JPH0714211B2 - Method and apparatus for removing block distortion in moving picture coding - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention divides a moving image signal into a pixel block such as a video conference and a videophone, and performs calculation such as conversion for each pixel block.
The present invention relates to a method and an apparatus for digital transmission with quantization for data compression to reduce a bit rate.

(Prior Art) When transmitting an image signal, many attempts have been made to remove noises that are superimposed without impairing the sharpness of the original image. Among them, a large number of image filters have been proposed which remove edges such as random noise while preserving edges in the image.

In such a filter, for each pixel, a signal level of a peripheral pixel is referred to, an operation is performed based on the signal level, and a process of overwriting the pixel with a new signal level is repeatedly performed on the entire surface of the image.

Examples of the above-mentioned calculation are: (a) referring to the signal level of the pixel concerned and its surrounding pixels, multiplying the value by a coefficient for smoothing, and taking a weighted average; A method in which the direction of an edge is determined by the arrangement of signal levels, and a combination of weighted average coefficients is adaptively selected so as to save the edge direction. (C) A histogram is created for the signal levels of peripheral reference pixels, and the intermediate value is set. For example, the new signal level of the pixel is set.

(Problems to be Solved by the Invention) In a coding method in which a moving image signal is divided into pixel blocks, and calculation is performed for each pixel block and quantization is performed for data compression, when the transmission bit rate is low, Since coarse quantization is performed, an unnatural discontinuous state (block distortion) occurs at the pixel block boundary portion, which adversely affects the image quality.

In order to remove such distortion and to perform image processing that does not make the original edges of the image as blunt as possible, it is necessary to perform edge-preserving smoothing processing. However, since the conventional edge-preserving filter is a method that uniformly operates on the entire surface of the image, there is a problem that the block distortion is treated as an edge and effective smoothing is not performed.

An object of the present invention is to focus on the boundary portion of a pixel block and effectively remove the block distortion while preserving the original edge of the image.

(Means for Solving the Problem) A feature of the present invention is that it can be independently connected to a low bit rate moving picture decoder, and the original edge of the image (the ringing of the image is reduced by the output of the decoder and the size of the pixel block). ) Is an edge-preserving type block distortion eliminator that can effectively remove block distortion without slowing down.

FIG. 1 shows a block diagram of this device. An input unit 11 for selecting the pixel signal level around the block boundary and an input signal delay unit 12 for arranging the input signals as a reference pixel column.
And whether or not to perform smoothing processing on the block boundary based on the signal level of the reference pixel, that is, the smoothing determination unit 14 that performs block distortion detection, and whether or not smoothing is required by the smoothing determination unit (ON / OFF ) When the control signal is received (ON), an operation for smoothing the pixel is performed, that is, a smoothing operation unit 13 that performs block distortion removal and is configured to output an operation processing result. Type block distortion removal device.

(Example) Input unit for selecting pixel signal level around block boundary
At 11, the signal levels of four pixels which are symmetrical with respect to the block boundary and are arranged in the direction perpendicular to the block boundary are input. An example of selective input is shown in FIG. As shown in FIG. 2, a 4-pixel signal in the vertical direction is input to the block boundary in the horizontal direction, and a 4-pixel signal in the horizontal direction is input to the block boundary in the vertical direction. The output signal of the input unit 11 is arranged in the input signal delay unit 12 as a reference pixel column.

The smoothing determination unit 14 determines whether the block boundary portion needs to be smoothed from the signal level of the reference pixel column, that is, the block distortion detection process is determined and detected by the following determination algorithm.

FIG. 3 shows an example of the relationship between the pixel arrangement and the signal level calculated and determined in the algorithm. Judgment parameters d ₀ , d ₁ , and d ₂ are calculated from the signal level of the reference pixel.

_{d 0 = | S 0 -S '} 0 | d 1 = | S 1 -S 0 | d 2 = | S' 0 -S '1 | these values with the threshold Th _A, the magnitude comparison of Th _B, d ₀ ≦ Th _A : Smoothing disabled (OFF) d ₀ ＞ Th _A and (d ₁ ≦ Th _B and d ₂ ≦ Th _B ): Smoothing required (ON) d ₀ ＞ Th _A and (d ₁ ＞ Th _B or d ₂ ＞ Th _B ): Judge whether smoothing is required (OFF) and whether smoothing is required (ON / OFF). That is, when the signal level difference d ₀ between the two pixels in contact with the boundary is less than or equal to the first threshold value Th _A (d ₀ ≦ Th _A ), the signal level difference at the boundary portion is not so large. It is determined that smoothing is not required. The signal level difference d ₀ between two pixels touching the boundary is the first
Is larger than the threshold value Th _{A of the} two pixels (d ₀ > Th _A ), the signal level differences d ₁ and d ₂ between the pixels adjacent to these two pixels and each of these two pixels and the second threshold value Th A _By comparing with _B , it is determined whether the original edge portion of the image or the block distortion due to quantization. That is, the signal level difference d ₁
And d ₂ is less than or equal to the second threshold value Th _B (d ₀ > Th _A and d ₁ ≦ Th _B and d ₂ ≦ Th _B ), the signal level changes sharply on both sides of the boundary. Since the change in the signal level is gradual outside that, it is considered to be block distortion due to quantization, and it is determined that smoothing is required. Also, the signal level difference d ₁
And at least one of d ₂ is larger than the second threshold Th _B (d
₀ > Th _A and d ₁ > Th _B or d ₂ > Th _B ), the change in the signal level around the boundary is continuous and abrupt. It is determined that smoothing is not required.

