JP5586061B2 - Color component prediction type image encoding apparatus and decoding apparatus - Google Patents

Description

  The present invention relates to an inter-color component predictive image encoding apparatus and decoding apparatus, and more particularly to an efficient compression encoding for high-definition video and an inter-color component predictive image encoding apparatus for decoding the same. Relates to the device.

  Compression encoding systems for storage applications such as hard disk recorders and VTRs are likely to support a 4: 4: 4 color space represented by RGB and a 4: 2: 2 color space. As a conventional technique for improving the encoding efficiency in such a compression encoding system, there are the following.

  In Patent Document 1 below, for 4: 4: 4 video input, prediction reference between color components is performed prior to encoding, and only the residual signal is handled in the subsequent encoding process. To improve performance.

  Further, in Patent Document 2, an input image signal itself or a residual signal obtained by intra-frame prediction is determined as an application target of inter-color component prediction by analyzing an input image. Propose that.

  Non-Patent Document 1 shows that high coding performance can be achieved when a residual signal obtained by intra-frame prediction is used as an application target of inter-color component prediction.

JP 2006-310941 A Japanese Patent Application No. 2009-85661

2009 Video Coding Symposium “A Study on Inter-channel Prediction for Intra Prediction Residue of H.264”

  For the 4: 4: 4 video signal, the prediction between color components can be effectively utilized, and high compression performance can be achieved. However, in the techniques described in the above-mentioned patent documents and non-patent documents, 4: 4: 4 video is mainly applied, and luminance such as 4: 2: 2 video and 4: 2: 0 video is used. It was not intended to be applied to a color space where the signals are independent.

  In color spaces where luminance signals are independent, such as 4: 2: 2 video and 4: 2: 0 video, the encoding method applied to the luminance signal and the chrominance signal is usually different. In the case of applying inter-color component prediction to the signal, the improvement in compression performance was limited. This is because a different encoding method is applied to each color component, so that the correlation between the color components of the residual signal is reduced and the compression performance is affected.

  The present invention has been made in view of the above-described problems of the prior art, and its purpose is to allow a difference in encoding method between color components and to reduce performance of prediction between color components for a residual signal. An object of the present invention is to provide an inter-color component prediction type image encoding device and decoding device that can be reduced.

In order to achieve the above-described object, the present invention provides an inter-color component prediction type image encoding apparatus capable of adaptively setting a reference signal for inter-color component prediction, and a reference signal from an input image signal composed of a plurality of color components. A reference signal selection unit (3) that selects a non-reference signal, a first intra prediction unit (12) that obtains a prediction mode by intra-predicting the non-reference signal, and an intra decoded image or a locally decoded image of the reference signal A first resolution conversion unit (14a) that converts the resolution of the reference decoded image (hereinafter referred to as “reference decoded image”) to be equal to the resolution of the non-reference signal; From the second intra prediction unit (14c) that performs intra prediction in the prediction mode obtained by the intra prediction unit, and the resolution-converted reference decoded image, Subtraction to the obtained reference component predicted residual image intra prediction image predicted by the second intra prediction unit (15), a subtractor for subtracting from said residual signal of the intra prediction of the non-reference signal (e) (16 And a means for performing an encoding process on the inter-color component prediction error signal (f) obtained by the subtraction.

  Further, the present invention provides a second resolution conversion unit (31a) for converting the resolution of the reference decoded image so as to be equal to the resolution of the non-reference signal, before the first intra prediction unit (12). A second feature is that a subtracter (32) for subtracting the reference component predicted image obtained by the second resolution conversion unit from the non-reference signal is provided.

Further, the present invention provides a decoding device, a determination unit (42) for determining a reference signal and a non-reference signal from the encoded signal, and an intra decoded image or a locally decoded image (hereinafter referred to as “reference decoded image) of the reference signal. A third resolution conversion unit (55a) that converts the resolution of the non-reference signal to be equal to the resolution of the non-reference signal, and the variable-length decoding of the non-reference signal using the resolution-converted reference decoded image. A third intra prediction unit (55c) that performs intra prediction in the prediction mode (d) extracted from the subtracted intra prediction image predicted by the third intra prediction unit from the resolution-converted reference decoded image reference component prediction residual image obtained Te (56) and said non-reference signal of the variable-length decoding result (54) is a residual signal and an adder to obtain a decoded residual image by adding (58) the ( 7), wherein the intra-prediction image (60) obtained by the intra prediction in a non-reference signal prediction mode extracted from the variable length decoding of (d), the decoded residual image (58) an adder for adding the (59) has the third feature.

