JP2004056190A - Filter device - Google Patents

Abstract

Provided is a filter device capable of reducing or removing block distortion without deteriorating image quality of an interlaced decoded image without damaging high-frequency components of the decoded image.
A filter device performs a predetermined delay on a decoded image signal processed in a block unit, and a vertical block distortion determining unit that determines vertical block distortion of a block of the decoded image signal. 102, a vertical filter 103 for vertically filtering an image signal delayed by a predetermined amount in the line memory 101 based on the determination result obtained by the vertical block distortion determination unit 102, and an image signal output from the vertical filter 103. And a horizontal filter 105 that filters the image signal in the horizontal direction based on the determination result obtained by the horizontal block distortion determination unit 104.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to the field of digital video signal processing, and in particular, to block distortion caused by quantization noise in a decoded image of an encoding device that orthogonally transforms an image signal in block units and quantizes and encodes the coefficients. And a post-processing filter for making the filter inconspicuous.
[0002]
[Prior art]
FIG. 7 is a block diagram showing a configuration of a conventional apparatus for processing a video image signal. Referring to FIG. 7, a video signal recorded on a recording medium 10 such as a DVD passes through a system multiplex decoder 20, is decoded by a video decoder 30, and the decoded image is filtered by a filter device 40. , NTSC / PAL encoder 50 and output to a monitor (not shown).
[0003]
Video images can be represented by digital signals using a series of information bits. For certain bandwidth-limited lines, such as mobile and public telephone lines, low bit rate image coding is particularly beneficial for image transfer and communication. Therefore, the quality of the video image generated by the low bit rate encoding is a serious problem.
[0004]
In general, the quality of an encoded image is determined by the type of encoding technology used and the target bit rate. When the target bit rate is reduced, that is, when an image is compressed with high efficiency, image quality deterioration specific to various compression encoding techniques often occurs. For example, in the case of an encoding method in which image compression is performed in block units using DCT (Discrete Cosine Transform) or the like, block distortion in which block boundaries are discontinuous may occur.
[0005]
Block distortion is the most visually noticeable deterioration in image quality in a compression encoding technique that performs processing in units of blocks, and it is necessary to reduce or eliminate this in order to improve image quality. For this purpose, a conventionally performed method is to locally or globally perform a filtering process by smoothing on a decoded image including block distortion.
[0006]
FIGS. 9 and 10 are diagrams for explaining the filtering process of the block distortion due to the smoothing by the filter device 40 shown in FIG. FIG. 9 shows an example in which filtering by global smoothing is suitable, and FIG. 10 shows an example in which filtering by local smoothing is suitable. FIGS. 9A and 10A show the luminance values of the decoded image before the filter processing, and FIGS. 9B and 10B show the luminance values after the filter processing. In these figures, lines orthogonal to the block boundaries in the image are shown as representatives, and black circles represent the luminance levels of the pixels.
[0007]
As shown in FIGS. 9A and 10A, a difference in luminance level between the left and right of the block boundary, that is, block distortion in which the block boundary is discontinuous occurs. If there is such block distortion, the decoded image looks like a mosaic at that location, making it hard to see visually.
[0008]
Therefore, conventionally, as shown in FIG. 9 (B), the difference in luminance level at the block boundary is blunted by filtering processing using global smoothing, and block distortion is reduced by eliminating linear edges of luminance values. Alternatively, as shown in FIG. 10B, the difference in the luminance level at the block boundary is blunted by the filter processing based on local smoothing for pixels near the block boundary to reduce block distortion.
[0009]
FIG. 8 shows the configuration of a conventional filter device 40 that performs such processing. Referring to FIG. 8, a conventional filter device 40 receives a decoded image signal, a decoder 41, a delay circuit 42 for delaying the decoded image, and a decoded image and an image delayed by the delay circuit 42. And a control circuit 43 for controlling the filter 44 according to a control signal from the decoder 41.
[0010]
[Problems to be solved by the invention]
Conventional filtering has been performed as described above. However, if the filtering process using the two types of smoothing is performed in the vertical direction on the interlaced decoded image, when an object with a sharp outline moves up and down, the image quality may become double. Deterioration had occurred. Further, if the filtering process using the two types of smoothing is performed on all the block boundaries, there is a problem that the high-frequency components of the decoded image are removed and the image is entirely blurred.
[0011]
The present invention has been made in view of such a situation, and can reduce or remove block distortion without impairing the high-frequency components of a decoded image. It is an object of the present invention to provide a filter device capable of reducing or removing the noise.
[0012]
[Means for Solving the Problems]
A filter device according to the present invention includes a delay unit that performs a predetermined delay on a decoded image signal processed in units of blocks, and a first determination that determines block distortion in a first direction of a block of the decoded image signal. Means, filter means for filtering an image signal delayed by a predetermined amount by the delay means based on the determination result obtained by the first determination means in the first direction, and output from the first filter means. A second determining means for determining block distortion in a second direction intersecting the first direction of the image signal, and filtering the image signal in the second direction based on the determination result obtained by the second determining means. And second filter means for processing.
