JP3149729B2 - Block distortion removal device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a block distortion removing apparatus for removing block distortion caused by encoding when decoding an image signal which has been transformed and encoded in block units.

[0002]

2. Description of the Related Art In coding an image signal, the image signal is divided into a plurality of blocks, and each block is subjected to DCT.
A block coding method for coding image data by (discrete cosine transform) or the like is often used. In this method, a discontinuous level difference between adjacent blocks, that is, a so-called block distortion occurs, leading to deterioration in image quality.
Conventionally, an apparatus for removing block distortion by smoothing block boundaries has been proposed.

A conventional block distortion removing apparatus detects, for example, whether or not there is block distortion at a block boundary, performs a smoothing process when there is block distortion, and performs an edge process of an original image when there is no block distortion. Some store data. The smoothing processing includes a moving average, a median filter, and the like.

[0004]

However, in the above-mentioned conventional apparatus, the processing such as the moving average and the median filter performed for removing the block distortion can only linearly smooth the block distortion. It could not be removed according to the characteristics of the distortion.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a more accurate block distortion removing device by configuring a nonlinear noise filter that removes distortion in accordance with the characteristics of block distortion.

[0006]

SUMMARY OF THE INVENTION In order to solve the above-mentioned problems, a block distortion elimination apparatus according to the present invention, when decoding an image signal divided into a plurality of blocks, uses a level difference between pixels at the block boundaries. When the magnitude is within the range of the probability distribution in which block distortion occurs, the product of the probability according to the level difference and the correction amount represented by a straight line having a negative slope passing through the origin is calculated as the block distortion removal amount. And adding the respective levels so that the levels of the two pixels are close to each other.

[0007]

According to the above arrangement, the probability of block distortion is obtained from the level difference of pixels at the block boundary, and the larger the probability is, the larger the block distortion removal amount is, and the smaller the probability is, the smaller the removal amount is. I have. By making the characteristic of distortion removal non-linear in this way, distortion can be removed with higher accuracy.

[0008]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of a block distortion removing apparatus according to the present invention. FIG.
1 shows a configuration of a block distortion removing device according to an embodiment of the present invention.

First, the removal of block distortion in the horizontal direction will be described. The delay circuits 1 and 2 delay the input signal by one pixel. If the signal of the target pixel is the output signal of the delay circuit 1 and its XY coordinates are (x, y), the target pixel and its left and right The level difference between the pixels (x-1, y) and (x + 1, y) can be obtained by the subtracters 3 and 4.

When the level difference is input to the non-linear processing circuits 5 and 6, the block distortion removal amounts calculated according to the magnitudes are summed up in the adder 7. The adder 8 removes block distortion from the target pixel.

The block boundary detection circuit 9 and the ON / OFF
The off control circuit 10 adds the outputs from the non-linear processing circuits 5 and 6 to the signal of the target pixel when the target pixel is on the block boundary, and otherwise prevents the addition.

When the input level difference is within a predetermined range, the non-linear processing circuits 5 and 6 determine that there is block distortion and output distortion removal data. In the case of, edge data is determined and edge data is stored.

Specifically, the characteristics of the nonlinear processing circuits 5 and 6 are determined as follows. That is, the probability distribution in which the level difference at the block boundary is the block distortion is a normal distribution up to the level difference t as shown in FIG. If the level differences are all block distortions, the correction amount of the target pixel for removing the block distortions increases according to the level difference as shown in FIG. 3, and the level difference has a positive value. Is a negative value, and the level difference is a positive value if the level difference is a negative value.

If the block distortion removal amount is a product of the probability distribution of the block distortion and the correction amount, a block distortion removal amount having a magnitude proportional to the probability that the level difference between pixels is block distortion is given. It will be. Therefore, the characteristics shown in FIG. 4 are obtained. Block distortion is removed when the level difference is in the range of -t to t, and 0 is output at other levels to preserve the edge. t is a value for determining whether it is a block distortion or an edge.

The block distortion removal processing will be specifically described with reference to FIGS. FIG. 5A shows the pixel arrangement and the position of the block boundary, and FIG. 5B shows the signal levels of the reference pixels A, B, C, and D near the block boundary.
a indicates a block boundary, and b indicates a pixel to be processed. The level difference between the A and B pixels is d1, the level difference between the B and C pixels is d2, and the level difference between the C and D pixels is d3. Therefore, in this case, the level difference at the block boundary is d2.

In calculating the block distortion removal amount, not only the level difference d2 between the pixels B and C at the block boundary but also the level differences d1 and d between the reference pixels A and D on the opposite side to the block boundary.
3 is also used, but this is to remove block distortion more accurately as described later.
Assuming that d1 = 0 and d3 = 0 as shown in FIG.
Processing is performed only with the level difference d2 at the block boundary.

First, the processing will be described with the pixel of interest as a B pixel. If the absolute value of the level difference d2 at the block boundary with the C pixel is smaller than the reference value t, it is determined that block distortion has occurred, and as shown in FIG. The block distortion removal amount h is output from the non-linear processing circuit 5, and the value of the B pixel is reduced by the removal amount h. At this time, since the level difference d1 from the A pixel is 0, the output of the nonlinear processing circuit 6 is 0.

