KR101051331B1 - Adaptive Fast Mode Decision Method Using Dynamic Threshold - Google Patents

Abstract

The present invention relates to video encoding. In the fast mode determination method using the dynamic threshold according to the present invention, mode determination is performed early by determining whether an optimal mode of a current block is a skip mode or a 16 × 16 block size mode by using neighboring block information of a current block to be encoded. Quit. This can result in significant time savings.

In addition, the RDC values for the skip mode and the 16x16 block size mode are compared with the cost values of the 8x16 block size mode, so that there is an opportunity to end the mode decision early again.

Accordingly, the present invention provides a time gain in video encoding by providing an opportunity for terminating the mode decision early, and furthermore, has little effect on the compression performance.

Video encoding, fast algorithm, mode determination

Description

Adaptive Fast Mode Decision Method using Dynamic Thresholds

The present invention relates to a method for determining a mode at high speed in video encoding.

Recently, the compression transmission and encoding technology of digital video is rapidly progressing with the development of communication technology such as broadcasting and internet. In particular, various specifications have been proposed in the field of compression or encoding of digital video, and various efforts have been made to improve performance.

New video coding jointly established by the Joint Video Team (JVT), recently formed by the ITU-T Video Coding Experts Group (VCEG) and the ISO / IEC Motion Picture Expert Group (MPEG). As a standard, H.264 / AVC has been proposed.

H.264 / AVC has a high encoding performance capable of transmitting high quality video over low bit rate channels. It is also applicable to wide spectrum video applications, enabling new video services over cable, satellite and mobile networks. In 2004, Europe approved H.264 / AVC as a digital video broadcasting (DVB) standard for broadcast TV standards. In Korea, H.264 / AVC standards are used for DMB services.

The H.264 / AVC specification generally performs a number of encoding functions, such as motion prediction / compensation, intra prediction, transform, quantization, and entropy coding. The complexity of the transform and quantization processes in H.264 / AVC is relatively low, due to the small size of the transform block compared to other compression processes. However, the process of selecting the optimal mode for each macroblock (MB) requires a lot of computation. The H.264 / AVC encoder uses adjacent samples of the current frame for intra prediction or samples of previous frame (s) for inter prediction. Seven macroblock modes are available for P slices.

1 shows a variable block size of the H.264 / AVC standard. Since the mode decision method affects compression efficiency, bit rate, and encoding time, such a variable encoding mode is an important problem in mode decision. Calculating for a variable block size is like finding the optimal value through a very accurate calculation. To this end, H.264 / AVC uses a rate distortion optimization (RDO) technique. However, RDO shows good performance in terms of visual quality and coding bit rate, but has a disadvantage in that the operation is complicated. In order to determine the optimal size of the block, the rate distortion cost (RDC) is calculated using the RDO method, and Equation 1 below represents the equation.

In the above equation, J is the distortion distortion cost (RDC), M is the mode, and D is the distortion produced by the sum of the products of the differences. R is the rate and λ is the Lagrange product. By using the above technique of H.264 / AVC, we can have accurate prediction and good performance, but it has the disadvantage of requiring complicated operation to find the optimal mode of all modes. Due to these shortcomings, H.264 / AVC has better performance than conventional coding techniques, but there are many problems in applying it to real-time applications requiring low complexity.

Currently, many algorithms have been proposed to solve this problem. Known inter-mode decision algorithms are methods of using correlation between previous and current frames, algorithms for filtering unnecessary modes for skipped mode prediction, and fast mode decision algorithms for DCT coefficients.

However, these conventional inter-mode decision algorithms have a performance reduction that is not negligible in terms of PSNR or bit rate, which is a measure of the performance of the rest of the image focused only on speed improvement. In addition, the application algorithm is often too complicated for the speed improvement amount.

Accordingly, an object of the present invention is to provide a video encoding method in which a speed improvement is noticeable in video encoding, but hardly affects compression performance.

According to the present invention, the object is achieved by determining whether the optimal mode of the current block is the skip mode or the 16 × 16 block size mode by using the neighboring block information of the current block to be encoded, and ending the mode decision early. Can be. This can result in significant time savings.

