TWI482501B - Fast mode decision method for scalable video coding multi-layer encoder control - Google Patents

Description

Adjustable video multi-layer coding control fast decision mode algorithm

The invention relates to a video coding algorithm, and more particularly to an adjustable video multi-layer coding control fast decision mode algorithm. By using the concept of Temporal Mode Score (TMS) to more accurately predict the coding mode of the enhancement layer, and accurately find the coding mode of the base layer and the enhancement layer and can follow the video sequence. The nature of the different pairing calculations, with lower coding distortion, reduced encoding time, adjustability and encoding mode with better bit rate-distortion performance (Rate-Distortion performance).

Due to the increasing variety of devices with varying capacities and the increasing number of video playback services on the network, single video encoding is no longer sufficient for these projects. For this reason, the Joint Video Team (JVT) has developed a new standard, Scalable Video Coding (SVC) based on H.264/AVC, which provides a single bit stream. The spatial, temporal, quality, and tunability of the combination between the three, using this tunable property to encode a single bit stream, can get different resolution and different quality services.

The tunable video coding has a single base layer and one or more enhancement layers, and the enhancement layer utilizes the information of the base layer as much as possible, and the coded multi-layer can achieve a reduced bit rate and can also achieve the expected picture quality. Traditional Adjustable Video Coding When determining the coding mode of the base layer and the enhancement layer, the Joint Scalable Video Model (JSVM) adopts a Bottom up encoder control. . In this encoding method, only the performance of the lower layer between the two layers is considered. Therefore, the traditional coding method will cause more serious distortion between the two layers, resulting in the selection of the base layer (BL) coding mode, without considering the enhancement layer (Enhancement Layer, EL for short). The coding efficiency, therefore, leads to unpredictable distortion errors in the enhancement layer.

In the early days, Schwarz et al. provided a method for giving appropriate weights between the base layer and the promotion layer, called Multilayer Encoder Control. When making coding mode decisions, the base layer uses a weighted Lagrangian cost decision method. As shown in Fig. 1, in the multi-layer coding mode decision, the basic unit is a pair of viewing blocks (Macroblock, MB), which are derived from the base layer and the relative position of the promotion layer. For the coding mode between the two layers, there are a total of 56 pairs. If you calculate the coding mode of these 56 groups and find the best one among them, compared with the traditional bottom-up coding method, there is obviously A lot of calculations, it can be seen that a good fast decision mode approach is very important.

Schwarz et al. proposed a two-stage fast coding mode decision algorithm. The approach is that when the base layer of the base layer makes the coding mode decision, each coding mode has a coding mode of the enhancement layer as a pairing. This pairing is called Dominant Mode Pair. As shown in FIG. 2, the pairing method is that when the base layer selects an inter prediction mode (Inter Prediciton) or an intra prediction (Intra Prediction) coding mode, the promotion layer correspondingly selects BL_skip or IntraBL as a prediction, and then the base layer utilizes this. Group pairing is used to calculate the bit rate-distortion-cost (Rate-Distortion-cost, RD-cost for short), and the coding mode with the smallest RD-cost is selected as the coding mode of the base layer, and finally the corresponding view in the promotion layer The block then performs a complete search based on the selected base layer coding mode, and finds the smallest RD-cost coding mode for compression. But this method has some serious problems:

(1) When selecting the coding mode of each block in the base layer, the 7 sets of pairing combinations in Fig. 2 are used for calculation, and the adjustment is not performed according to the characteristics of the video sequence. This pairing method is too monotonous and does not have any adjustability.

(2) This pairing mode may be different from the actual encoding mode because it is only 7 of the 56 pairing modes. If we use this method to estimate the coding mode of the enhancement layer, it may be too different from the coding mode of the actual enhancement layer, so it will cause the base layer to make an error in the selection of the coding mode. If this error is used, the base layer coding is used again. The mode to find the coding mode in the enhancement layer will cause unpredictable severe distortion.

In view of the shortcomings of the prior art, it is necessary to propose an adjustable image processing method. By using the temporal coding mode score, the amount of computation required for encoding can be reduced and the above mentioned problems can be solved.

In view of the above disadvantages of the prior art, the main object of the present invention is to provide a scalable video multi-layer coding control fast decision mode algorithm, which can determine whether the sub-band of wavelet transform is suitable for directional disassembly, and then according to all directions. The signal characteristics on the frequency band propose a new adjacent table to achieve better compression.

