AU2021103378A4 - A self-adaptive n-depth context tree weighting method - Google Patents

Abstract

The present disclosure relates to a self-Adaptive N-depth Context Tree Weighting method (SANDCTW) to overcome the limitations of Context tree weighting (CTW) method, which applied in Context Adaptive Binary Arithmetic Coding(CABAC). CABAC uses KT estimators and relies on beginning with Bayesian approach to determine true distribution of the next symbol to select for data compression. This approach is suitable only if true distribution is stationary, the proposed SANDCTW uses discounted KT estimators, which is suitable if distribution is non-stationary and it reduces computation and memory cost. Additionally, Block size sustained intra mode detection (BSSIMD) is proposed based on mass-center and sub-sampling approach. In this approach, all correlation directions about entire block is associated to the intra-prediction and DC mode directions determined by using mass-center vector. Then, modes corresponding to the determined directions selected as best intra-prediction candidates during intra-coding process for computing Rate-Distortion Optimization (RDO) with less complexity. 31 100 appyiri a self-ada pte N-depithcontexttr eemghng rgteqej5AN r nWmontet ada ptebiry ar tic 102 codirg|CABAC to oercome theflimrtatiraofcontext treoxeghitirg(tTWajmethd -104 prqxs UicirimsustairediramodedetectiolESSIMDonmmass-centerandsub-sampfm appradAi 106 asscfatirgallcoelat directicrrat cFt entebcktothe tra-predctionmodearddeemini DCmode dictis b ig mass-centevector selectirgmodescorrespondirgtothedetermineddiectis astestintra pedctincandidatesdurirgtheitra- 108 codingpricessforcomputirgrate-ditortionoptmiaton|RDC withesscomplexity Figure1 Fi xd Bins: 1-14 Regular bins Bypass bins Tninrnted unary ck EGO iode a bs leave I (bins) Figure 2 L r L- I mo w I I HaT L 4i -- , p T- ft&id t-)EI Mafs"dn F rud Figure 3 Fig ure 4

Description

appyiri a self-ada pte N-depithcontexttr eemghng rgteqej5AN r nWmontet ada ptebiry ar tic 102 codirg|CABAC to oercome theflimrtatiraofcontext treoxeghitirg(tTWajmethd

-104 prqxs UicirimsustairediramodedetectiolESSIMDonmmass-centerandsub-sampfm appradAi

106 asscfatirgallcoelat directicrrat cFt entebcktothetra-predctionmodearddeemini DCmode dictisb ig mass-centevector

selectirgmodescorrespondirgtothedetermineddiectis astestintra pedctincandidatesdurirgtheitra- 108 codingpricessforcomputirgrate-ditortionoptmiaton|RDC withesscomplexity

Figure1

Fi xd Bins: 1-14

Regular bins Bypass bins

Tninrnted unary ck EGO iode a bs leave I (bins)

Figure 2

L r

L- I HaT I mo w I

L 4i -- , p T- ft&id t-)EI Mafs"dn

F rud

Figure3 Fig ure 4

A SELF-ADAPTIVE N-DEPTH CONTEXT TREE WEIGHTING METHOD FIELD OF THE INVENTION

The present disclosure relates to a self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimator. BACKGROUND OF THE INVENTION

In modern years, video communication plays a major role because of the expansion of digital video applications and video compression standards. Among completely different standards, the trendy video compression standards are (H.265/HEVC) High Efficiency Video coding. When correlated with Advanced Video Coding- (H.264/AVC), HEVC offers double the data com- pression quantitative relation at the equal level of video quality or bit rate. H.265/HEVC- utilizes the hybrid video encoding techniques for deciding, however, the frame is partitioning on the blocks and the way the pixels of individual blocks are to be predicted. Initially, a group of non-overlapping pixel blocks is utilized for dividing each edge of a video succession into littler areas by applying intra and inter prediction methods. Intra prediction extrapolates the pixel esteems within the block by the contiguous pixel estimations of neighbour pieces. Inter expectation uses pixel esteems from some district of the prior coded video outline. Then, the residual signal is getting by subtracting the predicted pixel values, and conjointly the obtained leftover flag mapped into the recurrence area upheld the two-dimensional discrete Fourier change. Then, the trans- formed signal quantized that reduced the information on redundancy. Thus, the changed and quantized residue and supplemental information on compression techniques utilized for entropy coding.