On the other hand, the above object is to compare the cost values for the skip mode and the 16 × 16 block size mode with the cost values of the 8 × 16 block size mode, and thus once again have the opportunity to terminate the mode decision early. By having it. In this way, a time gain can be obtained by providing an opportunity to terminate the mode decision early.
According to the present invention, in the adaptive fast mode determination method using the dynamic threshold, (a) the rate distortion cost (RDC ₀ ) and the 16 × 16 block for the skip mode (SKIP MODE) of the current block for encoding Calculating a rate distortion cost RDC ₁ for the size mode; calculating a mean value RDC _L of a rate distortion cost of a neighboring block having a skip mode or a 16 × 16 block size mode as an optimal mode among neighboring blocks adjacent to the current block that is already encoded; (c) calculating an average value (RDC _H ) of a rate distortion cost of a neighboring block having a mode other than a skip mode and a 16x16 block size mode among the neighboring blocks adjacent to the current block that are already encoded as an optimal mode; (d) determining whether to terminate the mode decision early by comparing the RDC _L with the RDC _H , comparing the RDC ₀ with the RDC _L , and comparing the RDC ₁ with the RDC _L. It can be achieved by an adaptive fast mode decision method using a dynamic threshold.
In the step (d), when the RDC _L is smaller than the RDC _H , the RDC ₀ is smaller than the RDC _L , and the RDC ₁ is smaller than the RDC _L , the RDC ₀ is smaller than the RDC ₁ . Determining a mode having an optimal mode of the current block and ending mode determination early, and wherein the RDC _L is greater than the RDC _H , the RDC ₀ is greater than the RDC _L , or the RDC ₁ is the RDC. _If greater than _L , calculating the rate distortion cost (RDC ₃ ) for the 8 × 16 block size mode, and comparing the RDC ₀ , the RDC ₁ , and the RDC ₃ again to determine whether to terminate the mode decision early By including, the mode determination of the current block can be dynamically and adaptively performed according to the image characteristics of the neighboring blocks.
Here, re-determining whether to terminate the mode decision early may include: when the RDC ₀ is smaller than the RDC ₃ and the RDC ₁ is smaller than the RDC ₃ , the RDC ₀ is smaller than the RDC ₁ . Determining a mode having an optimal mode of the current block and terminating the mode decision early; And if the RDC ₀ is larger than the RDC ₃ or the RDC ₁ is larger than the RDC ₃ , calculating a rate distortion cost for all remaining modes to determine a mode having a minimum value as an optimal mode. You have the opportunity to end the mode decision early.
On the other hand, the above object, according to the present invention, in the adaptive fast mode determination method using the dynamic threshold, (a) the rate distortion cost (RDC ₀ ) and 16 for the skip mode (SKIP MODE) of the current block for encoding Calculating a rate distortion cost (RDC ₁ ) for a x16 block size mode; (b) the skip mode or the optimal mode of the current block based on the optimal mode of the neighboring blocks adjacent to the current block and the rate distortion cost value of the optimal mode of the neighboring block and the RDC ₀ and RDC _1; Primarily determining whether to determine early in any of the 16x16 blocksize modes; And (c) if in step (b) is not the best mode of the current block is determined at an early stage, the calculated rate distortion cost (RDC ₃₎ of the current block 8 × 16 block size mode and the RDC ₀ and And adaptively determining whether to optimally determine the optimal mode of the current block as the skip mode or the 16 × 16 block size mode through comparison with the RDC ₁ . It can also be achieved by a fast mode determination method.
The step (b) may include: calculating an average value RDC _L of a rate distortion cost of a neighboring block having a skip mode or a 16 × 16 block size mode as an optimal mode among neighboring blocks adjacent to the current block that is already encoded; Calculating an average value (RDC _H ) of a rate distortion cost of a neighboring block having a mode other than a skip mode and a 16 × 16 block size mode among the neighboring blocks adjacent to the current block that are already encoded as an optimal mode; And when the RDC _L is smaller than the RDC _H , the RDC ₀ is smaller than the RDC _L , and the RDC ₁ is smaller than the RDC _L , comparing the RDC ₀ with the RDC ₁ has a mode having a smaller value. By determining the optimal mode of the block and prematurely terminating the mode decision, once again there is an opportunity to terminate the mode decision early.
In the step (c), when the RDC ₀ is smaller than the RDC ₃ and the RDC ₁ is smaller than the RDC ₃ , the mode having a smaller value is compared to the optimal mode of the current block by comparing the RDC ₀ with the RDC ₁ . Decision and early termination of the mode decision.
The adaptive fast mode determination method using the dynamic threshold may include determining whether the current block is a boundary block positioned at upper, lower, left, and right boundaries of a frame. If the current block is the boundary block, the optimal mode is determined by calculating a rate distortion cost for all modes, and if the current block is not the boundary block, steps (a) to (c) are performed to perform the boundary block. The invention applies only when this is not the case.