  According to the encoding device and / or the decoding device of the present invention, it is possible to perform inter-color component prediction on a residual signal even in a color space where luminance signals are independent, such as 4: 2: 2 signals and 4: 2: 0 signals. Performance degradation can be reduced. For this reason, prediction between color components can be made to function with high accuracy.

  In particular, an intra-decoded image or a locally-decoded image (“reference decoded image”) of the reference signal is converted to the same resolution as the non-reference signal, and then intra prediction (intraframe prediction) applied to the non-reference signal is performed. Since the reference component prediction residual image is obtained, it is possible to reduce the performance degradation of inter-color component prediction for the residual signal. For this reason, the prediction between color components can be functioned with high accuracy, and the amount of codes can be reduced.

  According to the experiment results of the present inventors, it was confirmed that the code amount was reduced by about 10%.

1 is a block diagram showing a schematic configuration of a first embodiment of the present invention. It is a block diagram which shows the structure of the example of the 1st inter-component prediction part of FIG. It is a block diagram which shows the structure of the outline of 2nd Embodiment of this invention. It is a block diagram which shows the structure of the example of the 2nd inter-component prediction part of FIG. It is a block diagram which shows the schematic structure of the decoding apparatus corresponding to the said 1st Embodiment. FIG. 6 is a block diagram illustrating a configuration of a specific example of a first inter-component prediction unit in FIG. 5. It is a block diagram which shows the schematic structure of the decoding apparatus corresponding to the said 2nd Embodiment. It is a block diagram which shows the structure of the example of the 2nd inter-component prediction part of FIG.

  Hereinafter, the present invention will be described in detail with reference to the drawings. In video encoding, as a method or method for performing prediction between color components, (1) a method or method for the input image itself, and (2) a residual signal when applying intra-frame prediction for each color component There are methods or methods that target Among these, for the former (1), since the prediction of the input image is performed by referring to the locally decoded image, the difference in the coding method for each color component does not affect the efficiency reduction.

  The present invention is directed to the latter (2). This (2) can be divided into the following two cases.

  (A) A case in which a method or method for a residual signal when intra-frame prediction (intra prediction) is applied for each color component is independently applied as a method or method for performing prediction between color components.

  (B) As a method or method for performing prediction between color components, a method or method for the input image itself and a residual signal when intra-frame prediction (intra prediction) is applied for each color component are targeted. Cases where methods or methods are used together.

  Next, embodiments of the encoding apparatus in the cases (a) and (b) will be described individually. In the following, H.C. The image compression method such as the H.264 / AVC method will be described, but the present invention is not limited to this image compression method, and other image compression methods may be used.

  First, the configuration and operation of the first embodiment (case (a)) of the present invention will be described with reference to the block diagram of FIG. Note that the processing described below is repeated for each macroblock.

  When an input image 1 having an independent luminance signal, such as 4: 2: 2 video or 4: 2: 0 video, is input, the preprocessing analysis unit 2 causes the color components (luminance signal and color difference signal) of the input image 1 to be input. For example, the resolution is analyzed. The reference signal selection unit 3 selects a color component (reference component) as a reference signal. For example, according to the resolution of the color component obtained by the preprocessing analysis unit 2, the color component having the highest resolution is selected as the reference signal. In general, the luminance signal (Y) is often a reference signal (reference component), and the color difference signals (Cb, Cr) are often non-reference signals (non-reference components).

