JPH0497678A - Picture decoding system - Google Patents

Abstract

PURPOSE:To prevent deterioration in picture quality by reflected distortion by applying inverse conversion to a block of the same size as that of a coder side, applying low pass filtering processing to a picture element value being an inverse conversion output and reducing number of picture elements by subsampling. CONSTITUTION:The system consists of a DCT(Discrete Cosine Transformation) transformation section 1, a quantization section 2, a code assignment section 3, a reception code decoding section 4, an inverse DCT transformation section 6, a low pass filtering processing section (LPF) 7, and a subsampling section(SS) 8. Then remaining transformation coefficients corresponding to high frequency components are made zero and inverse conversion is implemented by a same block size in the case of coding and low pass filtering processing is implemented to a picture element being an inverse conversion output and number of picture elements is reduced by subsampling. Thus, since the low pass filtering processing is implemented after inverse conversion, occurrence reflected distortion is prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image decoding method for decoding a coded image signal to reproduce an image having a resolution and number of pixels different from those of the original image. The method of the present invention is applied, for example, to a case where encoded data of an HDTV image or a part of the data is received and a current standard definition television image is reproduced from the received data.

(Prior Art) As mentioned above, DCT (Discrete Cos.
There is a method using ine transform (discrete cosine transform).

An example of its configuration is shown in FIG. 4, in which 1 is a DCT conversion unit;
2 is a DCT coefficient quantization unit, 3 is a code assignment unit for quantized DCT coefficients, 4 is a reception code decoding unit, and 5 is an inverse DCT
This is the conversion section.

To explain the operation on the encoder side, in the DCT transform unit 1,
8 pixels in the horizontal direction x 8 lines in the vertical direction (hereinafter referred to as "8 x 8 for simplicity"
A two-dimensional DCT is performed using 64 pixels (1) of ``'') as one block to convert the pixel domain to the DCT coefficient domain (2). The signal is quantized in step 2, and then represented by an appropriate code in code assignment section 3 and sent out.

On the decoder side, a reception code decoder 4 first obtains DCT transform coefficients from the received encoded data.

In the inverse DCT transform unit 5, as shown in FIG. 6, 4×4 (16 pieces/block) corresponding to low frequency components
A 4×4 block is constructed using only the DCT transform coefficients, and 4×4 inverse DCT transform (1) is performed.

This results in an image consisting of 16 pixels per block (2
) is reproduced, but compared to the original image, the resolution is reduced in both the horizontal and vertical directions, and the number of pixels is also reduced to 1/2 in both the horizontal and vertical directions.

(Problem M to be Solved by the Invention) In the prior art described above, the low frequency component representing the 4×4 transform coefficient in FIG. 6 covers the entire frequency band represented by the original 8×8 transform coefficient. It is assumed that the band is limited to 172 or less in both the horizontal and vertical directions.

However, in reality, the conventional method of truncating the transform coefficient corresponding to the DCTO high frequency component may not achieve sufficient band limitation. That is, in DCT, transformation is performed block by block, but due to the nature of DCT, sufficient band limitation cannot be performed for images that do not have symmetry within the block.

For this reason, if the number of pixels is reduced, aliasing distortion will occur and the reproduced image will deteriorate. This is a particularly noticeable deterioration when edge portions of the image exist near the boundaries of transformed blocks.

(Objective of the Invention) An object of the present invention is to solve the problems of the prior art as described above, and to provide an image decoding method that does not cause aliasing distortion in resolution conversion and pixel number conversion using transform coding. It's about doing.

(Means for Solving the Problems) In order to solve the above problems and achieve the objectives, the present invention decodes image signals by inverse transformation from image signal encoded data that has been orthogonally transformed block by block of a certain size. In doing so, the transform coefficient values obtained from the encoded data are used for a predetermined number of orthogonal transform coefficients corresponding to low frequency components, and the values of the remaining transform coefficients corresponding to high frequency components are set to zero. Inverse transformation is performed using blocks of the same size as the encoding side, and after low-pass filtering is performed on the pixel values that are the inverse transformation output, the number of pixels is reduced by subsampling, resulting in higher resolution and higher resolution than the original image. It is characterized by reproducing an image with a reduced number of pixels.

(Function) The present invention sets the values of the remaining transform coefficients corresponding to high frequency components to zero, performs inverse transform with the same block size as in the case of encoding, and performs inverse transform on the pixel value that is the inverse transform output. Furthermore, after performing low-pass filtering processing, the number of pixels is reduced by subsampling. By further performing low-pass filtering processing after the inverse transformation in this manner, generation of aliasing distortion can be prevented and a high-quality image decoding system can be obtained.

