American-Eurasian Journal of Scientific Research 11 (5): 356-360, 2016 ISSN 1818-6785 IDOSI Publications, 2016 DOI: 10.5829/idosi.aejsr.2016.11.5.22951 Region Based Texture Approximation Technique for Efficient Video Compression 1 2 Tamil Selvi and A. Rajiv Kannan 1 Research Scholar, Anna University, Chennai, India 2 Department of Computer Science and Engineering, K.S.Rangasamy College of Engineering, Thiruchengode, India Abstract: To improve e performance of video compression produced by e previous solution, we propose a region based texture approximation technique. There are situations where e tiny part of e image would vary and transmitting e entire image will not necessary. To overcome such problem, e meod splits e video into frames and for each successive frame e meod splits e frame into number of regions. Between regions of two successive frames e meod computes e feature similarity. Based on e feature similarity e meod sends e selective region to e oer end, where e meod retains e feature of e previous frames for e un received region. This improves e performance of e video compression an e previous meod. Key words: Digital Auentication E-Learning Video Compression Feature Approximation INTRODUCTION When you consider e video conferencing, e video must be transmitted in short time and if ere is a The modern information technology uses various delay en e video will become meaningless. In order to forms of data from numerical and textual to multi media. achieve e data transmission efficiency, e video must Most of e business process is performed rough e be transmitted in short time. To achieve is, e video internet and to maintain e secrecy of sensitive sequence has to be compressed, so at e bandwid information, e organization use multimedia information. consumption will be reduces. To perform video Furer e educational institutions have shifted to compression, ere are number of meods discussed provide E-learning. E-learning is e process learning earlier. But e meods suffer to achieve e compression e subject rough internet. By providing E-learning, ratio and suffer to produce video compression in short e learner can educate from eir own location and time. ey need not go to e institution. There will be The most meods transmit e entire video frame and number of tutorials available and to provide E-learning, at also increase e bandwid occupation. If you e entire seminar video has to be transferred to e user consider two adjacent frames of any video, ere will be a [1]. limited feature variation. So transmitting e entire video To provide e video to e user who request e frame of same snapshot will not produce any efficiency in video, it must be transferred to e user side. The video is video transmission. In some oer meods, e meod large in size and it occupies more bandwid. The network transmits e variation values of two successive frames, has intermediate nodes and has only limited bandwid. which also introduces e overhead. To overcome such In order to provide efficient service e video has to be deficiency, e image can be split into number of small transmitted wiin short time. The video is a collection of regions. By computing e feature similarity between e frames and ere will be number of frames in a single regional images, e region which differs can be sent to second video. So transmitting e video in e original e remote side. This can help e video compression to form requires more bandwid and increases e latency be performed in efficient manner and can reduce e also [2]. bandwid utilization. Corresponding Auro: Tamil Selvi, Research Scholar, Anna University, Chennai, India. 356
Related Works: There are number of video compression motion parameter and inter-view residual prediction for techniques discussed earlier and is section discusses coding of e dependent video views are developed and about e meods. integrated. Furermore, for dep coding, new intra Video Compression Algorim Based on All Phase coding modes, a modified motion compensation and Biorogonal Transform and MPEG-2 [1], works based on motion vector coding as well as e concept of motion e all phase biorogonal transform (APBT) eory, parameter inheritance are part of e HEVC extension. which has ree kinds of forms in accordance wi A novel encoder control uses view synesis different transform matrices, referred to as e all phase optimization, which guarantees at high quality Walsh biorogonal transform (APWBT), e all phase intermediate views can be generated based on e discrete cosine biorogonal transform (APDCBT) and e decoded data. The bitstream format supports e all phase inverse discrete cosine biorogonal transform extraction of partial bitstreams, so at conventional 2D (APIDCBT). Compared wi e conventional DCT, APBT video, stereo video and e full multi-view video plus reduces e inter-pixel redundancy and e computational dep format can be decoded from a single bitstream. complexity using e uniform quantization for e intra Efficient intra prediction algorim for smoo regions frames transform-coding. Experimental results show at in dep coding [8-10], aiming to efficiently encode e peak signal to noise ratio (PSNR) of e proposed