  The reference signal a is sent to the intra prediction unit 4, while the non-reference signal b is sent to the subtractor 11 and the intra prediction unit 12. The intra prediction units 4 and 12 are, for example, H.264. H.264 / AVC system intra prediction is performed. The reference signal a is intra-predicted by the intra-prediction unit 4, and the prediction mode with the minimum code amount, the minimum RD cost, and / or the minimum error power is selected. Next, the reference signal a is orthogonally transformed by the DCT unit 5, and the orthogonal transformation coefficient is quantized by the quantization unit 6. Subsequently, the variable length coding unit 7 performs variable length coding and outputs the result. The signal variable-length encoded by the variable-length encoder 7 is input to the intra decoder 8 and is intra-decoded. From the intra decoding unit 8, the intra decoded image c of the reference component is output. This intra decoded image c of the reference component is sent to the first inter-component prediction unit 14. The reason will be described later. Note that a local decoded image c obtained by performing inverse quantization and inverse DCT on the signal quantized by the quantization unit 6 without performing intra decoding from the signal subjected to variable length coding by the variable length coding unit 7 is the first. You may make it send to the prediction part 14 between these.

  Next, the non-reference signal b is intra-predicted by the intra prediction unit 12, and a prediction mode with the minimum code amount, the minimum RD cost, and / or the minimum error power is selected. H. In the H.264 / AVC format, for example, there are nine types of prediction modes 0 to 8 for a block of 4 × 4 pixels, and therefore the minimum prediction mode such as the code amount is selected. The selected prediction mode d is notified to the first inter-component prediction unit 14 and sent to the variable length coding unit 19. Further, the intra prediction image 13 obtained according to the selected prediction mode is subtracted from the non-reference signal b by the subtractor 11 to obtain an intra-frame prediction residual signal e. The residual signal e is sent to the subtracter 16 and also sent to the first inter-component prediction unit 14.

  Here, a specific example of the first inter-component prediction unit 14 will be described with reference to the block diagram of FIG. The first inter-component prediction unit 14 includes a resolution conversion unit 14a that receives the intra decoded image c or the local decoded image c of the reference component, and a conversion converted to the same resolution as the non-reference signal b by the resolution conversion unit 14a. The post-reference image 14b, the intra-prediction unit 14c for intra-predicting the converted reference image 14b in the prediction mode d, the intra-prediction image 14d obtained according to the intra-prediction, the post-conversion reference image 14b and the intra-prediction image 14d A subtractor 14e that obtains the difference x and a linear transformation unit 14f that outputs a residual signal y after linear transformation obtained by linear transformation of the prediction residual x, that is, an output y, are formed. Here, the resolution conversion unit 14a converts the intra decoded image c or the local decoded image c of the reference signal so that, for example, the number of samples constituting the screen is the same as that of the non-reference signal. The linear conversion unit 14f performs linear conversion of y = αx + β. The prediction coefficients α and β are determined so that the output y is close to the residual signal e of the intra-frame prediction. The prediction coefficients α and β are sent to the variable length coding unit 19 as the first prediction coefficient z1. The output y is the same as the reference component prediction residual image 15.

  Next, the reference component prediction residual image 15 obtained as described above is subtracted from the residual signal e of the intra-frame prediction by the subtracter 16, and the prediction error between color components of the residual signal of the intra-frame prediction is obtained. A signal f is formed. As described above, the reference component prediction residual image 15 is subjected to the same intra prediction as the intra prediction of the non-reference signal after being made equal to the resolution of the non-reference signal by the resolution conversion unit 14a. Since the component prediction residual signal 15 is obtained, it is possible to reduce the performance degradation of inter-color component prediction for the residual signal. Further, since the residual signal is corrected so as to be close to the residual signal e by the linear conversion unit 14f, the inter-color component prediction error signal f is further reduced and the code amount is reduced.

  Next, the inter-color component prediction error signal f is DCTed by the DCT unit 17, quantized by the quantizing unit 18, further variable-length coded by the variable-length coding unit 19, and variable-length coded. It is output together with the first prediction coefficient z1 and the prediction mode d.

  Further, the variable length encoded error signal g is locally decoded by the local decoding unit 20 and sent to the adder 21. The adder 21 adds the error signal g to the reference component prediction image 15, so that a decoded residual image 22 is obtained. It is formed. The decoded residual image 22 is added to the intra predicted image 13 by the adder 23 to obtain an intra decoded image (non-reference component) 24. The intra decoded image 24 is sent to the intra prediction unit 12 and is used for intra prediction as a reference pixel. The above processing is sequentially performed on the two non-reference signals.