To achieve the above objective, the present invention provides an adjustable video multi-layer coding control fast decision mode algorithm for compression and encoding of an image signal, the method comprising at least the following steps: Step (1): predicting the first of a hypothesis The coding mode of the viewing block in the enhancement layer; step (2): predicting the coding mode of the first picture of the hypothesis in the enhancement layer, and using the second picture of the hypothesis in the enhancement layer The coding mode of the block or the second picture of the plurality of hypotheses has an encoding mode of the viewing block around the encoding mode of the viewing block of the enhancement layer; step (3): using a calculation method of a temporal coding mode, The second picture of the hypothesis is calculated by the encoding mode of the viewing block of the enhancement layer and the plurality of second pictures of the hypothesis, and the coding mode of the viewing block with the viewing block of the enhancement layer is calculated. Obtaining a temporal coding mode score; step (4): determining a probability distribution according to the temporal coding mode score and a plurality of video continuation contents; and step (5): determining a first according to the probability distribution a possible combination of the coding mode of the viewing block in the enhancement layer and the coding mode of the viewing block of the first picture in the base layer; step (6): viewing blocks in the enhancement layer for the first picture The coding mode is combined with the possible coding mode of the first picture in the base layer, and the bit rate-distortion cost is calculated; step (7): according to the first bit rate-distortion cost A coding mode of the viewing block in the base layer determines a coding mode of the first picture in the base layer; and step (8): according to the first picture, the view block in the base layer Encoding mode and all feasible coding modes of the first picture in the enhancement layer, performing bit rate-distortion cost calculation; step (9): according to having a minimum bit rate-distortion cost first The coding mode of the picture in the enhancement layer determines the coding mode of the first picture in the enhancement layer.

In summary, the adjustable video multi-layer coding control fast decision mode algorithm of the present invention has the following effects:

1. Use the nature of Temporal to predict the coding mode in the enhancement layer, and then pair this prediction with the coding mode of the base layer. First, we predict the coding mode of the enhancement layer. Therefore, there is no fixed Dominant Mode Pair concept. When selecting the coding mode of the base layer, different viewing blocks may be calculated with different enhancement layer coding modes to reduce the coding time.

2. Providing the concept of Temporal Mode Score (TMS), TMS can more accurately predict the coding mode of the enhancement layer, and accurately find the coding mode of the base layer and the enhancement layer.

3. According to the nature of the video sequence, each of the viewing blocks first adjusts the encoding mode with the highest probability of the layer prediction as the base layer pairing, so it is adjustable.

The above and other objects, features, and advantages of the present invention will become more apparent and understood.

While the invention may be embodied in various forms, the embodiments illustrated in the drawings It is not intended to limit the invention to the particular embodiments illustrated and/or described.

Referring now to Figure 3, there is shown a schematic diagram of the architecture of the fast decision mode method of the present invention. In the figure, A is the viewing block of the current picture in the base layer, B is the viewing block relative to the A position in the promotion layer, and C is the nine pictures of the previous picture in the promotion layer relative to B. Block location.

Referring now to Figure 4, there is shown a flow chart of the steps of the adjustable video multi-layer coding control fast decision mode algorithm of the present invention, which mainly comprises nine steps.

Firstly, in step (1), it is a coding mode for predicting the first picture of the hypothesis in the enhancement layer, mainly because it is necessary to predict the enhancement when searching for the coding mode of the observation layer of the base layer. The coding mode of the layer with the view block.

In the step (2), the coding mode of the first picture of the hypothesis is predicted in the enhancement layer, and the second picture of the hypothesis is used in the coding mode of the enhancement layer or the plurality of hypotheses. The second picture has an encoding mode of the viewing block around the encoding mode of the viewing block of the enhancement layer.

In the step (3), the calculation method of the temporal coding mode is used, and the second picture of the hypothesis is encoded in the enhancement layer and the second picture of the hypothesis is in the enhancement layer. The coding mode of the viewing block is calculated by the coding mode of the viewing block, and a Temporal Mode Score (TMS) can be obtained.

Referring now to FIG. 5, there is shown an encoding mode selected by a viewing block (EL) of the present invention in a corresponding nine-frame position of the second picture, the figure showing one of the first picture base layers. The position of the viewing block corresponding to the viewing layer in the first picture enhancement layer is E. That is, if a coding mode with a viewing block is to be determined in the base layer, it is necessary to first find the encoding mode predicted by the viewing block of the E position as a pairing of the first picture base layer. In addition, the coding mode of the first picture E position hypothesis is that the coding mode of the viewing block using the second picture in the enhancement layer and the plurality of second pictures of the hypothesis are around the coding mode of the viewing block of the enhancement layer. The coding mode of the block. That is, the coding mode assumed by E in the figure is calculated by using the coding mode of the A ^' ~ I ^' position in the figure, and the calculation method of the temporal coding mode with the relative position of the block in Fig. 6(a). Further, a schematic diagram of the temporal coding mode scores of the various coding modes of FIG. 6(b) can be obtained.