The fast mode detection technique primarily incorporated with every micro and macro-level mechanisms in HEVC encoder. The Mode Search has pro- posed at the micro-level for by selection checking the potential modes. Some effective candidates are chosen for succeeding Rate Distortion Optimized quantization (RDOQ) to figure the Rate-Distortion (RD) optimal mode. Also, an early RDOQ skips schemes introduced for reducing the additional complexity. Moreover, the primary CU (Coding Unit) split termination were introduced at the macro-level if the estimated RD cost was larger than this CU RD costs. However, it desires additional improvement for achieving higher complexness reduction and coding efficiency trade-off. Fast bit estimation methodology in RD was proposed for intra coding units in HEVC. During this approach, a technique has proposed for evaluating the mode bits per intra bearing indicator all through unpleasant mode choice advance. The proposed methodology supported the regression for estimating the context adaptive binary bits for direction mode and TU in view of the greatness of non-zero coefficients. Moreover, the data of quantized TU coefficients and quantization parameters have utilized. However the coding productivity lower than the whole RD system. Computationally-scalable technique and its hardware design have presented that have the flexibility for supporting the intra encoding. This proposed encoder has used for permitting the exchange between the compression efficiency and through- put. The pre choice in view of the forecasts produced from the examples was performed with a proportionate asset as the normal handling. Moreover, the hardware cost has reduced and better throughputs as obtained. However, the complexness of the proposed design was high. Adaptive transform size decision technique have proposed for HEVC inter coding. During this approach, three methods are incorporated like change side step in view of the skip mode recognition, content-based change estimate choice, and early end of the variable- sized change. Initially, the coding data from close-by inter and intra CU or Tree pieces and the parent cu in the upper profundity levels has checked. Then, just few change sizes chose in the change estimate choice strategy in light of the coding data from adjacent CU. Quick TU estimate choice calculation was furthermore proposed utilizing Bayesian theorem detection. In this approach, the TU size was selected based on the utilization of Bayesian choice hypothesis and the connection between the fluctuation of lingering coefficients, However, the time consumption of coding is high compared with the other strategies.

High throughput CABAC entropy coding have proposed in H.265. In this approach, the key techniques have addressed which used for enabling HEVC to achieve higher throughput potentially. Such techniques were consisting of minimizing parsing dependencies and gathering bins with all possibility's ways Moreover, CABAC needs more memory that also reduced significantly. However, the complexity of the technique was high. Zhu, Liu, Wang, Han, and Song (2013) subsequently by proposing with Hadamard change based RD cost appraisal for HEVC intra coding. In this approach, the low complexity RD estimation techniques were recommended based on Hadamard trans- form. Such techniques have composed of two characteristics. Initially, Hadamard transforms as applied for PU prediction mode decision process. At that point, the bias estimation has rearranged by evacuating the recreation activities. This approach has used for improving the pro- cessing speed and reducing the forward transform complexity. However, the computational unpredictability of the calculation is high. Improved adaptive arithmetic coding has been proposed in AVC compression standard. The adaptive arithmetic encoder has proposed which is dependent on the improvement of CABAC strategy with context pattern of larger size based on CTW scheme which has used in MPEG H.264/AVC. The major objective of the approach is presenting the possibilities of further improvement in CABAC coding efficiency by utilizing the more complex data statistics modelling. However, the performance of encoder was analyzed only for MPEG-AVC video compression standard.

Probability estimation with high prediction has proposed for CABAC. This technique has proposed based on the multiple estimators by using different models. This approach efficiently realised in integer arithmetic. The major objective of this approach has to utilise the various estimations and a weighted average of such estimations for increasing the prediction probability. For achieving high prediction probability, exponential smoothing and the auto-correlated signal used. However, the computation complexity of the approach was high. In this approach, a VLSI based CABAC entropy codec presented for HEVC, and its performance has analyzed regarding effective processing throughput. However, the architecture designed for processing qualified two formal bins per unit time for improving high throughput when pipeline stall have minimised as the build in dependencies of CABAC by using different optimizations along with con-text forwarding and speculative decoding. However, this approach requires more accurate context prediction according to the intra and inter-prediction.

Gao et al. have proposed for HEVC. During this approach, CABAC has proposed that comprises of the enhanced setting displaying for change coactive levels and a Binary Arithmetic coding (BAC) motor (engine) with little memory request. Here, the ambience model record as significance map relied on and registered in light of the measure of the huge neigh- bours secured by the nearby layout and its situation inside TB. The aggregate number of context models was limited by part TB into various districts as indicated by the coefficient position. Within the BAC engine with low memory request, the likelihood assessed in light of the multi- parameter probability update mechanism. Also, an increase with low piece limits used in the interim coding sub-division for substituting the substantial look-into table for lessening its memory utilisation. However, the computation complexness of the approach is high. The dynamic entropy coding strategy used in the video encoders is CABAC. This CABAC for HEVC enhanced by utilizing more exact plan of contingent likelihood estimation.

The improvement of CABAC achieved by using CTW and the anticipate with the limited matching method. This CTW method performed by using KT estimators for approximating the current probability distribution of symbols. This approximation is most suitable if the true distribution is stationary and also the problem of KT estimator is that it operates at very slow for updating process while many samples have collected. Therefore, the modificationin distribution can't be learned quickly by KT estimator. Also, the computation complexity of schemes in the encoding process is high. Additionally, the required number of memory cradle reservation is more than H.264, so the expansion in memory requests isn't satisfactory for real-time processing. HEVC utilizes the angular prediction modes for PU which causes high computation complexity than the directional modes. On the other hand, full RDO techniques not used due to its complexity and high com- pression time. These are the problems self-addressed in H.265/HEVC based on the different block sizes such as 16 x 16, 32 x 32, or 64 x 64. Hence in this research, Self- Adaptive N-Depth Context Tree Weighting (SANDCTW)Technique and Block Size Sustained Intra Mode Detection (HEVC- BSSIMD) is proposed to enhance the HEVC standards.