[0013]
In the present invention, the presence or absence of distortion is determined in advance in each direction for each of the first direction and a second direction that intersects with the first direction, and thereafter, the filtering process is performed in each direction. Therefore, since the filtering process is performed only when the distortion is large, all the high frequency components of the decoded image are not removed, and the filtering process can be performed only in the direction in which the image quality deteriorates in the interlaced image.
[0014]
As a result, it is possible to provide a filter device that can reduce or remove block distortion without impairing the high-frequency component of the decoded image, and can also reduce or remove block distortion without deteriorating the image quality of an interlaced decoded image.
[0015]
Preferably, the first direction is a vertical direction and the second direction is a horizontal direction.
More preferably, a means for judging whether or not a pixel near the block boundary has a low frequency before the second judging means in the horizontal direction is included.
[0016]
In order to determine the presence or absence of block distortion after determining whether or not the frequency is low, the presence or absence of block distortion is determined only when necessary.
[0017]
Note that the first and second determination units each include a unit that compares an absolute value of a difference between adjacent pixels and a predetermined first threshold value with respect to pixels near the block boundary, and a unit that compares two pixels adjacent to the block boundary with each other. Means for comparing the absolute value of the difference with a predetermined second threshold value; and calculating the absolute value of the difference between two pixels adjacent to the block boundary and the maximum value of the absolute value of the difference between two adjacent pixels near the block boundary. Means for comparing.
[0018]
More preferably, the first filter means or the second filter means performs a filtering process on a predetermined number of pixels around the block boundary.
[0019]
BEST MODE FOR CARRYING OUT THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
[0020]
FIG. 1 is a diagram corresponding to FIG. 8 of the related art, showing one embodiment of a filter device for reducing or removing block distortion according to the present invention.
[0021]
The configuration and operation of the filter device 100 will be described with reference to FIG. Referring to FIG. 1, filter device 100 includes a line memory 101, a vertical block distortion determination unit 102, a vertical filter 103, a horizontal block distortion determination unit 104, and a horizontal filter 105.
[0022]
Next, the operation will be described with reference to FIGS. Here, a description will be given of a process performed when a decoded image of an image compressed by an encoding device that performs orthogonal transform on an image signal in block units (for example, 8 × 8 blocks) and quantizes and encodes the coefficients is input. .
[0023]
FIG. 2 shows a block boundary and pixels near the block boundary when an input decoded image is an interlaced image in a certain field. Here, since processing is performed for each field in the case of an interlaced image, as shown in FIG. 2, when the input decoded image is an interlaced image, a block boundary exists every four pixels in the vertical direction, and in the horizontal direction. There is a block boundary every eight pixels. FIG. 3 shows vertical pixels when displayed as a frame image. In FIG. 3, v0, v1, v2, and v3 in FIG. 2 are pixels of line 0, 2, 4, and 6, respectively. In the next field, v0 ', v1', v2 ', and v3' are line 1, 3, 5, and 7 pixels. That is, v0, v1, v2, v3 and v0 ', v1', v2 ', v3' are data at the same position.
[0024]
The decoded image is input to the line memory 101, subjected to a predetermined delay, and sent to the vertical block distortion determination unit 102 and the vertical filter 103.
[0025]
The vertical block distortion determination unit 102 receives the signal delayed by a predetermined time in the line memory 101, and determines the presence or absence of block distortion according to the flowchart shown in FIG. Here, the predetermined delay refers to the next time. That is, in the vertical filter, the pixels to be filtered need to be at the same time. For example, in the case of filtering by v0, v1, and v2 in FIG. 2, v0 is delayed by two lines and v1 is delayed by one line so as to be at the same position in time.
[0026]
Here, the flowchart of FIG. 4 will be described. First, for pixels near the block boundary (v0 to v7 in FIG. 2), the absolute value of the difference between adjacent pixels is compared with a predetermined threshold value thv1 to determine whether or not the frequency is low (step 11, hereinafter step). Is omitted). For example, the absolute value of the difference between adjacent pixels in v0 to v7 in FIG. 2 can be expressed by the following d0, d1, d2, d4, d5, and d6.
[0027]
d0 = | v1-v0 |
d1 = | v2-v1 |
d2 = | v3-v2 |
d4 = | v5-v4 |
d5 = | v6-v5 |
d6 = | v7−v6 |
If all the absolute values (d0, d1, d2, d4, d5, d6) of the difference between the adjacent pixels are smaller than a predetermined threshold value thv1, the vicinity of the block boundary is flat, and it is determined that the frequency is low. judge.
