminating unit that outputs a vertical block boundary signal, and a second block boundary discriminating unit 816 in the figure is a second block that cumulatively adds the vertical block boundary signals output from the second block boundary discriminating unit 815 in the vertical direction. A vertical accumulation processing unit 817 in the figure is a second vertical accumulation window generation unit that defines an area in which accumulation processing is performed by the second vertical accumulation processing unit 816, and 818 in the figure is a second vertical accumulation process. 819 is a second false detection and removal unit that corrects the vertical block boundary signal output from the second block boundary determination unit 815 using the accumulated result in the unit 816 and outputs a new vertical block boundary signal. Holds the vertical block boundary signal output from the second false detection removal unit 818 for a plurality of frames (fields) and changes the hysteresis in the frame (field) direction according to the change. By adding a scan characteristics is a second hysteresis characteristic adding unit for outputting a new vertical block boundary signal These perform vertical block noise detection. In the figure, reference numeral 820 denotes a frame difference processing unit that takes a frame difference of the video input signal 801, and reference numeral 821 in the figure indicates that the frame difference value in the frame difference processing unit 820 is compared with an arbitrary threshold value for each pixel. A binarization processing unit that outputs 1 or 0 depending on the size, 822 in the figure is a horizontal / vertical accumulation processing unit that performs accumulation processing in the horizontal / vertical direction on a signal output from the binarization processing unit 821, In the figure, reference numeral 823 denotes an accumulation window generating section for defining an area to be accumulated in the horizontal / vertical accumulation processing section 822. 825 in the figure is output from the first hysteresis characteristic adding unit 810 in the horizontal block noise detection process based on the motion determination result in the motion determination unit 824. 1 is a first motion adaptation processing unit that performs a correction process on a horizontal block boundary signal according to whether it is a moving image or a still image, and outputs a new horizontal block boundary signal. 825 is a first buffer for holding the horizontal block boundary signal obtained in 825 for one frame (field). Reference numeral 827 in the figure denotes the first buffer in the vertical block noise detection process based on the motion determination result in the motion determination unit 824. 2 is a second motion adaptation processing unit that performs a correction process on the vertical block boundary signal output from the hysteresis characteristic adding unit 819 and outputs a new vertical block boundary signal. The middle 828 is a second buffer for holding the vertical block boundary signal obtained by the second motion adaptation processing unit 827 for one frame (field). In the figure, reference numeral 829 indicates that the block boundary (block noise) is filtered by the horizontal block boundary signal and the vertical block boundary signal held in the first buffer 826 and the second buffer 828 to remove block noise. The filter processing unit generates the output signal 830.
[0042]
The operation will be described below with a specific example.
[0043]
802 to 806 in the figure are the same as the horizontal block noise detection example in (Embodiment 1), and 807 to 809 in the figure are the horizontal blocks in (Embodiment 2). This is the same as the noise detection example, and a hysteresis characteristic is added to the output from the first false detection removal unit 809 in the figure as in the example in (Embodiment 5). 811 to 815 in the figure are the same as the vertical block noise detection example in (Embodiment 3), and 816 to 818 in the figure are blocks in the vertical direction in (Embodiment 4). This is the same as the noise detection example, and a hysteresis characteristic is added to the output from the second false detection removal unit 818 in the figure as in the example in (Embodiment 5). Further, reference numerals 820 to 824 in the figure are blocks for performing motion determination as described in (Embodiment 6), and according to the result, the first motion adaptation processing unit 825 (for horizontal block boundary) and the first block are used. Similarly, in the second motion adaptation processing unit 827 (for vertical block boundary), motion adaptation processing in units of frames (fields) as in (Embodiment 6) is performed, and the results are respectively obtained in the first buffer 826 and the second buffer 826. It is held in the buffer 828. The filter processing unit 829 performs filter processing on the video input signal 801 using the horizontal block boundary signal and the vertical block boundary signal held in the first buffer 826 and the second buffer 828. For example, by performing appropriate filter processing on the pixels sandwiching the block boundary and several adjacent pixels, a steep waveform caused by the difference in DC level between adjacent coding blocks that causes block noise can be made smoother. Block noise can be removed by changing to a waveform. This makes it possible to detect and remove block noise from only pixel data.