In order to overcome the aforementioned drawbacks, there exists a need to develop a self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimator.

SUMMARY OF THE INVENTION

The present disclosure seeks to providea self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimator for non-stationary distributions.

In an embodiment,a self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimatoris disclosed. The method comprises:

applying a self-adaptive N-depth context tree weighting technique (SANDCTW) in context adaptive binary arithmetic coding (CABAC) to overcome the limitations of context tree weighting (CTW) method; proposing block size sustained intra mode detection (BSSIMD) on mass-center and sub-sampling approach; associating all correlation directions about entire block to the intra prediction mode and determining DC mode directions by using mass center vector; and selecting modes corresponding to the determined directions as best intra-prediction candidates during the intra-coding process for computing rate-distortion optimization (RDO) with less complexity.

In an embodiment, block mass-center direction is perpendicular to the block correlation direction, wherein such directions are horizontal-up, horizontal-down, vertical left, and vertical-right.

In an embodiment, the correlation direction of these blocks determined with mass-center direction by forming their corresponding symmetric sub-block.

In an embodiment, proposed two sub-sampling techniques are impair lines and half columns (ILHC) and impair columns and half lines (ICHL).

In an embodiment, ILHC is applied to the vertical-left, and vertical right directional correlation blocks and ICHL applied to the horizontal-up and horizontal-down correlation direction of blocks.

In an embodiment, ILHC sub-sampling method forms the square sub-blocks by subsampling the block pixels with two major steps such as the impair lines of the blocks selected, and then the half columns in the middle of the block are selected.

In an embodiment, eight correlation directions of the blocks are determined if the vector direction of the sub-block formed by the subsampling techniques is determined.

In an embodiment, the block size sustained intra-prediction in H.265/HEVC is proposed in the view of mass-center and two subsampling techniques.

An objective of the present disclosure is to determine the true distribution of the next symbol to select for data compression.

Another object of the present disclosure is to determine all possible correlation directions of the blocks corresponding to the intra-prediction mode.

Yet another object of the present invention is to deliver an expeditious and cost-effective method for replacing conventional CTW using disconnected KT estimator.

To further clarify advantages and features of the present disclosure, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.

BRIEF DESCRIPTION OF FIGURES

These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:

Figure 1 illustrates a flow chart of a self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimatorin accordance with an embodiment of the present disclosure; Figure 2 illustrates a binarization scheme for SANDCTW -CABAC in accordance with an embodiment of the present disclosure; Figure 3 illustrates an ILHC subsampling method in accordance with an embodiment of the present disclosure; Figure 4 illustrates a block diagram of SANDCTW-CABAC-BSSIMD in accordance with an embodiment of the present disclosure; Figure 5 illustrates comparison and experimental result of PSNR (dB) for four video sequences in accordance with an embodiment of the present disclosure; Figure 6 illustrates comparison of SAE for four video sequences in accordance with an embodiment of the present disclosure; Figure 7 illustrates comparison and experimental result of Compression Ratio (%) for four video sequences in accordance with an embodiment of the present disclosure; Figure 8 illustrates Table 1 depicts performance comparison for existing techniques in accordance with an embodiment of the present disclosure; and Figure 9 illustrates Table 2 depicts performance comparison for proposed technique in accordance with an embodiment of the present disclosure.

Further, skilled artisans will appreciate that elements in the drawings are illustrated for simplicity and may not have been necessarily been drawn to scale. For example, the flow charts illustrate the method in terms of the most prominent steps involved to help to improve understanding of aspects of the present disclosure. Furthermore, in terms of the construction of the device, one or more components of the device may have been represented in the drawings by conventional symbols, and the drawings may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the drawings with details that will be readily apparent to those of ordinary skill in the art having benefit of the description herein.

DETAILED DESCRIPTION

For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates.

It will be understood by those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not intended to be restrictive thereof.

Reference throughout this specification to "an aspect", "another aspect" or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, appearances of the phrase "in an embodiment", "in another embodiment" and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment.

The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process or method that comprises a list of steps does not include only those steps but may include other steps not expressly listed or inherent to such process or method. Similarly, one or more devices or sub-systems or elements or structures or components proceeded by "comprises...a" does not, without more constraints, preclude the existence of other devices or other sub-systems or other elements or other structures or other components or additional devices or additional sub-systems or additional elements or additional structures or additional components.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The system, methods, and examples provided herein are illustrative only and not intended to be limiting.

Embodiments of the present disclosure will be described below in detail with reference to the accompanying drawings.

Referring to Figure 1, a flow chart of a self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimatoris illustrated in accordance with an embodiment of the present disclosure.A self-adaptive N-depth context tree weighting technique to overcome the limitations of context tree weighting (CTW) method, which applied in context adaptive binary arithmetic coding (CABAC). At step 102, the method 100 includes applying a self-adaptive N-depth context tree weighting technique (SANDCTW) in context adaptive binary arithmetic coding (CABAC) to overcome the limitations of context tree weighting (CTW) method.