As described above, according to the present invention, there is provided a video encoding method in which speed improvement is noticeable but hardly affects compression performance.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings.

dynamic  Threshold ( Dynamic Throshold ) first  filter

In general, blocks having a lot of spatial information tend to be encoded in a skip mode (hereinafter referred to as mode 0) or 16 × 16 block size mode (hereinafter referred to as mode 1), and blocks having a large amount of temporal information are 16x8 blocks. There is a tendency to be encoded in a size mode (hereinafter referred to as mode 2), an 8x16 block size mode (hereinafter referred to as mode 3), and an 8x8 block size mode (hereinafter referred to as mode 8). These results can be easily seen from many related algorithms and experimental results. In addition, it is possible to extract quite accurate mode information on the optimum mode determination from the experimental results for the conventional mode determination method.

Tables 1 to 3 below show the distribution of optimal modes for various sequences.

The above sequences are all CIF images, Table 1 is QP: 28, Frames: 50, Table 2 is QP: 32, Frames: 50, Table 3 is QP: 36, Frames: 50 Experiment.

As can be seen from the above tables, it can be seen that the ratio of mode 0 to mode 1 is quite large in a sequence composed of frames having low motion such as "Akiyo" or "Container". On the other hand, in the high-motion sequences such as "Stefan" or "Coastguard", although the ratio of mode 0 to mode 1 is large, the ratio of mode 2, mode 3, and mode 8 is quite large.

However, it can be seen that the ratio of the sum of Mode 0 and Mode 1 is more than 50% in common. In light of these facts, we can see that we can significantly reduce the encoding time if we can find Mode 0 and Mode 1 effectively.

The present invention has been conceived based on this fact, and has proposed a dynamic threshold technique.

The present invention uses the peripheral spatial information shown in FIG. 2 to find Mode 0 and Mode 1 which occupy most of the optimal mode early.

Whether the optimal mode of the current block is mode 0 or mode 1 based on the four spatially closest adjacent blocks (LEFT, LEFT-UP, UP, and UP-RIGHT) based on the current block (Current MB) to be encoded. Determine. Since these four neighboring blocks are already encoded blocks, their optimal mode and rate distortion cost (RDC) can be known.

In the present invention, the following equation 2 is used to determine whether the optimal mode is mode 0 or mode 1.

Equation 2 does not apply to the boundary portion of the frame. 3 illustrates an area of a frame to which Equation 2 is applied. Equation 2 is not applied to the boundary part because neighboring blocks used in Equation 2 do not exist intact.

Equation 2 may be available if a block having an optimal mode of mode 0 or mode 1 among four adjacent blocks of the current block.

First, an average value of the RDCs of the blocks in which the optimal block is Mode 0 (MO) or Mode 1 (M1) is calculated. This average value is represented by RDC _L in Equation 2. If the optimal mode is other than mode 0 and mode 1, their RDC average value is calculated. This average value was represented by RDC _H in equation (2). RDC ₀ and RDC ₁ are RDC values of mode 0 and mode 1 of the current block, respectively. Where RDC _{L and} RDC _H are calculated and The order of calculation of RDC ₀ and RDC ₁ can be reversed.

Three comparisons are performed by substituting the calculated values into Equation 2. That is, if the RDC is not _L is smaller than the RDC _{H, 0} RDC RDC is less than _L, the RDC ₁ compares whether or not _L is smaller than the RDC.

The purpose of the above comparison is to determine whether there are mode 0 and mode 1 that satisfy all three comparisons. If all three comparison conditions are met, we conclude that the optimal mode of the current block is mode 0 or mode 1, and finally, after determining that the optimal mode is the smaller of RDC value between mode 0 and mode 1, Terminate the mode decision process.

If the condition of Equation 2 is not satisfied, the second filter is used to provide an opportunity to finish the mode decision process early again.

In Equation 2 above, we used a threshold value for mode determination. This value is not a predetermined value, but a variable value generated using neighboring blocks for each block, and can be used as a fairly accurate threshold value. .

Table 4 below shows the results after applying Equation 2.

In the above table, the hit ratio (HR) and the accumulation ratio (AR) are defined as Equations 3 and 4 below.

The HR of Equation 3 represents the ratio of the total number of modes 0 and 1 to the optimal mode and the number of modes 0 and 1 correctly determined through Equation 2. Referring to Table 4, the HR of a sequence having a lot of motion is less than 20%, and the HR of a sequence having less information on the motion is 40% or more.