[0044]
(Embodiment 9)
Next, details of the invention according to claim 9 will be described.
[0045]
According to the present invention, in the block noise removing apparatus according to the eighth aspect, filter processing for removing block noise using information obtained from the block noise detection process is adaptively changed according to the block noise level. FIG. 13 shows an example of the configuration. 901 to 928 in the figure are the same as blocks 801 to 828 in the figure in FIG. 12 shown in (Embodiment 8). In the figure, reference numeral 929 denotes a filter coefficient generation unit that generates a desired filter coefficient for each pixel in accordance with the horizontal and vertical difference levels in the horizontal difference processing unit 902 and the vertical difference processing unit 911. In the figure, 930 denotes a first buffer. The block boundary (block noise) is filtered by the horizontal block boundary signal and the vertical block boundary signal held in 926 and the second buffer 928. At this time, the filter coefficient generated by the filter coefficient generation unit 929 is It is a filter processing unit that generates an output signal 931 from which block noise is removed by performing adaptive filter processing according to the block noise level.
[0046]
The operation will be described below with a specific example.
[0047]
A horizontal block boundary signal and a vertical block boundary signal are obtained by the method described in (Embodiment 8) from 902 in the figure to 928 in the figure. The filter coefficient generation unit 929 generates a desired filter coefficient for each pixel according to the horizontal and vertical difference levels in the horizontal difference processing unit 902 and the vertical difference processing unit 911. For example, the stronger the difference, the stronger the filter The coefficient is determined so that. Then, the filter processing unit 930 performs filter processing on the block noise (block boundary) using this coefficient, so that the filter processing can be adaptively performed according to the block noise level. Here, the filter coefficient is determined according to the difference level in both the horizontal and vertical directions, but it is also possible to generate the filter coefficient based on the accumulated result after the weighting process, for example.
[0048]
(Embodiment 10)
Next, the details of the invention according to claim 10 will be described.
[0049]
According to the present invention, in the block noise removing apparatus according to any one of claims 8 and 9, it is determined whether or not each pixel is a high-band part from the video input signal, and a filter process for removing block noise is controlled based on the result. Thus, it is possible to prevent portions other than the block noise located near the block boundary from being blurred by the filter processing. FIG. 14 shows an example of the configuration. In the figure, reference numerals 1001 to 1029 are the same as blocks 901 to 929 in FIG. 13 shown in (Embodiment 9). In the figure, reference numeral 1030 denotes a picture detection unit that determines whether or not each pixel is a high-band part from the video input signal 1001. In the figure, 1031 denotes a horizontal block boundary held in the first buffer 1026 and the second buffer 1028. An output signal 1032 from which block noise is removed by performing adaptive filter processing according to the block noise level using the filter coefficient generated by the filter coefficient generation unit 1029 with respect to the block noise (block boundary) by the signal and the vertical block boundary signal. However, the filter processing can be partially limited by the detection result of the pattern detection unit 1030 at this time.
[0050]
The operation will be described below with a specific example.
[0051]
A horizontal block boundary signal and a vertical block boundary signal are obtained by the method described in (Embodiment 8) from 1002 in the figure to 1028 in the figure. The filter coefficient generation unit 1029 generates filter coefficients by the method described in (Embodiment 9). The picture detection unit 1030 determines from the video input signal 1001 whether or not each pixel is a high-band part. For example, if a desired HPF process is performed and the result is a level equal to or higher than a certain threshold value, the filter process is not performed. A control signal is sent to the filter processing unit. Even if block noise (block boundary) is detected in the filter processing unit 1031, the filter processing is controlled for each pixel by the control signal (no filter processing, weak filter processing, etc.), and a high-band portion not caused by block noise Can be prevented from being blurred by the filtering process.