[0028]
Next, the absolute value d3 = | v4-v3 | of the difference between two pixels (v3 and v4 in FIG. 2) adjacent to the block boundary is compared with a predetermined threshold value thv2, and two pixels adjacent to the block boundary are further compared. Is compared with the maximum value maxv1 (the maximum value of d0, d1, d2, d4, d5, d6) of the absolute value of the difference between two adjacent pixels near the block boundary to determine whether there is block distortion. Is determined (S12).
[0029]
That is, when the absolute value d3 of the difference between two pixels adjacent to the block boundary is larger than the threshold thv2, it is determined that the edge is included in the image. Also, when the absolute value d3 of the difference between two pixels adjacent to the block boundary is smaller than maxv1, it is determined that the edge is included in the image. When it is determined that the image is not an edge included in the image under the above two conditions, it is determined that there is block distortion.
[0030]
Here, the values such as the threshold value thv2 are empirical values determined by looking at actual image data input to the filter device 100. Therefore, it differs depending on the resolution and bit rate of the input image.
[0031]
In the vertical filter 103, a signal delayed by a predetermined time in the line memory 101 is input, and when the vertical block distortion determination unit 102 determines that there is block distortion, four pixels (v2 in FIG. To v5) is subjected to a filtering process (S13).
[0032]
The signal processed by the vertical filter 103 is input to the horizontal block distortion determination unit 104 and the horizontal filter 105. The horizontal block distortion determination unit 104 determines the presence or absence of block distortion for the input signal according to the flowchart shown in FIG. Here, the flowchart of FIG. 5 will be described. First, for the pixels near the block boundary (h0 to h7 in FIG. 2), the absolute value of the difference between adjacent pixels is compared with a predetermined threshold thh1 to determine whether or not the frequency is low (S21). For example, the absolute value of the difference between adjacent pixels at h0 to h7 in FIG. 2 can be expressed by the following D0, D1, D2, D4, D5, and D6.
[0033]
D0 = | h1-h0 |
D1 = | h2-h1 |
D2 = | h3-h2 |
D4 = | h5-h4 |
D5 = | h6-h5 |
D6 = | h7−h6 |
If all the absolute values (D0, D1, D2, D4, D5, D6) of the difference between the adjacent pixels are smaller than thh1, the vicinity of the block boundary is flat, so that it is determined that the frequency is low, and otherwise. In the case of, it is determined that the frequency is low.
[0034]
If it is determined that the frequency is low (YES in S21), the absolute value (D3 = | h4-h3 |) of the difference between two pixels (h3 and h4 in FIG. 2) adjacent to the block boundary and a predetermined threshold value thh2 And the maximum value maxh1 (D0, D1, D2, D4, D5, D6) of the absolute value of the difference between the two pixels adjacent to the block boundary and the absolute value of the difference between the two adjacent pixels near the block boundary. (Maximum value) is determined to determine whether there is block distortion (S22).
[0035]
If the absolute value D3 of the difference between the two pixels adjacent to the block boundary is larger than the predetermined threshold thh2, it is determined that the edge is included in the image. Also, when the absolute value D3 of the difference between two pixels adjacent to the block boundary is smaller than maxh1, it is determined that the edge is included in the image. Therefore, when it is determined that the image is not an edge included in the image under the above two conditions (YES in S22), it is determined that there is block distortion.
[0036]
When it is determined that the pixel row near the block boundary has a non-low frequency (NO in S21), the absolute value (D3) of the difference between two pixels (h3 and h4 in FIG. 2) adjacent to the block boundary and a predetermined value are determined. The threshold value thh3 is compared, and the absolute value D3 of the difference between two pixels adjacent to the block boundary and the maximum value maxh1 of the difference between two adjacent pixels near the block boundary are compared. The presence or absence is determined (S24). If the absolute value D3 of the difference between two pixels adjacent to the block boundary is larger than the predetermined threshold thh3, it is determined that the edge is included in the image. Also, when the absolute value D3 of the difference between two pixels adjacent to the block boundary is smaller than maxh1, it is determined that the edge is included in the image. Therefore, when it is determined that the image is not an edge included in the image under the above two conditions (YES in S24), it is determined that there is block distortion.
[0037]
The horizontal filter 105 receives the signal that has been processed by the vertical filter 103 and, when the horizontal block distortion determination unit 104 determines that there is block distortion at a low frequency, 8 pixels centering on the block boundary (FIG. Filter processing is performed on h0 to h7) of FIG. Also, when the horizontal block distortion determination unit 104 determines that there is block distortion at a non-low frequency, it is considered that a high-frequency component is included, so that the entire block is filtered to prevent the image from being blurred. , Filter processing is performed only on two pixels (h3 and h4 in FIG. 2) centered on the block boundary.