  As described above, according to the present embodiment, even if the resolution between color components is different, the resolution of one color component is made equal to the resolution of the other color component and applied to the other color component. Since the processing is performed after performing the prediction, it is possible to greatly reduce the decrease in the correlation of the prediction between the color components as compared with the conventional method. Moreover, since the prediction between color components can be functioned with high accuracy, it is possible to expect an improvement in coding performance. According to the experiment results of the present inventors, it was confirmed that the code amount was reduced by about 10%.

  Next, the configuration and operation of the second embodiment (the case (b)) of the present invention will be described with reference to the block diagram of FIG. 3, the same reference numerals as those in FIG. 1 denote the same or equivalent parts.

  In this embodiment, a method or method for the input image itself is added to the first embodiment, and in order to implement this, the second inter-component prediction unit 31 is added to the apparatus of FIG. A subtractor 32 and a reference component prediction image 33 are added.

  As shown in FIG. 4, the second inter-component prediction unit 31 includes a resolution conversion unit 31a and a linear conversion unit 31b. The intra decoded or local decoded image c of the reference component is input to the resolution conversion unit 31a and converted to the same resolution as the non-reference signal b. The reference signal after resolution conversion is input to the linear conversion unit 31b, and linear conversion processing similar to that performed by the linear conversion unit 14f in FIG. 2 is performed. That is, linear conversion of y = γx + δ (where γ and δ are weighting coefficients or prediction coefficients) is performed so that the reference signal x after resolution conversion approximates to the non-reference signal b (generally Cb or Cr). , Output as y. The prediction coefficients γ and δ are sent to the variable length coding unit 19 as the second prediction coefficient z2. The output y is the same as the reference component predicted image 33.

  The subtracter 32 subtracts the component prediction image 33 from the non-reference signal b to form a residual signal i. The residual signal i is sent to the subtractor 11 and the intra prediction unit 12, and an intra prediction image 13 is obtained from the intra prediction unit 12. Subsequent processing is the same as that in the first embodiment, and thus description thereof is omitted. Note that the intra predicted image 13 and the intra decoded image 24 in FIG. 1 are changed to the intra predicted image 13 and the intra decoded image 24, respectively.

  Also in this embodiment, the resolution conversion unit 31a of the second inter-component prediction unit 31 makes the resolution of the reference signal equal to that of the non-reference signal, and the subtracter 32 subtracts it from the non-reference signal. Therefore, since the reference signal and the non-reference signal can be correlated, the correlation between the reference signal and the non-reference signal can be significantly increased as compared with the conventional method. In other words, it is possible to significantly reduce the decrease in correlation between color component predictions as compared to the conventional method. Also, the code amount can be greatly reduced.

  Next, the decoding device of the present invention will be described. First, the configuration and operation of the third embodiment of the present invention (case (a)) will be described with reference to the block diagram of FIG.

  As shown in the figure, when the encoded output of the reference signal (reference color component) and the non-reference signal (non-reference color component) output from the encoding device of FIG. If the data is stored in a storage device via a line or is read from the storage device, it is input as an input stream 41 to the decoding device. The input stream 41 is discriminated by the reference signal discriminating unit 42 into a reference signal m and a non-reference signal n as determined at the time of encoding. The stream of the reference signal m is sent to the variable length decoding unit 43 and subjected to variable length decoding, and then inverse quantization and inverse DCT are sequentially performed by the inverse quantization unit 44 and the inverse DCT unit 45. By these processes, the reference signal is output as the intra decoded image 46. The intra decoded image 46 is also sent to the first inter-component prediction unit 55.

  The stream of the non-reference signal n is sent to the variable length decoding unit 51. By the variable length decoding, a variable length decoded image signal, a prediction mode d, and a first prediction coefficient z1 are obtained. The variable length decoded image signal is sent to the inverse quantization unit 52, and the prediction mode d and the first prediction coefficient z1 are sent to the first inter-component prediction unit 55. The prediction mode d is also sent to the intra prediction unit 62.