[0052]
【The invention's effect】
As described above, according to the present invention, the spatial difference of the video input signal is weighted using the feature of block noise, and the result of the weighting is cumulatively processed in the spatial direction to perform block noise detection. Furthermore, random noise can be reduced by performing cumulative processing in the spatial direction (in another direction), adding hysteresis characteristics to the detection result in the frame (field) direction, and performing motion adaptation processing on the detection result. It is possible to prevent erroneous detection due to influences and the like, and to suppress an increase in circuit scale, and to detect block noise with high accuracy from only pixel data without using encoded information. Furthermore, by constructing a block noise removal device using this block noise detection, and further using adaptive filter processing based on the block noise level and filter control based on the detection result of the high band part not caused by block noise It is possible to remove block noise with high accuracy without using it.
[Brief description of the drawings]
FIG. 1 is a diagram showing a block noise detection apparatus according to a first embodiment of the present invention.
FIG. 2 is a diagram showing an example of horizontal difference processing according to Embodiment 1 of the present invention.
FIG. 3 is an explanatory diagram of weighting processing according to Embodiment 1 of the present invention.
FIG. 4 is an explanatory diagram of vertical accumulation processing according to Embodiment 1 of the present invention.
FIG. 5 is a diagram showing a block noise detection apparatus according to a second embodiment of the present invention.
FIG. 6 is a diagram showing a block noise detection apparatus according to a third embodiment of the present invention.
FIG. 7 is a diagram showing a block noise detection apparatus according to a fourth embodiment of the present invention.
FIG. 8 is a diagram showing a block noise detection apparatus according to a fifth embodiment of the present invention.
FIG. 9 is an explanatory diagram for adding hysteresis characteristics according to Embodiment 5 of the present invention;
FIG. 10 is a diagram showing a block noise detection apparatus according to a sixth embodiment of the present invention.
FIG. 11 is an explanatory diagram of a block noise detection device according to a seventh embodiment of the present invention.
FIG. 12 is a diagram showing a block noise removing apparatus according to Embodiment 8 of the present invention.
FIG. 13 is a diagram showing a block noise removing apparatus according to Embodiment 9 of the present invention.
FIG. 14 is a diagram showing a block noise removing apparatus according to Embodiment 10 of the present invention.
[Explanation of symbols]
101 Video input signal
102 Horizontal difference processing unit
103 Weighting processing unit
104 Vertical accumulator
105 Vertical accumulation window generator
106 Block boundary discriminator
107 Horizontal block boundary signal

Claims (10)

For the video input signal obtained by dividing the video signal into blocks consisting of a plurality of pixels, encoding and decoding again, a horizontal difference processing unit for taking a difference from a pixel adjacent in the horizontal direction for each pixel and each pixel A weighting processing unit that performs block boundary determination using the difference signal from the horizontal difference processing unit for the reference pixel and a plurality of peripheral pixels in the horizontal direction as a reference pixel, and weights and outputs the result according to the level And a vertical accumulation processing unit for accumulating and outputting the output signal from the weighting processing unit in the vertical direction, a vertical accumulation window generating unit for defining a region for accumulation processing in the vertical accumulation processing unit, and the vertical accumulation processing unit A block boundary discriminator that determines the boundary position of the block based on the accumulated result and outputs a horizontal block boundary signal, and includes it in the video input signal. Block noise detecting apparatus capable of high-precision horizontal direction of the block noise detection from only the pixel data to suppress the influence of random noise. 2. The block noise detection apparatus according to claim 1, wherein a horizontal accumulation processing unit that performs horizontal addition processing on a horizontal block boundary signal output from the block boundary determination unit and a region in which accumulation processing is performed by the horizontal accumulation processing unit are defined. A horizontal accumulation window generation unit that corrects a horizontal block boundary signal output from the block boundary determination unit using a result of accumulation in the horizontal accumulation processing unit, and outputs a new horizontal block boundary signal; The block noise detection apparatus according to claim 1, wherein erroneous detection that is likely to occur in block noise detection using only pixel data is prevented. For a video input signal obtained by dividing a video signal into blocks composed of a plurality of pixels, encoding and decoding again, a vertical difference processing unit that takes a difference from a pixel adjacent in the vertical direction for each pixel and each pixel A weighting processing unit that performs block boundary determination using the difference signal from the vertical difference processing unit for the reference pixel and a plurality of peripheral pixels in the vertical direction as the reference pixel, and weights and outputs the result according to the level And a horizontal accumulation processing unit that outputs an output signal from the weighting processing unit in a cumulative direction in the horizontal direction, a horizontal accumulation window generation unit that defines a region in which the accumulation processing is performed in the horizontal accumulation processing unit, and the horizontal accumulation processing unit A block boundary discriminator that determines the boundary position of the block based on the accumulated result and outputs a vertical block boundary signal, and includes it in the video input signal. Block noise detecting apparatus capable of only from the precise vertical block noise detection pixel data by suppressing the influence of random noise. 