[0038]
Here, it is conceivable that the vertical filter 103 is also subjected to filter processing in the case of a non-low frequency. However, if filter processing is performed on two pixels centering on a block boundary, which is a filter processing in the case of non-low frequency, on an interlaced image, an input decoded image is an image in which an object whose contour is sharp moves up and down. In such a case, there is a problem that the image quality is deteriorated, for example, the outline may appear double.
[0039]
This will be described with reference to FIG. FIG. 6A shows an example of an interlaced image before the filtering process. The pixels in the first field are V0 to V7, and the pixels in the second field are V0 'to V7'. When it is determined that the first field in FIG. 6A has block distortion at a non-low frequency, and that it is determined that there is no block distortion in the second field, V3 and V4 in FIG. Filter processing is performed. FIG. 6B shows an interlaced image after the filter processing. Assuming that V0 to V3 and V0 'to V2' in FIG. 6A represent white, and V4 to V7 and V3 'to V7' represent black, filtering is performed on V3 and V4 in FIG. As a result, V3 and V4 in FIG. 6B become gray. When FIG. 6B is displayed as a frame, V2 ′, V3, V3 ′, V4, and V4 ′ become white, gray, black, gray, and black, and the outline looks double. Therefore, when the input decoded image is interlaced, non-low-frequency filter processing is not performed in the vertical direction, thereby preventing the above-described generation of gray pixels and improving image quality.
[0040]
As described above, according to the present invention, block distortion can be reduced or eliminated without impairing the high-frequency components of the input decoded image, and when the input decoded image is interlaced, a non-low-frequency filter is provided in the vertical direction. By not performing the processing, it is possible to improve the image quality.
[0041]
In the present embodiment, the case where the block unit is 8 × 8 blocks has been described, but other block units may be used. In FIG. 1, the block distortion in the vertical direction is reduced or removed first, and then the block distortion in the horizontal direction is reduced or removed. However, the order may be reversed.
[0042]
The embodiments disclosed this time are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.
[0043]
【The invention's effect】
As described above, according to the present invention, the presence or absence of distortion is determined in advance in each direction for each of the first direction and the second direction that intersects with the first direction, and then the filtering process is performed in each direction. I have. Therefore, since the filtering process is performed only when the distortion is large, all the high frequency components of the decoded image are not removed, and the filtering process can be performed only in the direction in which the image quality deteriorates in the interlaced image. As a result, it is possible to provide a filter device that can reduce or remove block distortion without impairing the high-frequency component of the decoded image, and that can reduce or remove block distortion without deteriorating image quality even for an interlaced decoded image.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of a filter device according to the present invention.
FIG. 2 is a diagram for explaining a block boundary and pixels in the vicinity thereof;
FIG. 3 is a diagram for explaining pixels of a frame image.
FIG. 4 is a flowchart illustrating a process performed by a vertical block distortion determination unit.
FIG. 5 is a flowchart illustrating a process performed by a horizontal block distortion determination unit.
FIG. 6 is a diagram for describing filter processing of an interlaced image.
FIG. 7 is a block diagram illustrating an overall configuration of a video image processing device.
FIG. 8 is a block diagram illustrating a configuration of a filter device.
FIG. 9 is a diagram illustrating a filtering process of block distortion due to smoothing.
FIG. 10 is a diagram for describing filtering processing of block distortion due to smoothing.
[Explanation of symbols]
100 filter device, 101 line memory, 102 vertical block distortion determination unit, 103 vertical filter, 104 horizontal block distortion determination unit, 105 horizontal filter.

Claims (5)

Delay means for performing a predetermined delay on a decoded image signal processed in block units;
First determining means for determining block distortion in a first direction of a block of the decoded image signal;
First filtering means for filtering an image signal delayed by a predetermined delay by the delay means based on the determination result obtained by the first determining means in the first direction;
A second determination unit that determines block distortion in a second direction intersecting with the first direction of the image signal output from the first filter unit;
A second filtering unit configured to filter the image signal in the second direction based on the determination result obtained by the second determining unit. The filter device according to claim 1, wherein the first direction is a vertical direction and the second direction is a horizontal direction. 3. The filter device according to claim 2, further comprising: a unit that determines whether or not a pixel near the block boundary has a low frequency before the second determination unit in the second direction. Means for comparing the absolute value of the difference between adjacent pixels with a predetermined first threshold value for each of the pixels near the block boundary;
Means for comparing the absolute value of the difference between two pixels adjacent to the block boundary with a predetermined second threshold value;
3. The filter device according to claim 2, further comprising means for comparing an absolute value of a difference between two pixels adjacent to the block boundary and a maximum absolute value of a difference between two adjacent pixels near the block boundary. 3. The filter device according to claim 2, wherein the first filter unit or the second filter unit performs a filter process on a predetermined number of pixels around a block boundary. 4.

Images