HARDEEPSINH JADEJA UTA ID:1000721847 The operation of converting video in one format to another format. It is the ability to take existing video content and change the bitrate and/or resolution in order to view it on another video playback device MPEG-2 [4] has been a widely accepted video coding standard for various applications ranging from DVD to digital tv. broadcast. The most important goal of MPEG-2 was to make the storage and transmission of digital AV material more efficient. However new H.264 AVC standard [1] has an even broader perspective to support high and low bit-rate multimedia applications on existing and future networks. H.264 requires half of the space to store video content when compared to MPEG-2[1]. Hence there is need to find compatibility between MPEG-2 and H.264 devices.Fig.1 MPEG-2 encoder/decoder [1]Fig.2 H.264 Encoder [9].MPEG-2 H.264 Macroblock prediction with ¼ pel accuracy No, only ½ pel accuracy yes Macroblock modes 16x16 16x16,16x8,8x16, 8x8,8x4,4x8, 4x4 Multiple reference prediction no yes Direct modes in B frames no yes Use of B frames as reference frames no Allowed, can be selected by the userFig.3 Different macroblocks in H.264[8].MPEG-2 H.264 Macroblock modes supported 8x8 16x16 with 4 directional modes, 4x4 and 8x8 with 9 directional modes Type of intra prediction Fixed prediction of D.C. coefficient Adaptive directional prediction of 4x4 or 16x6 pixel blocks Transform 8x8 DCT 4x4 Integer transform In H.264, inter frame motion estimation is performed using different MB sizes from 16 16 to 4 4 [8]. For each MB, all different sizes are evaluated and the one leading to the minimum rate-distortion (RD) cost is then selected. This guarantees the optimal block size for the final encoding but at the expense of a high-computational cost. Intra mode selection algorithm[10]. Only intra spatial-prediction scheme[11]. Fast mode decision algorithm for intra prediction for H.264/AVC [12]. Dynamic control of motion estimation search parameters for low complexity H.264[13]. Adaptive algorithm for fast motion estimation [17]. Adaptive algorithm for fast motion estimation in H.264/MPEG-4 AVC [16]. 12 It is a branch of science which develops algorithms to allow computers to evolve or become smart. Machine learning algorithms are applied in large number fields: machine vision, medical diagnostics, image processing, wireless communication are just few among them. 13 It was developed by J.R.Quinlan. It uses divide and conquer approach to develop a tree. Uses two possible criteria to carry out a test at each node of the tree: information gain and gain ratio. 14 Select number of frames of a video sequence in QCIF and CIF as training sequences. Obtain the required attributes off line Encode the training sequence using full complexity H.264 encoder Store the attributes calculated off line and mode decision taken by encoder in ARFF file Feed this ARFF file to weka tool, which will give decision tree. 15 Mask the motion estimation part in the actual H.264 encoder. Overwrite that with if-else statements based on the decision tree. Compare the performance of the simple cascaded transcoder with our transcoder. 16 Creating the Training Files Fig.4 Training algorithm for decision tree[2].Decision Trees Fig.5 Decision tree for deciding the macro block mode of H.264 encoder [1].ResultsQP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 40.246 40.194 167.25 170.23 1309.124 364.59 32 37.654 37.602 103.24 105.07 1120.317 312 36 35.186 35.134 68.07 67.246 934.22 260.18 40 32.638 32.586 46.45 47.27 819.643 228.27 Test sequence:Akiyo_CIF No. of Frames:100 Width:352 Height:288 PSNR:-0.052(dB) Bitrate(kbps)=1.78% Encoding time(sec.)=-72.55% Table 1. Time,PSNR and Bit rate for Akiyo CIF sequence.QP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 38.775 38.738 60.52 61.076 220.425 62.049 32 35.783 35.746 39.87 40.236 216.823 61.0356 36 33.125 33.088 27.08 27.329 211.004 59.397 40 30.515 30.4781 18.98 19.15 183.513 51.659 Test sequence:Akiyo_qcif No. of Frames:100 Width:176 Height:144 PSNR:-0.037(dB) Bitrate(kbps)=0.92% Encoding time(sec.)=-71.85% Table 2. Time,PSNR and Bit rate for Akiyo QCIF sequence.QP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 36.568 36.489 796.99 825.76 1040.353 280.48 32 33.282 33.203 481.00 498.37 834.965 225.106 36 30.226 30.147 283.88 294.13 799 215.41 40 27.388 27.309 173.26 179.52 797 214.87 Test sequence:Paris_CIF No. of Frames:100 Width:352 Height:288 PSNR:-0.079(dB) Bitrate(kbps)=3.61% Encoding time(sec.)=-73.04% Table 3. Time,PSNR and Bit rate for ParisCIF sequenceQP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 35.757 35.679 618.87 652.165 260.857 72.23 32 32.116 32.046 407.52 429.44 201.505 55.795 36 28.785 28.722 240.44 253.37 199.260 55.175 40 25.707 25.65 140.49 148.05 197.326 54.64 Test sequence:Paris_QCIF No. of Frames:100 Width:176 Height:144 PSNR:-0.216(dB) Bitrate(kbps)=5.38% Encoding time(sec.)=-72.31% Table 4. Time,PSNR and Bit rate for Paris QCIF sequenceQP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 35.351 35.339 2317.97 2340.182 1078.516 328.516 32 31.415 31.4031 1215.71 1207.38 1008.546 307.203 36 28.088 28.0763 571.88 571.83 1006.631 306.619 40 25.324 25.312 301.13 304.02 999.369 304.408 Test sequence: Mobile_CIF No. of Frames:100 Width:352 Height:288 PSNR: -0.0115(dB) Bitrate(kbps)=0.96% Encoding time(sec.)= -69.54% Table 5. Time, PSNR and Bit rate for Mobile CIF sequenceQP PSNR Bit rate Encoding time JM Proposed JM Proposed JM Proposed 28 34.3 34.186 557.06 579.47 241.382 69.71 32 30.328 30.176 283.61 295.02 233 67.29 36 27.106 26.992 141 146.67 227 65.56 40 24.447 24.333 79.29 82.48 214.875 62.05 Test sequence:Mobile_qcif No. of Frames:100 Width:176 Height:144 PSNR:-0.052(dB) Bitrate(kbps)=1.78% Encoding time(sec.)=-72.55% Table 6. Time, PSNR and Bit rate for Mobile QCIF sequence0510152025303540450 500 1000 1500 2000 2500PSNR(dB) bit rate(kbps) Bit Rate v/s PSNR using JM reference software for CIF akiyo_cifparis_cifmobile_cifFig.6 Bit Rate v/s PSNR using JM reference software for CIF.0510152025303540450 500 1000 1500 2000 2500PSNR(dB) Bit rate(kbps) Bit Rate v/s PSNR for CIF using Proposed method
View Full Document
Unlocking...