4. The block noise detection apparatus according to claim 3, wherein a vertical accumulation processing unit that vertically adds vertical block boundary signals output from the block boundary determination unit and a region in which accumulation processing is performed by the vertical accumulation processing unit are defined. A vertical accumulation window generation unit that corrects a vertical block boundary signal output from the block boundary determination unit using a result of accumulation in the vertical accumulation processing unit, and outputs a new vertical block boundary signal; The block noise detection apparatus according to claim 3, wherein erroneous detection that is likely to occur in block noise detection using only pixel data is prevented. 5. The block noise detection apparatus according to claim 1, 2, 3 and 4, wherein the horizontal block boundary signal or the vertical block boundary signal finally output is held for a plurality of frames (fields). A hysteresis characteristic addition unit that adds hysteresis characteristics in the frame (field) direction according to changes is added to prevent small fluctuations for each frame (field) due to false detection that is likely to occur in block noise detection using only pixel data. The block noise detection apparatus according to claim 1, 2, 3, and 4, which can be performed. In the block noise detection device according to any one of claims 1, 2, 3, 4 and 5, a frame difference processing unit for obtaining a frame difference with respect to the video input signal and a frame difference processing unit The absolute value of the frame difference output is compared with an arbitrary threshold value, and binarization processing is performed for each pixel according to the magnitude thereof, and the signals from the binarization processing unit are accumulated in the horizontal and vertical directions. Motion discrimination for each frame (field) using a horizontal / vertical accumulation processing unit that performs processing, an accumulation window generation unit that defines an area to be accumulated in the horizontal / vertical accumulation processing unit, and an accumulation result in the horizontal / vertical accumulation processing unit The horizontal block boundary signal or the previous signal output from the block noise detection device according to the result of the motion determination unit and the motion determination unit A motion adaptation processing unit that performs correction processing on a vertical block boundary signal is added to enable desired motion adaptation processing at a frame (field) rate for a video input signal mixed with moving images and still images. The block noise detection apparatus according to claim 1, claim 2, claim 3, claim 4, and claim 5. 4. The screen is divided into a plurality of blocks for a video input signal obtained by dividing the video signal into blocks made up of a plurality of pixels, encoding and decoding again, and claim 1 or claim 2 or claim 3 in each block. Alternatively, block noise is detected for each screen block by using the block noise detection device according to claim 4 or 5 or 6 for a video input signal in which different video sources such as multi-screen synthesized video are mixed. Block noise detection device that enables detection. Claims 1, 2, 3, 4, 5, and 6 for video input signals obtained by dividing a video signal into blocks of a plurality of pixels, encoding, and decoding again. The block noise detection device according to claim 7 is used alone or in combination to perform block noise detection, and the horizontal block boundary signal and the vertical block boundary signal output from the block noise detection device are held in one frame (field). And a filter processing unit that performs block processing for each pixel by the horizontal block boundary signal and the vertical block boundary signal held in the buffer and removes block noise, thereby adding block noise only from pixel data. Block noise that can be detected and removed Removed by the device. 9. The block noise removing apparatus according to claim 8, further comprising: a filter coefficient generating unit that generates a filter coefficient to be used in the filter processing unit based on information obtained from a block noise detecting process in the block noise detecting unit, and blocking the filter processing The block noise removing apparatus according to claim 8, wherein the block noise removing apparatus can be adaptively changed according to a noise level. 10. A block noise removal apparatus according to claim 8, wherein a picture detection unit for determining whether or not each pixel is a high-band part from a video input signal is added, and detection by the picture detection unit is performed in the filter processing unit. 10. The block noise removing apparatus according to claim 8, wherein a part other than the block noise located near the block boundary can be prevented from being blurred by the filter process by controlling the filter process according to the result.

Images