Equation 4 shows the ratio of how accurately the mode 0 and the mode 1 were determined when the optimum mode was determined. In other words, when mode 0 or mode 1 is determined to be the optimal mode, it is the ratio that indicates whether the optimal mode is incorrectly determined as mode 0 or mode 1 when the other mode is the optimal mode.

Looking at Table 4, it can be seen that the value of AR is close to about 90% on average. As a result, it can be seen that the initial determination method of Mode 0 and Mode 1 using Equation 2 was performed with considerable accuracy.

The present invention makes it possible to determine the optimum mode early by simply using information of neighboring blocks of the current block. The threshold value used herein is generated by using a corresponding neighboring block for each block, and can be changed according to the characteristics of the block, thereby further increasing the accuracy of mode determination of each block.

mode  Using correlation second  filter

As can be seen from Table 4 above, the ratio of HR after filtering Mode 0 and Mode 1 using the above-described dynamic threshold is about 40% on average. This shows that we still have the possibility of finding Mode 0 or Mode 1 at approximately 60% of the total. The present invention proposes a second filter based on this fact.

The first step in creating a second filter is to find the correlation between the modes. In particular, since the main object of the present invention is to determine the mode 0 and mode 1 early and to end the mode determination step, the experiment was to find out which mode is most correlated with the mode 0 and mode 1.

Experiments were conducted for Mode 2, Mode 3, and Mode 8, and data on the correlation between Mode 0 and Mode 1 RDC values were collected for each of the three modes.

Table 5 is a table showing the experimental results for the correlation between each mode.

Through the above experiments, we want to determine the best comparison mode to use as the comparison object for Mode 0 and Mode 1. Equations 3 and 4 above were used as it is to calculate HR and AR.

Referring to Table 5, it is easy to see that Mode 8 is the best in terms of HR only. However, in terms of accuracy (AR), it can be seen that Mode 2 or Mode 3 is better than Mode 8.

In other words, mode 8 is better in case of 'Stefan', 'Mobile', and 'Tempete' which has a lot of motion information, but mode 8 is less excellent in case of 'Akiyo' and 'Container' which is small in motion information. Can be.

Mode 2 and Mode 3 show considerable similarity in terms of HR and AR, and it can be seen that Mode 3 is slightly better than Mode 2. Considering the above experimental results, it would be better to select Mode 8 as the most suitable comparison target, but for the stability of performance, one more experiment was conducted.

Table 6 below shows the results of calculating the performance and temporal gain of the encoder when Mode 2, Mode 3, and Mode 8 are used as comparison targets for Mode 0 and Mode 1, respectively.

From the above table, it is better to select Mode 8 in terms of the performance of the encoder, but considering the temporal gain, it can be seen that almost no gain can be obtained when Mode 8 is compared. Note that in time gain, Mode 2 or Mode 3 is three times better than Mode 8.

On the other hand, considering the performance of the encoder, mode 8 is excellent, but it can be seen that mode 2 or mode 3 does not differ significantly. Considering this, it can be seen that it is reasonable to determine mode 3 as a comparison target for use in the fast mode decision algorithm.

The above formula of the second filter is as follows.

If {(RDC ₀ <RDC ₃ ) & (RDC ₁ <RDC ₃ )}

Then best mode is MO or M1 and early terminated

Otherwise calculate for all modes

Here, RDC ₀ , RDC ₁ , and RDC ₃ represent RDCs corresponding to mode 0, mode 1, and mode 3, respectively. If the above equation is satisfied, the mode corresponding to the smaller value is determined by comparing the two values, and otherwise, the optimal mode is determined by calculating the RDC for all modes.

The second algorithm used for the initial determination of mode 0 and mode 1 is relatively simple, but very accurately it is possible to find out which block is mode 0 or mode 1 as the optimal block.

4 is a flowchart summarizing the above-described inter mode determination method according to the present invention.

In FIG. 4, the present invention is described as being applied when encoding P-slice after encoding an I-frame. Referring to FIG. 4, it is first determined whether a current block corresponds to a boundary block of a frame (S10). If the frame is a boundary block, the optimal mode is determined by calculating RDC values for all modes as in the conventional method (S19).

On the other hand, if the current block is not a boundary block of the frame, first, an RDC (ie, RDC ₀ and RDC ₁ ) for mode 0 and mode 1 is calculated (S11), and the RDC _L and the RDC values of the neighboring blocks of FIG. RDC _H is calculated (S12).