smoo regions in dep maps. By taking e textureless algorim performs close to e DCT for e tested frames characteristics of dep maps into account, only one and ere is no difference in visual quality [3]. single prediction direction instead of multiple prediction Low-complexity dep map compression in directions is sufficient in intra prediction of dep maps. HEVC-based 3D video coding [4], discuss a low- Consequently, coding of e prediction direction can be complexity algorim is proposed to reduce e complexity skipped which results in lower computational complexity of dep map compression in e high-efficiency video and higher coding efficiency for synesized views [11]. coding (HEVC)-based 3D video coding (3D-HEVC). Dep and dep color coding using shape-adaptive Since e dep map and e corresponding texture video wavelets [12-15], present a novel dep and dep color represent e same scene in a 3D video, ere is a high codec aimed at free-viewpoint 3D-TV. The proposed correlation among e coding information from dep map codec uses a shape-adaptive wavelet transform and an and texture video. An experimental analysis is performed explicit encoding of e locations of major dep edges. to study dep map and texture video correlation in e Unlike e standard wavelet transform, e shape-adaptive coding information such as e motion vector and transform generates small wavelet coefficients along prediction mode. Based on e correlation, we propose dep edges, which greatly reduces e bits required to ree efficient low-complexity approaches, including early represent e data. The wavelet transform is implemented termination mode decision, adaptive search range motion by shape-adaptive lifting, which enables fast estimation (ME) and fast disparity estimation (DE). computations and perfect reconstruction. We derive a Standardized extensions of high efficiency video simple extension of typical boundary extrapolation coding [5], describes extensions to e High Efficiency meods for lifting schemes to obtain as many vanishing Video Coding (HEVC) standard at are active areas of moments near boundaries as away from em. We also current development in e relevant international develop a novel rate-constrained edge detection standardization committees. While e first version of algorim, which integrates e idea of significance HEVC is sufficient to cover a wide range of applications, bitplanes into e Canny edge detector. needs for enhancing e standard in several ways have All e above discussed meods produces poor been identified, including work on range extensions for compression rate and produces higher time complexity. color format and bit dep enhancement, embeddedbitstream scalability and 3D video. Regional Texture Approximation Technique: The 3D high efficiency video coding for multi-view video regional feature approximation technique splits e image and dep data [6, 7], describes an extension of e high into small scale frames. For each tiny image generated e efficiency video coding (HEVC) standard for coding of meod extracts e features and compares e features to multi-view video and dep data. In addition to e known find out e variation. The process has been split into concept of disparity-compensated prediction, inter-view number of stages and each will be discussed clearly. 357
Input Video Multi Level Covariance Measure Based Video Compression Preprocessing Fig. 1: Architecture of Region based approach Covariance Matrix Generation Video Compression The Figure 1, shows e architecture of e proposed region based approach and shows e functional components. Preprocessing: At is stage, e input video is taken into processing and splits e entire video into number of sub sampling image. The generated image is applied wi histogram equalization, which improves e quality image and removes e noise from e image. The generated image will be used to perform feature extraction in e next stage. e meod computes features similarity between e oer in e next frame. Based on e similarity value e meod selects required regions to be transmitted. Region Based Texture Approximation Algorim: Input: Frame Set Fs Output: Region Set Rs Start For each Frame Fi from Fs Read Previous Frame Pf. Split Frames into sectional images. size( Pf ) i = 1 size( Cf ) i = 1 Preprocessing Algorim: SI p = Crop( Pf, size) Input: Video V Output: Frame set Fs SI c = Crop( Cf, size) Start Read Input Video V. For each sectional image Si from SI Split Video into frames. compute similarity measure. size( SI ) Fs = Split (V) Rsim = ΣDist( SI () i SIC() i i = 1 For each frame Fi from Fs size Fs(i) = Histogram-Equalization(Fi). If Rsim>Th en Add region to region set. Stop. The above discussed preprocessing algorim splits Stop. e video into number of frames and for each frame generated, e histogram equalization is applied to The region based approximation technique splits e enhance e image quality. image into number of sectional images and for each region e sectional similarity is computed. Based on computed Region Based Texture Approximation: For any two similarity value e region will be selected. The similarity successive frames of e video, e meod splits e value decides wheer e region has to be sent as it is or frame into number of small scale regions. For each region not. 358