In the smoothing calculation unit 13, only when a signal indicating that smoothing is required (ON) is received from the smoothing determination unit 14, for example, the following smoothing process, that is, block distortion removing process is simultaneously performed on two pixels that are in direct contact with the block boundary. . This is a weighted averaging process. Taking FIG. 3 as an example, the smoothed corresponding pixel X ₀ ,
The new signal levels S _0-new and S ′ _0-new of X ′ ₀ are S _0-new = (S ₁ + 2 × S ₀ + S ′ ₀ ) / 4 S ′ _0-new = (S ₀ + 2 × S It is given as ′ ₀ + S ′ ₁ ) / 4.

Test video C actually used for CCITT standardization work
About 32 frames of LAIRE encoded with 64KBIT / S,
As a result of applying the present invention using the above algorithm, it was confirmed that the original image S / N of the decoded image to which the present invention is not applied is improved. Furthermore, it was found by visual inspection by the naked eye that the visual impression was improved more than the improvement of the S / N ratio. The smoothing required (ON) rate in the smoothing determination unit at this time is 17.30%.

As a result of applying the present invention directly to the original image before encoding as a guideline for showing the effect of the present invention to preserve edges, the smoothing required (ON) rate is 2.42% and the S / N ratio is 67.18 dB, which is unnecessary. It was found that the smoothing process was hardly performed.

(Effects of the Invention) The present invention preserves the original edge of an image for an unnatural discontinuous state (block distortion) at a pixel block boundary that occurs when the encoding bit rate is low in encoding a moving image. However, it can be effectively removed. This means that it is possible to eliminate visual stigma due to block distortion.

Further, since the present invention does not use variable parameters used in encoding, such as the motion compensation amount and the quantization step size, it can be independently connected to the decoder, and the function addition by the device of the present invention is easy. . Moreover, since a relatively simple smoothing determination algorithm and smoothing calculation algorithm are used, the device configuration is simple and the edge-preserving block distortion removal processing can be effectively and easily performed.

[Brief description of drawings]

FIG. 1 shows a block diagram of the device of the present invention. FIG. 2 shows the positional relationship between pixel blocks, block boundaries, and reference pixels. FIG. 3 shows the relationship between the signal level of the reference pixel and the ON / OFF determination parameter. 11 ... Input unit, 12 ... Input signal delay unit, 13 ... Smoothing operation unit,
14 ... Smoothing determination section.

Claims (2)

[Claims] 1. A block distortion removing method in a transmission method in which an image is divided into a plurality of pixel blocks and quantization is performed for each block, and the block distortion removing method is arranged in one column in a vertical direction at a boundary of the pixel blocks. In each of the two pixel blocks having a signal level difference between two boundary adjacent pixels adjacent to each other across the boundary of the pixel is larger than a first threshold value, the pixel adjacent to the boundary adjacent pixel and the column above A first step of determining block distortion due to quantization only when the signal level difference between the pixel adjacent to the boundary adjacent pixel in the same block among the pixels A second step of performing a smoothing process for removing the block distortion on the pixel in contact with the block boundary when it is determined to be the block distortion. Block distortion removal method of reduction. 2. A block distortion eliminating device in a transmission device for dividing an image into a plurality of pixel blocks and quantizing each block, wherein the block distortion eliminating devices are arranged on one column in a vertical direction at a boundary of the pixel blocks. In each of the two pixel blocks having a signal level difference between two boundary adjacent pixels adjacent to each other across the boundary of the pixel is larger than a first threshold value, the pixel adjacent to the boundary adjacent pixel and the column above Determination means for determining the block distortion due to quantization only when the signal level difference between the pixel adjacent to the boundary adjacent pixel in the same block in the same pixel is smaller than the second threshold value; Block distortion for moving picture coding, characterized in that it has smoothing processing means for removing the block distortion for the pixels in contact with the block boundary when the means determines that the block distortion is present. Removed by the device.