When the calculated values are substituted into Equation 2 to satisfy Equation 2 (S13), the RDC (ie, RDC ₀ ) for mode ₀ and the RDC (ie, RDC ₁ ) for mode 1 are compared (S14). If RDC ₀ is small, the optimal mode of the current block is determined as mode 0. If RDC ₁ is small, mode 1 is determined.

On the other hand, in step S13, if the optimal mode of the neighboring blocks do not exist mode 0 or mode 1, or does not satisfy the equation (2), filter the mode 0 and mode 1 through the second filter.

In detail, an RDC (ie, RDC ₃ ) for mode 3 is calculated (S17), and the corresponding values are substituted into Equation 5 to determine whether Equation 5 is satisfied (S18).

That is, if the RDC values of the mode 0 and the mode 1 are smaller than the RDC values of the mode 3, the operation S14 to S16 is performed to determine the mode 0 or the mode 1 as the optimal mode.

On the other hand, if the equation 5 is not satisfied, the optimal mode of the current block is determined by calculating RDCs for all modes (S19).

As such, by applying the two initial mode determination methods, the present invention can considerably reduce the time required for compression while the encoding performance is hardly degraded. The present invention can be very useful in applications requiring fast encoding.

Experiment result

In order to verify the performance of the technical content proposed in the present invention, several experiments were conducted using JM 12.4 reference software.

Table 7 shows the experimental conditions of the codec, and Table 8 shows the result of comparing the present invention with Zhan's algorithm. Here, three things such as PSNR (ΔPSNR), bit rate difference (ΔBR), and time reduction (ΔTime) were compared.

Here, the parameter for time reduction is defined as follows.

Where T _proposed And T _{JM 12.4} mean the encoding time of the proposed algorithm and JM 12.4 reference software, respectively. To obtain the simulation results, seven test sequences were used.

For fair experiments, the proportions of the motion information referenced different sequences. The 'Stefan', 'Foreman', and 'Table-tennis' sequences have a lot of motion, but the 'Akiyo', 'Container' and 'Silent' sequences have relatively little motion.

Referring to Table 8, in terms of time reduction, the present invention can obtain an average of 42.5% of gain, while showing a negligible difference in terms of PSNR. The average of the difference in PSNR between the present invention and the reference software is approximately -0.03.

However, the mean PSNR reduction mean of the chance algorithm is larger as -1.15.

The increase in bit rate only occurred in the 'Akiyo' sequence. The mean value of ΔBR is significantly reduced compared to JM software. Although the ΔBR of the chance algorithm is also reduced, it can be seen that the ΔPSNR has dropped considerably.

From the above experimental results, it can be seen that the present invention has much better performance than the conventional chance algorithm.

5 to 7 are graphs showing the comprehensive results of the performance compared through the above experiment.

The average reduction rate of the PSNR of the present invention is 0.03% and the average reduction rate of the bit rate is 0.79%, which is a result confirming that the present invention shows a fairly good performance.

While the present invention can achieve an average of 42.5% gain in terms of time reduction, nearly similar results can be obtained in terms of PSNR.

The present invention described above may be implemented by a software algorithm, and may also be implemented by a device including a memory in which the algorithm is stored, a processor on which the algorithm is executed, and the like. It is also applicable to all displays which process and display an image.

In particular, the present invention can be applied to all applications using the video compression technology, for example, a mobile phone supporting a mobile broadcast and many DMB players, cameras, HDTV, IPTV, PC, such as many devices that require a video compression technology.

Although some embodiments of the invention have been shown and described, it will be apparent to those skilled in the art that modifications may be made to the embodiment without departing from the spirit or spirit of the invention. . It is intended that the scope of the invention be defined by the claims appended hereto and their equivalents.

1 shows a variable block size of the H.264 / AVC standard.

2 illustrates peripheral spatial information of the current block used in the present invention.

3 illustrates an area of a frame to which Equation 2 of the present invention is applied.

4 is a flowchart summarizing an inter mode determination method according to an embodiment of the present invention.

5 to 7 are graphs showing the results of comparing the performance with the conventional algorithm through an experiment according to an embodiment of the present invention.