Video Compression: To compress e input video given, e meod first performs preprocessing and en between each successive frames e meod performs e region based feature approximation. Based on e feature similarity value computed e meod selects required region of e image to be transmitted. On e oer end e meod reconstruct e frame wi e received and wi e regions of previous frame. RESULTS AND DISCUSSION The proposed region based feature approximation techniquefor video compression has been implemented using matlab and e performance of e meods has been evaluated using different videos. The meods have produced efficient result in compression ration and reduce e distortion ratio an oer meods. The Graph 1, shows e comparison of video compression ratio being achieved by different meods and it shows clearly at e proposed meods has produced more video compression ratio an oer meods. The Graph 2, shows e comparison of distortion ratio being produced by different meods and it shows clearly at e proposed meod has produced less distortion ratio and e meods reduces e distortion ratio rapidly. The Table 1, shows e comparison of different video compression measures produced and it shows at e proposed meod has produced efficient results. Graph 1: Comparison of video compression ratio Graph 2: Comparison of distortion ratio 359
Table 1: Comparison of various video compression measures Meod Compression Ratio % Distortion Ratio % Time Complexity in seconds SVC 11 11 87 SATD 14 9 81 HEVC-based 16 7 76 MCV 23 4.8 56 Region Based 29 2.1 45 CONCLUSION 6. Sullivan, G.J., J.R. Ohm, W.J. Han and T. Wiegand, In is paper, an efficient region based feature 2012. Overview of e high efficiency video coding (HEVC) standard, IEEE Trans. Circuits Syst. Video approximation technique has been discussed to improve Technol., 22(12): 1649-1668. e performance of video compression. The meod first 7. Müller, K., H. Schwarz, D. Marpe, C. Bartnik, splits e input video into number of frames and for each S. Bosse, H. Brust, T. Hinz, H. Lakshman, P. Merkle, frame generated histogram equalization has been applied. H. Rhee, G. Tech, M. Winken and T. Wiegand, 2013. Then e meod splits e image into number of small 3D high efficiency video coding for multi-view scale images and for each sectional image e meod video and dep data, IEEE Trans. Image Process, computes e feature similarity measure. The region based 22(9): 3366-3378. feature approximation returns a small set of region which 8. Zhang, L., G. Tech, K. Wegner and S. Yea, 2013. In can be transmitted to e oer end. If e number of Joint Collaborative Team on 3D Video Coding regions selected from e approximation technique is Extensions (JCT-3V) Document JCT3V-E1005, 5 higher an particular reshold en e entire image has Meeting, 3D-HEVC test model 5 (Vienna, Austria). to be sent. The proposed meod has produced higher 9. Wang, M., X. Jin and S. Goto, 2010. in Proc. 28 rate of video compression and reduces e time Picture Coding Symp., Difference detection based complexity. early mode termination for dep map coding in MVC, pp: 502-505. REFERENCES 10. Tsang, S., Y. Chan and W. Siu, 2012. Efficient intra prediction algorim for smoo regions in dep 1. Jiang Baochen, Chunxiao Zhang, Chengyou Wang coding, Electron. Lett., 48(18): 1117-1119. and Xiaoyan Wang, 2015. Video Compression 11. Cernigliaro, G., F. Jaureguizar, J. Cabrera and Algorim Based on All Phase Biorogonal N. García, 2013. Low complexity mode decision and Transform and MPEG-2, International Journal of motion estimation for H.264/AVC based dep maps Hybrid Information Technology, 8(3): 145-154. encoding in free viewpoint video, IEEE Trans. 2. Liu, X.G., K.Y. Yoo and S.W. Kim, 2010. Low Circuits Syst. Video Techn., 23(5): 769-783. complexity intra prediction algorim for MPEG-2 to 12. Maitrea, M. and M.N. Do, 2010. Dep and H.264/AVC transcoder, IEEE Transactions on dep color coding using shape-adaptive wavelets. Consumer Electronics, 56(2): 987-994. J. Vis. Commun. Image, 21(5-6): 513-522. 3. Wang, C.Y., 2012. Directional APBT and its 13. Milani, P. Zanuttigh, M. Zamarin and S. application in image coding, Proceedings of e 11 Forchhammer, 2011. in Proc. IEEE Int. Conf. IEEE International Conference on Signal Processing, Multimedia and Expo(ICME), Efficient dep map pp: 728-731. compression exploiting segmented color data 4. Zhang Qiuwen, Ming Chen, Xinpeng Huang, (Barcelona), pp: 1-6. Nana Li and Yong Gan, 2015. Low-complexity 14. Shen, L., P. An, Z. Liu and Z. Zhang, 2014. Low dep map compression in HEVC-based 3D video complexity dep coding assisted by coding coding, EURASIP Journal on Image and Video information from color video, IEEE Trans. Processing. Broadcasting, 60(1): 128-133. 5. Sullivan, G.J., J.M. Boyce, C. Ying, J.R. Ohm, C.A. 15. Dandu Anusha and Escalin Tresa, 2015. Hevc Video Segall and A. Vetro, 2013. Standardized extensions of Compression Using Dwt and Block Matching high efficiency video coding (HEVC), IEEE J. Sel. Algorim, ARPN Journal of Engineering and Top. Sign. Proces, 7(6): 1001-1016. Applied Sciences, 10(9). 360