(Example) As an example of the present invention, a case will be described in which a standard resolution TV image in which the resolution and number of pixels in both the horizontal and vertical directions are reduced to 1/2 from a DCT-encoded HDTV image is reproduced.

FIG. 1 is a block diagram of one embodiment of the present invention, and 1 is a DCT
2 is a quantization unit for DCT coefficients, 3 is a code assignment unit for quantized DCT coefficients, 4 is a reception code decoding unit, 6 is an inverse DCT transformation unit, and 7 is a low-pass filtering processing unit (
LP F), 8 is a sub-sampling section (SS).

Next, the operation will be explained based on this configuration, but the operation of each section from the DCT converter 1 to the received code decoder 4 is the same as in the conventional technique. That is, on the encoder side, DCT
In the conversion unit 1, the HDTV input image is subjected to two-dimensional DCT in which one block is 8×8 64 pixels (1) as shown in FIG. 5, and the DCT coefficient region (2
).

The obtained transform coefficients (64 pieces/block) are quantized in the quantization section 2, and then represented by an appropriate code in the code assignment section 3 and sent out.

On the decoder side, a reception code decoder 4 first obtains DCT transform coefficients from the received encoded data.

Now, in the next inverse DCT transform section 6, as shown in FIG.
For the DCT transform coefficients of the block), the DCT coefficients obtained from the received encoded data are used, and the remaining DCT coefficients corresponding to high frequency components are set to zero to create an 8x8 block (1), ×8 inverse DCT transform (2) is performed. As a result, an image consisting of 64 pixels per block is reproduced. This has the same number of pixels compared to the original image.

On the other hand, the resolution is reduced in both the horizontal and vertical directions, but as explained in the section "Problems to be Solved by the Invention" above, the degree of band limitation varies depending on the target image. Next, the LPF 7 performs low-pass filtering processing in both the horizontal and vertical directions. An example of this filtering process is shown in FIG. In this example, horizontal low-pass filtering [(x-, +2x, +x,)/4] is performed by weighting a total of three pixels: the target pixel X and two adjacent pixels before and after it, x, , X-1. I understand.

Similar filtering is performed in the vertical direction as well.

Finally, the 88 section 8 performs subsampling. That is, the number of pixels is reduced to 1/2 in both the horizontal and vertical directions. In this way, an image with a resolution and a reduced number of pixels of 172 is obtained as an 88-part 8 image.

In the above embodiment, an 8×8 DCT transform has been described, but the embodiment can be configured in the same way even in the case of an orthogonal transform different from this and a block size different from this. As for the low-pass filtering process, a configuration different from that of the embodiment, or a configuration in which the low-pass filtering process is performed only in the horizontal direction or only in the vertical direction is also possible. Furthermore, by changing the range of DCT coefficients to be set to zero and the characteristics of the low-pass filter, it is also possible to set the rate of reduction in the number of pixels to a value other than 172.

(Effects of the Invention) As explained above, according to the method of the present invention, inverse DCT transform is performed on blocks of the same size as those on the encoding side, and the pixel values that are the inverse transform output are subjected to low-pass filtering. Since the number of pixels is reduced by subsampling after processing, image quality deterioration due to aliasing distortion that occurs in conventional techniques can be prevented, and a high-quality image decoding system can be provided.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an embodiment of the present invention, Fig. 2 is an explanatory diagram of inverse DCT transform in the present invention, Fig. 3 is an explanatory diagram of low-pass filtering, and Fig. 4 is an image decoding method according to the prior art. Fig. 5 is an explanatory diagram of two-dimensional DCT transformation, Fig. 6 is a block diagram of
The figure is an explanatory diagram of inverse DCT transformation in the prior art. 1...DCT transformation section, 2...quantization section, 3
... code allocation section, 4 ... reception code decoding section, 5
, 6... Inverse DCT transform section, 7... Low pass filtering processing section (LPF), 8... Subsampling section (SS). Patent applicant Nippon Telegraph and Telephone Corporation Figure 1 Figure 2

Claims (1)

[Claims] When decoding an image signal by inverse transformation from image signal encoded data that has been orthogonally transformed in block units of a certain size, a predetermined number of orthogonal transform coefficients corresponding to low frequency components are extracted from the encoded data. Using the obtained transform coefficient values, and setting the values of the remaining transform coefficients corresponding to high frequency components to zero, inverse transform is performed in blocks of the same size as the encoding side, and the pixel value, which is the inverse transform output, is An image decoding method characterized by performing low-pass filtering processing on the image and then reducing the number of pixels by subsampling to reproduce an image with a lower resolution and a lower number of pixels than the original image.