It should be noted that the calculation method of the temporal coding mode is TMS = A ^' × 1 + B ^' × 2 C ^' × 1 + D ^' × 2 E ^' × 4 + F ^' × 2 G ^' × 1+ H ^' × 2 I ^' ×1, in addition, A', B', C', D', E', F', G', H', I' are the second picture of this hypothesis The coding mode selected by the nine-grid position corresponding to the coding mode of the layer. The temporal coding mode score ranges from 0 to 64. If A', B', C', D', E', F', G', H', I' appear in the intra-prediction mode, the value is 0.

In step (4), the probability distribution is determined according to the temporal coding mode score and the plurality of video continuation contents. The plurality of video continuation contents include Akiyo, Carphone, City, Coastguard, Highway, Stefan, and Walk video.

In the step (5), according to the probability distribution, a possible combination of the coding mode of the viewing block of the first picture in the enhancement layer and the coding mode of the viewing block of the first picture in the base layer is determined.

In step (6), the bit rate-distortion is performed for a possible combination of the encoding mode of the viewing block of the first picture in the enhancement layer and the encoding mode of the viewing layer of the first picture in the base layer. Calculation of cost.

In the step (7), the coding mode of the viewing block of the first picture in the base layer is determined according to the coding mode of the first picture in the base layer with a minimum bit rate-distortion cost. .

In step (8), the bit rate-distortion is performed according to the coding mode of the first picture in the base layer and the coding mode of the viewable block in the enhancement layer of the first picture. Calculation of cost.

In the step (9), the coding mode of the viewing block of the first picture in the enhancement layer is determined according to the coding mode of the first picture in the enhancement layer of the first picture having a minimum bit rate-distortion cost. .

In order to better understand the spirit of the present invention, a description of the fast decision mode method is made herein with a first embodiment. If the coding mode of the first picture in the enhancement layer is Inter 8x8, please refer to the 6th (a) and 7th pictures, and calculate the hypothetical second picture in the enhancement layer. The encoding mode of the block and the TMS of the plurality of second pictures of the hypothesis in the coding mode of the viewing block of the enhancement layer with the viewing block are calculated as follows:

TMS =0×1+1×2+1×1+1×2+4×4+2×2+2×1+2×2+1×1=32

Next, please refer to Figure 8 according to the probability distribution corresponding to TMS=32, and find the most probable coding mode as the possible combination of the coding mode of the first picture in the enhancement layer. Here, BL_skip is used as the possible combination of the coding mode of the viewing block of the first layer in the enhancement layer.

Referring now to FIG. 9, the possible combination of the coding modes of the viewing blocks in the enhancement layer according to the first picture is BL_skip, as shown by the solid line in the figure. Then, the bit rate-distortion cost is calculated for each possible coding mode of the solid line frame, and the code with the smallest bit rate-distortion cost is used as the first picture in the base layer. mode.

Please refer to FIG. 9 again, and calculate the bit rate-distortion cost according to the coding mode of the first layer in the base layer. Here, 16×16 is used as the first picture in the base layer. The coding mode, as indicated by the dashed box, calculates the bit rate-distortion cost for each possible coding mode in the dashed box, and uses the minimum bit rate-distortion cost as the first picture in the promotion layer. The coding mode of the block.

Please refer to Figure 7 again. Please refer to the brackets in the figure, that is, in the case of intra-picture prediction, the TMS is calculated as follows:

TMS =0×1+1×2+1×1+0×2+4×4+2×2+2×1+0×2+0×1=25

Since there is an intra-picture prediction coding mode, it is necessary to perform normalization. The method is:

Round [(25/11)×16]=36, where Round is the rounding operation, and dividing by 11 is due to its position in the nine grids, and the score of the inter-picture predictive coding mode position is increased to 11. Therefore, with the probability distribution corresponding to TMS=36, the coding mode with the highest probability is found as the base layer pairing.

In summary, the fast decision mode of the multi-layer coding control of the adjustable video coding of the present invention has the following effects:

1. Use the nature of Temporal to predict the coding mode in the enhancement layer, and then pair this prediction with the coding mode of the base layer. First, we predict the coding mode of the enhancement layer. Therefore, there is no fixed Dominant Mode Pair concept. When selecting the coding mode of the base layer, different viewing blocks may be calculated with different enhancement layer coding modes to reduce the coding time.

