CN102256126A - Method for coding mixed image - Google Patents

Abstract

The invention discloses a method for coding a mixed image with higher coding efficiency. In the method, a residual scalar quantization (RSQ) mode is used in the interframe coding of a mixed image, according to a principle that rate distortion cost is minimum, each transformation unit (TU) adaptively selects a discrete cosine transform (DCT) mode or the RSQ mode, and a bit is used for expressing the selection; and residual errors obtained after interframe prediction are directly quantized in the RSQ mode, and a step of transforming codes is omitted.

Description

Coding Method for Mixed Images

technical field

The invention belongs to the technical field of image processing, and in particular relates to a method for encoding mixed images.

Background technique

In existing technologies, coding of mixed images can be roughly classified into the following three categories.

1. Method based on image coding:

This method still uses the traditional encoding framework to encode the mixed image, but uses different quantization steps for different regions. For blocks containing text and graphics, the reconstruction effect after encoding will be very blurred, so when quantizing these blocks Reduce the quantization step size, allocate more bits to it, and make its reconstruction quality as good as possible. But for a fixed number of bits, if more bits are allocated for text and graphic blocks, then the number of bits allocated for natural image blocks will be correspondingly reduced, so generally speaking, this method still cannot achieve good results.

2. Method based on layered coding:

This method uses a hybrid raster model for compression. It divides an image or an image block into a foreground layer, a background layer and a plane layer of a binary operator. The plane layer of this binary operator indicates that each pixel of the image or image block belongs to the foreground layer or the background layer, and this binary operator plane can be compressed with JBIG standard or JBIG2 standard. The foreground and background layers blurred by the data filling algorithm can be compressed using the traditional image coding framework. The paper "Compressing compound documents" (The Document and ImageCompression Handbook, M.Barni, Ed.New York: Marcel-Dekker, 2005.) proposed this method, and it was proved by experiments that this method is compared with the traditional codec framework , a large gain can be obtained. But this method has some disadvantages: First, the gain depends largely on the segmentation algorithm, that is, how to divide the image into foreground and background. Second, the segmented holes will damage the encoding effect if they are not processed. Third, separately encoding the text color of the background layer and the text shape of the binary operator also reduces the efficiency of encoding. Therefore, this method still cannot meet the needs of the application well.

3. Block-based encoding method:

This approach can be divided into two categories:

One is to use the properties of the airspace, such as histogram statistics, gradients, the number of colors, etc., to classify the current coding block into one of several categories, such as text, charts, natural images, etc. Different categories are then compressed with different encoding frameworks to suit their statistical properties, as shown in Figure 1.

The other is to consider that the histogram of the text and graphics in the mixed image is relatively sparse, and decompose it into the form of the basic color and index table in the time domain, that is, the Base color and Index map (BCIM) mode.

Although the block-based coding method has achieved good gains compared to the two aforementioned techniques, all current coding methods have one thing in common, that is, only the spatial correlation is considered, that is, only the frame The internal part is improved, but the time correlation is ignored, the characteristics of inter-frame coding are not taken into account, and the coding scheme for inter-frame coding mixed images is not designed.

Contents of the invention

The problem solved by the technology of the present invention is to overcome the shortcomings of the prior art and provide a method for encoding mixed images with higher encoding efficiency.

The technical solution of the present invention is: this mixed image coding method uses RSQ (Residual scalar quantization, residual scalar quantization) mode for interframe coding of mixed images: according to the principle of minimum rate-distortion cost, each TU (Transform unit, transformation unit) adaptively selects DCT (Discrete cosine transform, discrete cosine transform) mode or RSQ mode, and uses one bit to represent the selection; RSQ mode quantizes the residual obtained after inter-frame prediction.

Since the present invention solves the problem that the mixed image is not suitable for transform coding, and since the residual error obtained after the inter-frame prediction will be smaller, the inter-frame coding efficiency for the mixed image will be higher.

Description of drawings

Figure 1 shows an existing RSQ mode, a block-based coding method that exploits spatial properties;

Fig. 2 shows a flow chart of a method for encoding mixed images according to the present invention.

Detailed ways

This mixed image coding method uses RSQ (Residual scalar quantization) mode for interframe coding of mixed images: according to the principle of the least rate-distortion cost, each TU (Transform unit, transformation unit) adaptive Select DCT (Discretecosine transform, discrete cosine transform) mode or RSQ mode freely, and use one bit to represent the choice; RSQ mode is to directly quantize the residual obtained after inter-frame prediction, and skip the step of transform coding.

When transmitting RSQ encoding, scan in row order and assign a context to it.

As shown in Figure 2, the encoding method of this mixed image includes the following steps:

(1) For the luminance component in the PU, obtain its prediction block, and then subtract it from the original value to obtain the residual, divide the residual block into different sizes, and obtain TU (Transform Unit, transformation unit);

(2) quantify the residual in the current TU;

(3) Send the quantized TU to the decoding end;

(4) Inverse quantization is performed on each TU at the decoding end;

(5) The TU is combined into a PU, which is added to the predicted value of the current PU after motion compensation to obtain the reconstruction value of the PU, and then the current CU (Coding Unit, coding unit) is reconstructed.

Preferably, the quantization in step (2) adopts the formula: C(S)=sign(S)×max(0, floor(|S|/Q+1-Z), wherein S is the input signal, and floor is downward Rounding function, Q is the quantization step size, and Z is the preset adjustment parameter.

Preferably, the transmission in step (3) is different from the HM standard in that the residuals are not scanned in zig-zag order, but each residual coefficient is scanned sequentially in raster scanning order, and then transmitted sequentially.

Preferably, the dequantization in step (4) adopts the formula: R(x)=sign(x)×round(q×|x|+p), wherein round is a rounding function, and q and p are The control parameter transformed by the parameter.

The above are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Any simple modifications, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention are still within the scope of this invention. The protection scope of the technical solution of the invention.

Claims (6)

1. The coding method of the mixed image is characterized in that: the residual scalar quantization RSQ mode is used for the interframe coding of the mixed image: according to the principle of the minimum rate-distortion cost, each transformation unit TU adaptively selects the discrete cosine transform DCT mode Or RSQ mode, and use one bit to represent the selection; RSQ mode is to quantize the residual obtained after inter-frame prediction. 2. The method according to claim 1, characterized in that: when transmitting RSQ codes, row sequential scanning is used and an independent context is assigned to it. 3. the coding method of mixed image, it is characterized in that: comprise the following steps: (1) For the luminance component in the prediction unit PU, the prediction block is obtained, and then subtracted from the original value to obtain the residual, and the residual block is divided into different sizes to obtain the transformation unit TU; (2) quantify the residual in the current TU; (3) Send the quantized TU to the decoding end; (4) Inverse quantization is performed on each TU at the decoding end; (5) The TUs are combined into PUs, which are added to the predicted value of the current PU after motion compensation to obtain the reconstruction value of the PU, and then the coding unit CU of the current coding unit is reconstructed. 4. The method according to claim 3, characterized in that: the quantization in the step (2) adopts formula: C(S)=sign(S)×max(0, floor(|S|/Q+1-Z ), where S is the input signal, floor is the rounding down function, Q is the quantization step size, and Z is the preset adjustment parameter. 5. The method according to claim 3, characterized in that: the transmission in step (3) is to scan each residual coefficient sequentially in a raster scanning order, and then transmit them sequentially. 6. The method according to claim 4, characterized in that: the inverse quantization in the step (4) adopts formula: R(x)=sign(x)×round(q×|x|+p), wherein round is The rounding and rounding function, q, p are the control parameters that change with the quantization parameters.

Images