At step 104, the method 100 includes proposing block size sustained intra mode detection (BSSIMD) on mass-center and sub sampling approach.

At step 104, the method 100 includes associating all correlation directions about entire block to the intra-prediction mode and determining DC mode directions by using mass-center vector.

At step 106, the method 100 includes selecting modes corresponding to the determined directions as best intra-prediction candidates during the intra-coding process for computing rate-distortion optimization (RDO) with less complexity.

In an embodiment, block mass-center direction is perpendicular to the block correlation direction, wherein such directions are horizontal-up, horizontal-down, vertical left, and vertical-right.

In an embodiment, the correlation direction of these blocks determined with mass-center direction by forming their corresponding symmetric sub-block.

In an embodiment, proposed two sub-sampling techniques are impair lines and half columns (ILHC) and impair columns and half lines (ICHL).

In an embodiment, ILHC is applied to the vertical-left, and vertical right directional correlation blocks and ICHL applied to the horizontal-up and horizontal-down correlation direction of blocks.

In an embodiment, ILHC sub-sampling method forms the square sub-blocks by subsampling the block pixels with two major steps such as the impair lines of the blocks selected, and then the half columns in the middle of the block are selected.

In an embodiment, eight correlation directions of the blocks are determined if the vector direction of the sub-block formed by the subsampling techniques is determined.

In an embodiment, the block size sustained intra-prediction in H.265/HEVC is proposed in the view of mass-center and two subsampling techniques.

Figure 2 illustrates a binarization scheme for SANDCTW -CABAC in accordance with an embodiment of the present disclosure. The proposed Self Adaptive N-depth CTW introduced for solving the constraints of CTW which uses KT estimators to comparative the current distribution. In CT approach, the KT estimator considered with the moving window. The probability of the subsequent bit is estimated using KT estimator which uses only the past k bits. Consider an array produced byBern(0),O 0 1. There is no pleasant excess headed for all cases, and the normal redundancy (R) for one bit is given as,

a k-a 0 log k±1 +(ok±l R (k; 0) = = kk )"(r) + a - -H[8] (1) (aL ka 2 2]il

In Eq. (1) r = (1 - )and H() denotes the entropy. But,the limitation of this approach is that a history of lengthkshouldbe kept. Hence in the proposed method, discountedversion of KT estimator is applied which overcomes theissue of storing the history of lengthk based on the windowing method. The proposed Self Adaptive N depth CTW is incorporated with CABAC context modeller and, therefore the binarization of proposed CABAC context encoder relies about the integration of the abbreviate single code and Othorder Exp-Golomb (EGO) code. In a binarizationword for the total estimation of the change coefficient leveli.e., abs level, the quantity of paired images coded withbinarization plot is displayed in Figure 2. Here, a better variety of bins are coded with an applied math modelling of transform data. Consider the discount ratey E [0,1), anda binary string yi:t. The concession rate at and bibt however different through the reduced number of zeros and ones iny:t are reserved.

Then, the next symbol observed as one of the result at and bt is maximised based on the recognized token and KT assessment is computed in view of the at and be. At that point the concession amount is used for updating the counts at and bt by,

at 1 = (1 - y)at, bt±i = (1 - y)bt (2)

The estimation of y need isn't a consistent. For elect thevalue of y, the consecutive number of possibilities isconsidered:

Full-context N Depth Visit-based (HEVC- SANDCTW-vl): By using this mechanism, the concession estimates at the leaf hub(node) respective through the present context computed. An observed sequence is yi:t where the array of bits in y:t is [y1:t]nwhich wind up in the leaf node n. Assume ko is the length of [y1:tlin, then the discount rate is computed as,

yn = ck-X (3)

For each hub(node) n' on the direction from n to the root hub(node) A,the concession rate is as follows:

yn, = ck-x (4)

Based on the Eqs. (3) and (4), the same concession rate used for all nodes in the path. Thus, the concession estimates turn into an element of the number of perceptions of the present length D context for a weighted context tree ofdepth D.

Leaf-Context N Depth Visit-based (HEVCSANDCTW-v2): This is like to the full-context visitbasedadaptive approach. However, it uses a flexible approach for updating the counts an and b. The concession rate is as similar as Eq. (3) and the counts an, and bo, are updated for hub(node) n' on the direction from to the root hub(node) by using the following expressions:

an, = an,, + an,,, bn, = bn, 1 + bn,, (5)

In Eq. (5), the left side and right-side children of n' are denoted as n', and nOr respectively. By utilising these methodologies, the concession rate does not affect the KT estimator tally commitments of some other profundity D context. This approach spares the property of CTW where the counts an and bn of a node(hub) n equal to the total of the means its youngster hubs (childnodes). Also, Block size sustained intra mode detection Method (BSSIMD) is introduced for HEVC intracodingprocess.