  Here, a specific example of the first inter-component prediction unit 55 will be described with reference to FIG. The first inter-component prediction unit 55 has almost the same configuration as that shown in FIG. 2, and includes a resolution conversion unit 55a, a post-conversion reference image 55b, an intra prediction unit 55c, an intra prediction image 55d, a subtractor 55e, and a linear conversion unit 55f. Composed. Here, these configurations correspond to the resolution conversion unit 14a, the converted reference image 14b, the intra prediction unit 14c, the intra prediction image 14d, the subtractor 14e, and the linear conversion unit 14f in FIG. The only difference from FIG. 2 is that the linear prediction unit 55f is given the first prediction coefficient z1 obtained by the processing of the variable length decoding unit 51, that is, the prediction coefficients α and β. Only the operation will be described, and the description of the operation of other processing units will be omitted. When the prediction coefficients α and β are given, the linear conversion unit 55f performs the operation y = αx + β used in the linear conversion unit 14f in FIG. 2 to reproduce the output y, that is, the reference component prediction residual image 56.

  Returning to FIG. 5 again, the description is continued. The variable-length decoded image signal is inversely quantized by the inverse quantization unit 52 and then inversely DCTed by the inverse DCT unit 52 to obtain an error signal of the decoding result 54. The error signal of the decoding result 54 corresponds to the prediction error signal f in FIG. Next, the error signal of the decoding result 54 is added to the reference component prediction residual signal 56 by the adder 57, and a decoded residual image 58 is obtained. The decoded residual image 58 corresponds to the residual signal e in FIG.

  Subsequently, the decoded residual image 58 is input to the adder 59 and added to the intra predicted image 60. As a result, an intra decoded image 61, that is, an intra decoded image 61 of a non-reference signal is obtained. The intra decoded image 61 is output as a decoded image and is sent to the intra prediction unit 62 as a reference pixel. The intra prediction unit 62 performs intra prediction in the prediction mode d notified from the variable length decoding unit 51. The intra decoded image 61 corresponds to the non-reference signal b in FIG.

  As described above, the reference signal and the non-reference signal encoded by the encoding device in FIG. 1 can be decoded, and the reference signal and the non-reference signal can be correlated. Be able to get.

  Next, the configuration and operation of the fourth embodiment of the present invention (case (b)) will be described with reference to the block diagram of FIG. The second inter-component prediction unit 71 is substantially the same as the configuration of FIG. 4 and includes a resolution conversion unit 71a and a linear conversion unit 71b as shown in FIG. Here, these configurations correspond to the resolution conversion unit 31a and the linear conversion unit 31b in FIG. The only difference from FIG. 4 is that the second prediction coefficient z2 obtained by the process of the variable length decoding unit 51, that is, the prediction coefficients γ and δ, is given to the linear conversion unit 71b. When the prediction coefficients γ and δ are given, the linear conversion unit 71b calculates y = γx + δ used in the linear conversion unit 31b of FIG. 4 and reproduces the output y, that is, the reference component prediction signal 72.

  Returning to FIG. 7 again, the adder 73 adds the intra decoded image 61 to the reference component predicted image 72 to obtain a decoded image of the non-reference signal. Here, the decoded residual image 58 and the intra decoded image 61 of FIG. 7 are the residual signal e and the residual signal i of FIG. 3, respectively, and the decoded non-reference image that is the output of the adder 73 is FIG. Correspond to the non-reference signal b.

  As described above, according to the fourth embodiment, the first and second inter-component prediction units 55 and 71 can decode the reference signal and the non-reference signal after correlating the reference signal and the non-reference signal. Therefore, a decoded image with good image quality can be obtained.

  DESCRIPTION OF SYMBOLS 1 ... Input image, 3 ... Reference signal selection part, 8 ... Intra decoding part, 14 ... 1st component prediction part, 14a ... Resolution conversion part, 14c ... Intra prediction , 14f: linear conversion unit, 31: second inter-component prediction unit, 31a: resolution conversion unit, 31b: linear conversion unit, 41: input stream, 42: reference Signal discriminating unit, 55 ... first inter-component prediction unit, 55a ... resolution conversion unit, 55f ... linear conversion unit, 71 ... second inter-component prediction unit, 71a ... resolution conversion Part, 71b ... linear conversion part.