When the pixel of interest shifts to the C pixel, the block distortion removal amount h corresponding to the level difference d2 between the C pixel and the B pixel is output from the nonlinear processing circuit 6, and the value of the C pixel is reduced to the removal amount h.
Increase by minutes. Also, the level difference d between the C pixel and the D pixel
Since 3 is 0, the output of the nonlinear processing circuit 5 is 0. Therefore, the values of the B pixel and the C pixel approach each other so that the block distortion is reduced. If the absolute value of the level difference d2 is larger than the reference value t, it is determined that the edge is an edge, and the nonlinear processing circuits 5 and 6 output 0, so that the edge is stored.

Next, the process of adding the reference pixel will be described. When the level difference d1 between the A pixel and the B pixel and the level difference d3 between the C pixel and the D pixel are added to the calculation, the block distortion can be removed with higher accuracy. For example, as shown in FIG. 7A, when the signal level continuously changes at the block boundary, it is considered to remove the block distortion based on only the level difference d2 at the block boundary as described above.

In this case, since the signal level of the B pixel decreases and the signal level of the point C increases by the block distortion removal amount h2 with respect to the level difference d2 at the block boundary, continuity may be lost due to the processing.

Then, the level difference d1 between the A pixel and the B pixel, C
Using the outputs h1 and h3 from the non-linear processing circuits 5 and 6 for the level difference d3 between the pixel and the D pixel, the levels of the B and C pixels are respectively corrected. The outputs h1 and h3 can maintain continuity because the level change of the B pixel and the C pixel due to the block distortion removal amount h2 is suppressed.

Regarding the removal of the vertical block distortion,
Similarly to the processing in the horizontal direction, when the target pixel is on the block boundary by the block boundary detection circuit 9, the target pixel (x, y) and the target pixel and its upper and lower pixels (x, y-
The difference between (1) and (x, y + 1) is obtained by the subtracters 3 and 4, and the above-described block distortion removal processing can be performed by the nonlinear processing circuits 5 and 6 from the difference data. In this case, the delay circuits 1 and 2 are one-line delay circuits.

In the above-described embodiment, only the processing for removing the block distortion at the block boundary is performed. If the block boundary has an edge, the original edge of the image is preserved. Therefore, if the characteristics shown in FIG. 8 are added to the above-described non-linear processing circuits 5 and 6, the edge can be enhanced and the sharpness of the image can be increased.

That is, the level difference between the target pixel and its left or right or upper and lower pixels calculated by the delay circuits 1 and 2 and the subtracters 3 and 4 is input to the nonlinear processing circuits 5 and 6. When the pixel of interest is on the block boundary, t can be used as a reference value for determining whether the distortion is a block or an edge. If the level difference of the signal at the block boundary is a value between -t and t, it is determined that the signal is a block distortion and the above processing is performed. If the level difference is smaller than -t or greater than or equal to t, it is determined to be an edge, and data for increasing the level difference is output. If the pixel of interest is not on the block boundary, the on / off control circuit 10 controls the non-linear processing circuits 5 and 6 to output the negative signal.
By setting the output values from t to t to 0, only contour emphasis can be performed. In this case, since the contour emphasis amount is the same as that at the block boundary, data continuity can be maintained.

[0025]

As described above, the block distortion removing apparatus of the present invention has a nonlinear characteristic so that the block distortion removal amount changes in accordance with the probability that the level difference detected between pixels at the block boundary is block distortion. , It is possible to perform highly accurate processing in accordance with the characteristics of block distortion. The continuity of the data of the original image can be maintained by correcting the removal of the block distortion between the pixels on the block boundary based on the level difference from the reference pixel adjacent to the pixel on the block boundary. Furthermore, by selectively performing not only block distortion removal but also edge enhancement between pixels at block boundaries, a sharp and distortion-free image can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a block distortion removing apparatus according to the present invention.

FIG. 2 is a probability distribution diagram of block distortion.

FIG. 3 is a characteristic diagram illustrating an example of a correction amount of block distortion.

FIG. 4 is an input / output characteristic diagram of the nonlinear processing circuit in the embodiment.

FIG. 5 is a diagram illustrating a pixel array and signal levels of pixels on a block boundary and reference pixels in the embodiment.

FIG. 6 is a view showing a distortion process between pixels at a block boundary in the embodiment.

FIG. 7 is a diagram showing a pixel on a block boundary and distortion processing including a reference pixel in the embodiment.

FIG. 8 is a characteristic diagram of a non-linear processing circuit that performs block distortion removal and edge enhancement in the embodiment.

[Explanation of symbols]

 1, 2 delay circuit 3, 4 subtractor 5, 6 nonlinear processing circuit 7, 8 adder 9 block boundary detection circuit 10 on / off control circuit

Claims (3)

(57) [Claims] When decoding an image signal divided into a plurality of blocks, a level difference between pixels at the block boundary is obtained, and when the magnitude is within a range of a probability distribution in which block distortion occurs, A product of the probability according to the level difference and a correction amount represented by a straight line having a negative slope passing through the origin is added as a block distortion removal amount so that the levels of the two pixels are close to each other. Block distortion removal device. 2. The block distortion removal amount is corrected by a value calculated from a level difference between a pixel at the block boundary and a reference pixel adjacent thereto in the same manner as the block distortion removal amount between pixels at the block boundary. 2. The block distortion removing apparatus according to claim 1, wherein the subtraction is performed from the pixels at the block boundaries. 3. A level difference is obtained between a pixel at a block boundary and a reference pixel adjacent thereto, and when the level difference is out of a range of a probability distribution in which block distortion occurs, the pixel between the block boundary and the reference pixel is determined. 3. The block distortion removing apparatus according to claim 1, wherein the pixel at the block boundary is corrected so as to emphasize the edge by increasing the level difference of the block distortion.