SANDCTW-CABAC technique:

Input: Binary tree of depth D

Output: Bit stream

1. Initialize 2.{ 3. Assign the node called n in the binary tree. 4. Consider yi:t is the binary string consists of a zeroes and b ones. 5. For ([y1:tn)

6.{

7. Discount KT estimator with eloquent window. Utilize only the last k bits. 8. Denote the discount rate y E[0,1). 9. If (HEVC-SANDCTW-v1 is used) 10. { 11. Compute discount (concession) rate using yn = ck-x. 12. For each and individual node n' on the path from n to the root hub(node) A, yn, = ckx 13. } 14. Else 15. { 16. Compute concession rate using yn = ck[x. 17. Update the counts an, and bo, for each and individual hub(node) n' on the path from n to the root hub(node) A as,

an, = an,, + anr

bn, = bn,, + bn,

18. } 19. Compute the discounted counts at and bt. Store the discounted counts at each node. Calculate the KT estimate as follows:

a1 Pkt(a + 1,b) b 2 Pkt(a,b) a+

b +1 2 Pkt ta, b + 1) = Pkt (a, b) a+

Pkt(0,0) = 1

20. Increment at and be as, at+1 (1 - y)at bt+ 1 (1 - y)bt

21. } 22. Encode the computed probability of a symbol using M encoder. 23. Obtain the output bit stream. 24. }

Figure 3 illustrates an ILHC subsampling method in accordance with an embodiment of the present disclosure. A Block Size Sustained intra prediction in H.265/HEVCis proposed in the view of mass-center and two subsampling techniques. This approach can able to determine all correlation bearings of the piece (block) that corresponds to the intra prediction mode directions of the HEVC. In this approach, the block mass center directionis perpendicular to the block correlation direction. Such directions are horizontal-up, horizontal-down, verticalleft,and vertical right. The correlation direction of these blocks determined with mass center direction by forming their corresponding symmetric sub-block. The proposed two sub-sampling techniques are Impair Lines and Half Columns (ILHC) and Impair Columns and Half Lines (ICHL). ILHC applied to the vertical-left, and vertical-right directional correlation blocks and ICHL applied to the horizontal-up and horizontal-down correlation direction of blocks.

When the block has a correlation direction like horizontal, vertical, diagonal-right, and diagonal-left, this block has a symmetry axis through the center of the current block. The direction of this axis is perpendicular to the direction of the homogenous pixels of the block. The ILHC sub sampling method forms the square sub-blocks by subsampling the block pixels with two major steps such as the impair lines of the blocks selected, and then the half columns in the middle of the block are selected. The ICHL sub-sampling forms the square sub-blocks by subsampling the block pixels by using two steps that are, the impair columns of the blocks selected, and then the halflines in the middle of the block are selected. Therefore, eachsub-block of these sub-blocks have a symmetry axis through the center of the current sub-block. Thus, the eight correlation directions of the blocks are determined if the vector direction of the sub-block formed by the subsampling techniques is determined. In the proposed model, the directions determined by the mass-centre method.

The symmetry axis direction of the blocks represented by the mass center vector direction which is verified by the following steps:

Vertical Direction Correlation Block: Consider an orthonormal coordinate system 0,7,j), 0,,7,j, the root is selected to be block's center i is the parallel direction and jj is the perpendicular direction to the bottom, and each block N X N, N = 2m + where m ={2, 4, 8} refers an impair value and the possible correlation directions of this block are vertical, parallel, corner to corner right and diagonal-left direction, then the mass center vector direction is parallel to the symmetry axis direction of the block. Consider the line through the origin 0 is L, and parallel to the vectorV = (x,), where x denotes an integer number except O.This line is the symmetry axis of the vertical direction correlation block. Therefore, all pixels in this block that are symmetry with this line L, have the equivalent intensity and defined as follows:

Iy = I_, I-m ! x,y !m (6)

The coordinates of the mass-center of this block denoted as follows:

G =Gx S =IX,y I YI Y, x x= N X Ix,O x x x=-my=-m x=-m m m

Gy = Y YI, x y y=-m x=-m

=_ Sym Y=-m /x,= y + Ei-nx -yx(y)] = 0 (7)

In Eq. (7), Ix,y refers the intensity (anxiety) of the pixel of location (x,y) of a block, (Gx;Gy) refers the coordinate of the mass-focus (center) of a block, SI is the total number of block pixels intensity values which is calculated as,

Si = ER-m Zym Ix,y (8)

Hence, the mass-center vector of this block is denoted asG=(k,O), where k refers the constant value. By using these computations, the direction of the mass center vectoris observed that it is parallel to the symmetry axis direction of this block. The two 5x5 blocks and their corresponding square sub blocks formed by ILHC method is shown inFigure3. From this figure it is observed that the sub-block formed by the ILHC subsampling technique for the block which is having vertical-left and vertical-right directional homogeneous pixels has a diagonal-left and diagonalrightdirectional homogeneous pixels. Therefore, eachsub-block of these sub-blocks has a symmetry axis through the center of this subblock.