Claims (8)

In an inter-color component prediction type image encoding device capable of adaptively setting a reference signal for inter-color component prediction,
A reference signal selection unit (3) for selecting a reference signal and a non-reference signal from an input image signal composed of a plurality of color components;
A first intra prediction unit (12) for intra-predicting the non-reference signal to obtain a prediction mode;
A first resolution converter (14a) that converts the resolution of an intra decoded image or a locally decoded image of the reference signal (hereinafter referred to as “reference decoded image”) to be equal to the resolution of the non-reference signal;
A second intra prediction unit (14c) for intra-predicting the resolution-converted reference decoded image in a prediction mode obtained by the first intra prediction unit;
Wherein the resolution transformed reference decoded image of said second subtracting the intra prediction image predicted by intra prediction unit obtained reference component predicted residual image (15), the intra prediction of the non-reference signal remaining A subtractor (16) for subtracting from the difference signal (e);
An inter-color component prediction type image encoding apparatus comprising: means for performing an encoding process on the inter-color component prediction error signal (f) obtained by the subtraction. In the inter-color component prediction type image encoding device according to claim 1,
The reference component prediction residual image is obtained by performing linear conversion on a signal obtained by subtracting the intra prediction image predicted by the second intra prediction unit (14c) from the resolution-decoded reference decoded image (14b). An inter-color component predictive image coding apparatus, further comprising a first linear conversion unit (14f) for obtaining (15). The inter-color component prediction type image encoding device according to claim 1 or 2,
Before the first intra prediction unit (12),
A second resolution converter (31a) that converts the resolution of the reference decoded image so as to be equal to the resolution of the non-reference signal, and a reference component prediction image obtained by the second resolution converter from the non-reference signal An inter-color component prediction type image encoding apparatus, comprising: a subtracter (32) for subtracting (33). The inter-color component prediction type image encoding device according to claim 3,
The image processing apparatus further includes a second linear conversion unit (31b) that performs linear conversion on the reference decoded image whose resolution is converted by the second resolution conversion unit (31a) to obtain the reference component predicted image (33). An inter-color component prediction type image encoding apparatus. The inter-color component prediction type image encoding device according to any one of claims 1 to 4,
The inter-color component prediction type image encoding apparatus, wherein the reference signal selection unit (3) selects a color component having a maximum resolution as the reference signal. In the decoding device of the signal encoded by the inter-color component prediction type image encoding device according to claim 1 or 2,
A discriminator (42) for discriminating a reference signal and a non-reference signal from the encoded signal;
A third resolution converter (55a) for converting the resolution of the intra decoded image or the locally decoded image of the reference signal (hereinafter referred to as “reference decoded image”) to be equal to the resolution of the non-reference signal;
A third intra prediction unit (55c) for intra-predicting the resolution-converted reference decoded image in a prediction mode (d) extracted from the variable-length decoding of the non-reference signal;
A reference component prediction residual image (56) obtained by subtracting the intra prediction image predicted by the third intra prediction unit from the resolution-converted reference decoded image and a variable length decoding result of the non-reference signal ( 54) an adder (57) that adds the residual signal to obtain a decoded residual image (58);
An adder (59) for adding the intra prediction image (60) obtained by intra prediction in the prediction mode (d) extracted from the variable-length decoding of the non-reference signal and the decoded residual image (58); A decoding apparatus comprising: The decoding device according to claim 6, wherein
First extracted from the variable-length decoding of the non-reference signal to a signal obtained by subtracting the intra-predicted image predicted by the third intra-prediction unit (55c) from the resolution-converted reference decoded image. A decoding apparatus, further comprising: a third linear transformation unit (55f) that performs linear transformation to which the prediction coefficient is applied to obtain the reference component prediction residual image (56). The decoding device according to claim 6, wherein
A fourth resolution conversion unit (71a) that converts the resolution of the reference decoded image to be equal to the resolution of the non-reference signal and the second prediction coefficient extracted from the variable-length decoding of the non-reference signal are applied. A fourth linear transformation unit (71b) that performs linear transformation to obtain a reference component predicted image (72), and an intra decoded image (61) is added to the reference component predicted image (72) to obtain a decoded image of a non-reference signal The decoding device further comprising an adder (73) for obtaining

Images