Claims (9)

In the adaptive fast mode decision method using the dynamic threshold, (a) calculating a rate distortion cost RDC ₀ for a skip mode (SKIP MODE) of the current block for encoding and a rate distortion cost RDC ₁ for a 16 × 16 block size mode; calculating a mean value RDC _L of a rate distortion cost of a neighboring block having a skip mode or a 16 × 16 block size mode as an optimal mode among neighboring blocks adjacent to the current block that is already encoded; (c) calculating an average value (RDC _H ) of a rate distortion cost of a neighboring block having a mode other than a skip mode and a 16x16 block size mode among the neighboring blocks adjacent to the current block that are already encoded as an optimal mode; (d) determining whether to terminate the mode decision early by comparing the RDC _L with the RDC _H , comparing the RDC ₀ with the RDC _L , and comparing the RDC ₁ with the RDC _L. Adaptive fast mode decision method using dynamic threshold. The method of claim 1, In the step (d), when the RDC _L is smaller than the RDC _H , the RDC ₀ is smaller than the RDC _L , and the RDC ₁ is smaller than the RDC _L , the RDC ₀ is smaller than the RDC ₁ . And determining a mode having an optimal mode as the optimal mode of the current block and terminating the mode decision early. The method according to claim 1 or 2, In step (d), if the RDC _L is greater than the RDC _H , the RDC ₀ is greater than the RDC _L , or the RDC ₁ is greater than the RDC _L , the rate distortion cost for an 8x16 block size mode Calculating (RDC ₃ ) and comparing the RDC ₀ , the RDC ₁ , and the RDC ₃ to re-determine whether to terminate the mode decision early. How to decide. The method of claim 3, Re-determining whether or not to terminate the mode decision early, When the RDC ₀ is smaller than the RDC ₃ and the RDC ₁ is smaller than the RDC ₃ , the mode having a smaller value is determined by comparing the RDC ₀ with the RDC ₁ to determine the optimal mode of the current block and to early the mode decision. Adaptive fast mode determination method using a dynamic threshold characterized in that it comprises the step of terminating. 5. The method of claim 4, Re-determining whether or not to terminate the mode decision early, When the RDC ₀ is larger than the RDC ₃ or the RDC ₁ is larger than the RDC ₃ , the rate distortion cost for all remaining modes is calculated to determine a mode having a minimum value as an optimal mode. Adaptive Fast Mode Decision Method. In the adaptive fast mode decision method using the dynamic threshold, (a) calculating a rate distortion cost RDC ₀ for a skip mode (SKIP MODE) of the current block for encoding and a rate distortion cost RDC ₁ for a 16 × 16 block size mode; (b) the skip mode or the optimal mode of the current block based on the optimal mode of the neighboring blocks adjacent to the current block and the rate distortion cost value of the optimal mode of the neighboring block and the RDC ₀ and RDC _1; Primarily determining whether to determine early in any of the 16x16 blocksize modes; And (c) If the optimal mode of the current block is not determined early in the step (b), the rate distortion cost RDC ₃ for the 8x16 block size mode of the current block is calculated and the RDC ₀ and the Adaptively determining whether to optimally determine the optimal mode of the current block as the skip mode or the 16 × 16 block size mode through comparison with RDC ₁ ; How to determine fast mode. The method of claim 6, The step (b) may include: calculating an average value RDC _L of a rate distortion cost of a neighboring block having a skip mode or a 16 × 16 block size mode as an optimal mode among neighboring blocks adjacent to the current block that is already encoded; Calculating an average value (RDC _H ) of a rate distortion cost of a neighboring block having a mode other than a skip mode and a 16 × 16 block size mode among the neighboring blocks adjacent to the current block that are already encoded as an optimal mode; And When the RDC _L is smaller than the RDC _H , the RDC ₀ is smaller than the RDC _L , and the RDC ₁ is smaller than the RDC _L , a mode having a smaller value is compared with the RDC ₀ and the RDC ₁ . Adaptive mode decision method using a dynamic threshold characterized in that it comprises the step of determining the optimum mode of the early mode termination. The method of claim 6, In the step (c), when the RDC ₀ is smaller than the RDC ₃ and the RDC ₁ is smaller than the RDC ₃ , the mode having a smaller value is compared to the optimal mode of the current block by comparing the RDC ₀ with the RDC ₁ . Adaptive fast mode determination method using a dynamic threshold characterized in that the decision and early termination of the mode decision. The method of claim 6, Adaptive fast mode determination method using the dynamic threshold, And determining whether the current block is a boundary block positioned at upper, lower, left, and right boundaries of a frame. If the current block is the boundary block, the optimal mode is determined by calculating a rate distortion cost for all modes, and if the current block is not the boundary block, steps (a) to (c) are performed. Adaptive fast mode determination method using a dynamic threshold.

Images