Horizontal Direction Correlation Block: Consider the line through the origin 0 is L and parallel to the vectorVLh = (O,y) , where y denotes an integer number except O.This line is the symmetry axis of the horizontal directioncorrelation block. Therefore, all pixels in this block that are symmetry with this line Lh have the equivalent intensity and defined as follows:

Iy = I_Xy' - m : xy m (9)

The coordinates of the mass -center of this block is denoted as follows:

Gy = ZY ENm M_Ix,y Exm-m x y = EmN oyxy (0_~ymN XIO,yX y (10)

By utilizing these computations, the mass-center vector of horizontal direction correlation block is denoted byG=(0,k), where j refers the constant value. Thus, the direction of the mass center vector G is observed that it is parallel to the symmetry axis direction of this block. Similar to Figure 3 it is observed that the sub-block formed by the ICHL subsampling technique for the block which is having horizontal-up and horizontal-down directional homogeneous pixels has a diagonal-left and diagonal-right directional homogeneous pixel. Therefore, each sub-block of these sub-blocks has a symmetry axis through the center of current sub-block.

Figure 4 illustrates a block diagram of SANDCTW-CABAC-BSSIMD in accordance with an embodiment of the present disclosure.

Figure 5 illustrates comparison and experimental result of PSNR (dB) for four video sequences in accordance with an embodiment of the present disclosure. Figure 5 shows the two 5x5 blocks and their corresponding square sub blocks formed by ICHL subsampling technique. From this figure, it is observed that the subblockformed by the ICHL subsampling technique for the block which is having horizontal-up and horizontal-down directional homogeneous pixels has a diagonal-left and diagonal-right directional homogeneous pixels. Therefore, each subblock of these subblocks has a symmetry axis through the center of this subblock.

Intra-prediction way. In this manner, the angle of the mass-focus (center) vector of the block utilized for deciding the intra-prediction mode of this block 4x4 and 16x16Luma Block Directional Correlation: The 4x4 luma blocks are more appropriate for predicting the pictures with important information. There are nine prediction modes such as one DC prediction mode and eight directional prediction modes. These directional prediction modes has further branched into three classes such as class 1, 2 and class3. Class 1 contains mode 0 (steep (vertical) prediction mode), mode 1 (parallel mode), mode 3 (corner to corner(diagonal) left mode), and mode 4 (corner to corner (diagonal)right mode). Class 2 includes mode 5 (steep (vertical)right mode), and mode 7 (steep (vertical) left mode)whereas class 3 contains mode 6 (parallel-down mode),and mode 8 (parallel-up mode). The corresponding mass centers of each mode of each class's modes prediction of4x4 blocks is determined by the Eq. (11).

Moreover, for 16x16 luma blocks, there are horizontal and vertical prediction modes incorporating with the plane and DC prediction mode. In this way, the mass-focus (center)bearing of this block and their relating forecast modes defined in Eq. (12).

Mode3/ 1 Ez] in,4 ~In] Mo p13x e4 + in, 51+ ilx Mode~fl T(,- T+ Modefl2 ]" + In, 1, + 1-x

Mode7fl2c In, e ln] Mode6fl ] + 1n, " +1n] Mode8f - + In, + ir] M~ode#C3~ F6 T6_jm 5 ~n (11) Mode l# 4 ] E11 -, 13 U- +, (12) 116 x 16 = (

Mo de0#f4 EC_,7 U ) + 7, + n)

In Eqs. (11) and (12), the angle f 1 is computed asfl1 = 01- 2' where 01 refers the angle of 4x4 block mass center. The angle #2 is calculated asp2 =02- 02 , where6 2refers the mass center angle computed for the sub blocks formed by the ICHL technique. The angle#3 is computedas/3 = 03

IT 2, where 03 refers the mass center angle whichis computed for the sub blocks formed by ILHC technique.The angle #4is determined byf#4 = 04 -

, where 04 refersthe mass center angle. Here, the mass center angle is computedas follows:

Ang(G) = IT arctan G (13)

In Eq. (13), G is the block mass center vector.Mode Decision Technique for Intra-Prediction-BSSIMD

The block diagram and flow diagram of proposed SANDCTW-CABAC BSSIMD (Improved High-Efficiency Video Coding (IHEVC)) shown in Figure 4 Thus,

Input: NxN blocks

Output: Candidate mode for intra-coding block size

1. Initialize 2.{ 3. For (each chroma block) 4. { 5. Define mass-center vector Gx andGy. 6. Compute the direction of the block mass-center vector as,

180 GY Ang(G) = -arctan()

7. Obtain one direction from partU and the other direction is from segment V .

8. Verify the direction of the mass -focus(center) vector is parallel to the relating symmetry axis of the block. 9.}

10. if (two intra-prediction modes of two components are identical) 11. { 12. Select this specific mode as competitor intra forecast modes. 13. Enumerate the RDO 14. } 15. Else 16. { 17. Select the DC mode 18. Calculate the RDO 19. } 20. For (16x16 block) 21. { 22. Compute the direction of mass-focus(center) vector. 23. Select the particular mode corresponding to the figured direction as a candidate intra prediction mode based on

Model 4 E- , -U] - +7, ] 116 x 16={ 3 8 8 '8 ModeO 4 E-,- U - +,-+T] 8 8 8 8

24. Choose the DC mode as another intra prediction mode candidate. 25. Determine the RDO. 26. For ((4x4 luma blocks)I(8x8 luma blocks)) 27. { 28. Form the subblocks by IHCL and ICHL subsampling method. 29. For (each sub-block formed by IHCL&ICHL) 30. { 31. Compute the directions of the mass- center vector. 32. Select the modes corresponding to the computed directions as a candidate intra prediction modes based on 14x4 (Eq. (11)).

