By :- Ramolia Pragnesh R.Guided by :- Dr. K.R.RaoDr. Dongil HanTerm :- Fall-20091Introduction to AVS-M Overview of AVS-M Complexity present in AVS-M encoder Various approaches to reduce complexity Introduction to data mining algorithm: C4.5 Project implementation steps AVS-M execution, and mode and attribute extraction.Future work.2Introduction to AVS-M AVS-M is the seventh part of video coding standard developed by AVS workgroup of China which aims at mobile applications. It has 9 different levels for different formats. It supports only progressive video coding hence codes frames only. It uses only 4:2:0 chroma sub-sampling format. It uses only I and P frames.3Different parts of AVS [10]PartName 1System2Video3Audio4Conformance test5Reference software6Digital media rights management7Mobile video8Transmit AVS via IP network9AVS file format10Mobile speech and audio codingTable 1: Different parts of AVS 4Layered Data Structure Sequence Picture Slice Macro Block BlockSequenceG.O.P.PictureSliceMacro blockBlock5AVS-M Codec Each MB needs to be intra or inter predicted. Switch S0(Fig. 1 ) is used to decide between inter and intra based type of MB. Unit size for intra prediction is block size of 4x4, and predictions are derived based on left and upper blocks. Inter predictions are derived on blocks of varying sizes: 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4 from locally reconstructed frames. Transform coefficients are coded by VLC. Deblocking filter is applied on reconstructed image.6EncoderFigure 1: Encoder of AVS-M [10] 7DecoderFigure 2: Decoder of AVS-M [10]8Major and Minor tools of AVS-M Network abstraction layer (NAL). Supplemental enhancement information (SEI). Transform –4x4 integer transform. Adaptive quantization of step size varying from 0-63. Intra prediction –9 modes (Fig. 3), simple 4x4 intra prediction and direct intra prediction. Motion compensation –16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4 block sizes. Quarter-pixel interpolation –8-tap horizontal interpolation filter and 4-tap vertical interpolation filter. Simplified in-loop de-blocking filter. Entropy coding. Error resilience .9Intra adaptive directional prediction [25]Figure 3: Intra adaptive directional prediction10Intra prediction Intra prediction scheme in AVS-M brings much simplicity as compared to H.264 baseline profile. It uses 4x4 block as the unit for intra-prediction. It uses 2 modes of prediction in intra prediction intra_4x4 and direct intra prediction. Intra_4x4 uses content based most probable intra mode decision as shown in Table 2 to save bits, where U and L represents the upper ad left blocks as shown in Fig. 4. Direct intra prediction brings much of the compression based on trade-off decision.Upper block[U]Left block[L]Current blockFig. 4 : Current block and neighboring block representation11Intra predictionUL -1012345678-18888888888080020002021821222222228222222222382123452724844244464458552555655686666666667877277767788012345678Table 2: Content based most probable mode decision table [25]Mode ‘-1’ is assigned to ‘L’ or “U’ when the current block does not have ‘Left’ or ‘Upper’ block respectively. 12Inter-frame prediction Size of the blocks in inter-frame prediction can be 16x16, 16x8, 8x16, 8x8, 8x4, 4x8, and 4x4 depending on the amount of information present within the macro-block. Motion is predicted up to ¼ pixel accuracy. If the half_pixel_mv_flag is 1 then it is up to ½ pixel accuracy. 8-tap filter F1 = (−1,4,−12,41,41,−12,4,−1) and 4-tap filter F2 = (−1,5,5,−1) are used for horizontal and vertical interpolations respectively for ½ pixel MV search and averaging (liner interpolation) is used for ¼ pixel accuracy as shown in Fig. 6.13Inter frame block sizes: 7 block sizes are present in AVS-M for inter frame prediction [9].08 × 8018 × 8440 18 × 8440 12 38 × 84 444016160116880 18 8160 12 38 888Figure 5: Inter frame prediction block sizes14sub-pixel motion estimation by interpolation[16]:Figure 6: interpolation of sub-pixels (hatched lines show half-pixels, empty circles are quarter-pixels, and capital letters represent full-pixels.)15Error concealment and resilience 3- techniques are used for error concealment –forward, backward and interactive error concealment. For error resilience supplemental enhancement information (SEI) is sent along with the bit-stream which has details of 1) The frame number from which particular block motion starts and 2) Type of motion (zooming out/in, transversal motion in plane etc.). SEI helps to recover any information lost due to transmission error.16AVS-M encoder complexity variable block sizes in Inter Mode. It supports 9 intra_4*4 mode and 1 Direct_intra prediction mode. Full search for motion estimation gives the optimum result, but that comes along with implementation complexity. For example, assuming FS(full search) and M block types, N reference frames and a search range for each reference frame and block type equal to +/- W, we need to examine N x M x (2W + 1)^2 positions compared to only (2W + 1)^2 positions for a single reference/block type.17Continued… 7 inter prediction modes because of 7 different block sizes, 9 intra_4*4 modes and 1 direct intra prediction mode. ¼ and ½ pixel accuracy in motion vector estimation.18Various techniques to reduce complexity Intra mode selection algorithm[26]. Only intra spatial-prediction scheme[27]. Fast mode decision algorithm for intra prediction for H.264/AVC [28]. Dynamic control of motion estimation search parameters for low complexity H.264[29]. Adaptive algorithm for fast motion estimation [30]. Data mining algorithm for fast motion estimation [2].19Data mining algorithm C4.5 Extracts information from data automatically, by computational and statistical methods. Based on the information extracted, develops trees. These trees give the decision statement for mode decision Takes decision based on metrics such as MB mean, MB variance, amplitude of edge detection, residual variance etc.20Goal of this project:-Implement data mining algorithm c4.5 to decide the inter prediction block mode.21Implementation steps:- Select number of frames of a video sequence in QCIF as training sequences. Obtain the required attributes
View Full Document
Unlocking...