2. Provide the concept of Temporal Mode Score (TMS). TMS can more accurately predict the coding mode of the enhancement layer and accurately find the coding mode of the base layer and the enhancement layer.

3. According to the nature of the video sequence, each of the viewing blocks first adjusts the encoding mode with the highest probability of the layer prediction as the base layer pairing, so it is adjustable.

While the present invention has been described in its preferred embodiments, it is not intended to limit the scope of the invention, and various modifications and changes can be made without departing from the spirit and scope of the invention. As explained above, all types of corrections and changes can be made without destroying the spirit of this creation. Therefore, the scope of the invention is defined by the scope of the appended claims.

Figure 1 is the coding mode of the base layer (BL) and the enhancement layer (EL) of the technology of the prior art;

Figure 2 is the pairing rule of the coding mode of the algorithm proposed by Schwarz et al.

Figure 3 is a schematic diagram showing the architecture of the adjustable video multi-layer coding control fast decision mode algorithm of the present invention;

Figure 4 is a flow chart showing the steps of the fast decision mode method of the present invention;

Figure 5 is a coding mode selected by a viewing block in the enhancement layer (EL) of the present invention at a position corresponding to nine locations of the second picture;

Figure 6 is a schematic diagram of the (a) calculated fraction of the temporal coding mode of the relative position of the viewing block and (b) the temporal coding pattern score of the various coding modes;

Figure 7 is a nine-frame corresponding coding mode of the first embodiment of the present invention;

Figure 8 is a diagram showing the probability distribution of the plurality of video continuation sequences corresponding to the different temporal coding mode scores of the first embodiment of the present invention;

Figure 9 is a combination diagram of the coding modes of the base layer (BL) and the enhancement layer (EL) of the first embodiment of the present invention.

Claims (6)

A tunable video multi-layer coding control fast decision mode algorithm for compressing and encoding an image signal, the method comprising at least the following steps: Step (1): predicting a first picture of a hypothesis in a viewing area of the enhancement layer Encoding mode of the block; step (2): using the second picture of the hypothesis in the coding mode of the viewing block of the enhancement layer or the second picture of the plurality of hypotheses in the surrounding of the coding mode of the enhancement layer Observing the coding mode of the block; step (3): using a calculation method of the temporal coding mode, the second picture of the hypothesis is encoded in the enhancement layer and the second picture of the hypothesis The coding mode of the viewing block of the enhancement layer is calculated by the coding mode of the viewing block, and a time-coded coding mode score is obtained; step (4): according to the temporal coding mode score and a plurality of video continuations Column content, determining a probability distribution; step (5): determining, according to the probability distribution, a coding mode of the first picture in the enhancement layer and a coding mode of the first picture in the base layer Possible combination; step (6): performing bit rate-distortion for the possible combination of the encoding mode of the viewing block of the first picture in the enhancement layer and the encoding mode of the first picture in the base layer Cost calculation; step (7): determining the encoding of the viewing block of the first picture in the base layer according to the coding mode of the first picture in the base layer with a minimum bit rate-distortion cost Mode; step (8): performing bit rate-distortion cost according to the coding mode of the first picture in the base layer and the coding mode of the viewable block of the first picture in the enhancement layer The calculation (step) (9): determining the coding mode of the viewing block of the first picture in the enhancement layer according to the coding mode of the first picture in the enhancement layer of the first picture with a minimum bit rate-distortion cost . For example, the tunable video multi-layer coding control fast decision mode algorithm of the first application patent scope, wherein the time coding mode is calculated as TMS = A ^' × 1 + B ^' × 2 C ^' × 1 + D ^' ×2+ E ^' ×4+ F ^' ×2+ G ^' ×1+ H ^' ×2+ I ^' ×1; where A', B', C', D', E', F', G'H',I' is the coding mode selected by the second picture of the hypothesis in the nine-grid position corresponding to the coding mode of the viewing block of the enhancement layer. For example, the tunable video multi-layer coding control fast decision mode algorithm of claim 2, wherein the action of step (1) is performed before searching for the coding mode of the base layer. For example, the adjustable video multi-layer coding control fast decision mode algorithm of the second application of the patent scope 'A',B',C',D',E', F',G',H',I' is on the screen The value of Inter prediction (Inter Prediciton) or intra-prediction (Intra Prediciton) is 0. For example, the tunable video multi-layer coding control fast decision mode algorithm of claim 1 of the patent scope, wherein the temporal coding mode score ranges from 0 to 64. For example, the tunable video multi-layer coding control fast decision mode algorithm of claim 1 of the patent scope includes the Akiyo, Carphone, City, Coastguard, Highway, Stefan, and Walk video continuation.