33. Select the DC mode as another intra prediction mode candidate. 34. Compute the RDO. 35. } 36. } 37. } 38. }

the proposed technique achieved minimum bit rate (60 to70fps) for different block size 64 * 64, 32 * 32, 16 * 16,8 * 8, and 4 * 4 in HEVC. Also, the proposed encoder supports the selection of candidate modes for 64 * 64, and32 * 32 blocks by reducing the computational complexity.

IHEVC: By combining both SANDCTW-CABAC and BSSIMD encoder in HEVC, its performance is enhanced by reducing computation complexity of RDO, memory cost and improving the data compression rate.

In this section, the various test sequence analyzed for existing AVC (Adaptive Video Coding), HEVC, HEVCCABAC,HEVC-CABAC-FSM (Finite State Machine),HEVC-CABAC-CTW (Context Tree Weighting) proposedHEVC-SANDCTW-vl, HEVC-SANDCTW-v2, HEVCBSSIMD and IHEVC. The numerous performances assessedinvarious benchmarks like, Sum of Absolute Error(SAE), Bit Rate (BR), Compression ratio and Peak Signal to-Noise Ratio (PSNR). Table 1 represents the performance metrics values of four video sequences for AVC,HEVC, HEVC-CABAC, HEVC CABAC-FSM and HEVC-CABAC-CTW techniques.

PSNR expressed that the portion of better extreme signal power and the corrupting noise power which influences the loyalty of its representations. For the most part, is characterized by utilizing Mean Square Error (MSE) and the arrived at the midpoint of PSNR esteems of luma (Y) and chroma (U,V) are computed as follows:

PSNR = 10 logo 5 (14)

In Eq. (14), MSE refers the averaged MSE and defined asfollows:

MSE 4XMSEy+MSEU+MSE" (15) 6

MQSE = =1EL |L(p, q) - L'(q, p)| 2 (16)

The following Table 2 demonstrates the performance metrics estimations of four different video sequences for HEVC- SANDCTW-vl, HEVC- SANDCTW -v2,HEVC- BSSIMD and Improved HEVC (IHEVC)techniques.

In Eq. (16), L and L' refers the original and reconstructedluma and chroma pictures respectively, (p, q) is the coordinates, and P and Q are the height and width of pictures.

Figure 5 shows the comparison and experimental result of PSNR for all video sequences. In this graph, the x-axis denotes the different types of video sequences, and the y- axis denotes the PSNR (dB) values. It observed that the IHEVC shows high performance when compared to the other techniques regarding maximised PSNR. For example, consider the Basketball Drive video sequence, if the bit rate is 18,850 kbps, then the PSNR value of IHEVC is 4 .2 6 % higher than AVC, 3 .9 2 % higher than HEVC, 3 .4 5% higher than HEVC-CABAC, 2 .9 7 % greater than HEVC CABAC-FSM, 2 .2 8 % higher than HEVC- CABAC-CTW, 1. 96 % higher than HEVC- SANDCTW -v1, 1. 26 % higher than HEVC- SANDCTW -v2 and 0.77% higher than HEVC- BSSIMD.

Figure 6 illustrates comparison of SAE for four video sequences in accordance with an embodiment of the present disclosure. The variance between the measured or incidental value of a quantity and its actual value is called the absolute error. The absolute error of the sum or distinction of some quantities is a smaller amount than or adequate to the sum of their absolute errors. Figure 6 shows the comparison of SAE for all video sequences. In this graph, the x-axis denotes the different types of video sequences, and the y- axis denotes the SAE values. It observed that the IHEVC is high performance when compared to the other techniques. For example, consider the Basketball Drive video sequence, if the bit rate is 18850Kbps, then the SAE value of IHEVC is 4 3 .9 3 % less than AVC. 40. 3 6 % less than HEVC, 3 6 .3 4 % less than HEVC-CABAC, 34 .11% less than HEVC-CABAC-FSM, 30. 6 1% less than HEVC-CABAC CTW, 2 6 . 4 5% less than HEVC- SANDCTW -v1, 17.51%less than HEVC SANDCTW -v2 and 11. 3 % less than HEVC- BSSIMD.

Figure 7 illustrates comparison and experimental result of Compression Ratio (%) for four video sequences in accordance with an embodiment of the present disclosure. Compression proportion (ratio) is well-defined that the proportion (ratio) between the uncompressed bit rate and compressed bit rate and given as follows:

CompressionRatio = Uncompressed BitRate (17)

Figure 7 shows the comparison and experimental result of compression ratio for all video sequences. In this graph, the x-axis denotes the different types of video sequences, along with the axis-y means the compression ratio (%) values. It is identified such the IHEVC is high performance when compared to the other techniques. For example, con- sider the Basketball-Drive video sequence if the bit rate is 18850Kbps, then the compression ratio value of IHEVC is 1 2 .3 9 % higher than AVC. 9 .9 6 % higher than HEVC, 8 .8 2 % higher than HEVC-CABAC, 6. 2 1% greater than HEVC-CABAC-FSM, 4 .50% higher than HEVC- CABAC CTW, 3 .3 2 % higher than HEVC-SANDCTW-v1, 2 .3 2 % higher than HEVC-SANDCTW-v2 and 1. 63% higher than HEVC- BSSIMD.

In an embodiment, it is concluded that the conventional video compression standard H.265/HEVC enhanced by improved CABAC version. In the proposed approach, a Self-Adaptive N-depth context tree weighting technique introduced for replacing the conventional CTW by utilizing only the disconnected KT estimator for non-stationary distributions. Moreover, a robust, quick block size sustained intra mode detection technique is proposed based on the mass-center theory and two sub-sampling techniques such as ILHC and ICHL. In the proposed strategy, the vector of mass-center applied for determining all possible correlation directions of the blocks corresponding to the intra-prediction mode. After that, the modes compare to the figured directions are chosen as the best intra-prediction candidates who utilized for RDO computation. Thus, this improved H.265/HEVC reduced the computation complexity of RDO, memory cost and improved data compression rate. Also, the pro posed method achieves the reduction of bit rate according to the distinct block ranges from 4 * 4 to 64 * 64. Finally, the trial comes about demonstrated that the proposed IHEVC strategy is preferred execution than the other techniques. Future extension of this work is to implement the proposed compression technique in fast inter prediction with new clustering method.

The usage of video content has increased in past ten decades. As a result, increase in usage of commercial video coding standards called High Efficiency Video Coding (HEVC/H.265). A self-Adaptive N-depth Context Tree Weighting technique (SANDCTW) is pro- posed to overcome the limitations of Context tree weighting (CTW) method, which applied in CABAC know as Context Adaptive Bin- ary Arithmetic Coding. This CABAC uses KT estimators and relies on beginning with the Bayesian approach to determine the true distribution of the next symbol to select for data compression. This approach is suitable only if the true distribution is stationary, the proposed SANDCTW that uses discounted KT estimators, which is suitable if the distribution is non-stationary and it reduces the computation and memory cost. Additionally, Block size sustained intra mode detection (BSSIMD) is proposed based on the mass-center and sub sampling approach. In this approach, all correlation directions about the entire block associated to the intra- prediction mode and DC mode directions determined by using mass-center vector. Then, the modes corresponding to the determined directions selected as the best intra prediction candidates during the intra-coding process for computing Rate Distortion Optimization (RDO) with less complexity. The bit rate of 60-70 frames per second (fps) achieved in this technique based on the different block, size. Thus, the bit rate is also reduced significantly compared with the preceding H.265 standard. The drawings and the forgoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, orders of processes described herein may be changed and are not limited to the manner described herein. Moreover, the actions of any flow diagram need not be implemented in the order shown; nor do all of the acts necessarily need to be performed. Also, those acts that are not dependent on other acts may be performed in parallel with the other acts. The scope of embodiments is by no means limited by these specific examples. Numerous variations, whether explicitly given in the specification or not, such as differences in structure, dimension, and use of material, are possible. The scope of embodiments is at least as broad as given by the following claims.

Benefits, other advantages, and solutions to problems have been described above with regard to specific embodiments. However, the benefits, advantages, solutions to problems, and any component(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential feature or component of any or all the claims.

Claims (8)

WE CLAIM

1. A self-adaptive N-depth context tree weighting method for replacing conventional CTW using disconnected KT estimator, the method comprises:

applying a self-adaptive N-depth context tree weighting technique (SANDCTW) in context adaptive binary arithmetic coding (CABAC) to overcome the limitations of context tree weighting (CTW) method; proposing block size sustained intra mode detection (BSSIMD) on mass-center and sub-sampling approach; associating all correlation directions about entire block to the intra prediction mode and determining DC mode directions by using mass center vector; and selecting modes corresponding to the determined directions as best intra-prediction candidates during the intra-coding process for computing rate-distortion optimization(RDO) with less complexity.

2. The method as claimed in claim 1, wherein block mass-center direction is perpendicular to the block correlation direction, wherein such directions are horizontal-up, horizontal-down, vertical left, and vertical right.

3. The method as claimed in claim 2, wherein the correlation direction of these blocks determined with mass-center direction by forming their corresponding symmetric sub-block.

4. The method as claimed in claim 1, wherein proposed two sub sampling techniques are impair lines and half columns (ILHC) and impair columns and half lines (ICHL).

5. The method as claimed in claim 4, wherein ILHC is applied to the vertical-left, and vertical-right directional correlation blocks and ICHL applied to the horizontal-up and horizontal-down correlation direction of blocks.

6. The method as claimed in claim 5, wherein ILHC sub-sampling method forms the square sub-blocks by subsampling the block pixels with two major steps such as the impair lines of the blocks selected, and then the half columns in the middle of the block are selected.

7. The method as claimed in claim 1, wherein eight correlation directions of the blocks are determined if the vector direction of the sub block formed by the subsampling techniques is determined.

8. The method as claimed in claim 1, wherein the block size sustained intra-prediction in H.265/HEVC is proposed in the view of mass-center and two subsampling techniques.