Pub Date : 2010-10-01DOI: 10.1109/MMSP.2010.5662040
S. Momcilovic, Yige Wang, S. Rane, A. Vetro
Most distributed source coding schemes involve the application of a channel code to the signal and transmission of the resulting syndromes. For low-complexity encoding with superior compression performance, graph-based channel codes such as LDPC codes are used to generate the syndromes. The encoder performs simple XOR operations, while the decoder uses belief propagation (BP) decoding to recover the signal of interest using the syndromes and some correlated side information. We consider parallelization of BP decoding on general-purpose multi-core CPUs. The motivation is to make BP decoding fast enough for realtime applications. We consider three different BP decoding algorithms: Sum-Product BP, Min-Sum BP and Algorithm E. The speedup obtained by parallelizing these algorithms is examined along with the tradeoff against decoding performance. Parallelization is achieved by dividing the received syndrome vectors among different cores, and by using vector operations to simultaneously process multiple check nodes in each core. While Min-Sum BP has intermediate decoding complexity, a “vectorized” version of Min-Sum BP performs nearly as fast as the much simpler Algorithm E with significantly fewer decoding errors. Our experiments indicate that, for the best compromise between speed and performance, the decoder should use Min-Sum BP when the side information is of good quality and Sum-Product BP otherwise.
{"title":"Toward realtime side information decoding on multi-core processors","authors":"S. Momcilovic, Yige Wang, S. Rane, A. Vetro","doi":"10.1109/MMSP.2010.5662040","DOIUrl":"https://doi.org/10.1109/MMSP.2010.5662040","url":null,"abstract":"Most distributed source coding schemes involve the application of a channel code to the signal and transmission of the resulting syndromes. For low-complexity encoding with superior compression performance, graph-based channel codes such as LDPC codes are used to generate the syndromes. The encoder performs simple XOR operations, while the decoder uses belief propagation (BP) decoding to recover the signal of interest using the syndromes and some correlated side information. We consider parallelization of BP decoding on general-purpose multi-core CPUs. The motivation is to make BP decoding fast enough for realtime applications. We consider three different BP decoding algorithms: Sum-Product BP, Min-Sum BP and Algorithm E. The speedup obtained by parallelizing these algorithms is examined along with the tradeoff against decoding performance. Parallelization is achieved by dividing the received syndrome vectors among different cores, and by using vector operations to simultaneously process multiple check nodes in each core. While Min-Sum BP has intermediate decoding complexity, a “vectorized” version of Min-Sum BP performs nearly as fast as the much simpler Algorithm E with significantly fewer decoding errors. Our experiments indicate that, for the best compromise between speed and performance, the decoder should use Min-Sum BP when the side information is of good quality and Sum-Product BP otherwise.","PeriodicalId":105774,"journal":{"name":"2010 IEEE International Workshop on Multimedia Signal Processing","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117225061","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2010-10-01DOI: 10.1109/MMSP.2010.5662039
Alberto Corrales-García, José Luis Martínez, G. Fernández-Escribano
Distributed Video Coding (DVC) provides a new coding paradigm based on lower complex encoders than decoders. On the decoder side some missed frames have to be estimated by means of available frames and a correlation noise model. In addition, parity bit chunks can be requested to the encoder across the feedback channel to correct the mismatches of these frames. This is an iterative procedure which collets most of the complexity of the decoder. In this work, a novel approach is proposed to parallelize the DVC decoding process in a multicore system. In this way, each bitplane is decoded at the same time by a different core and they exchange information to update the integration limits of the probably model, reaching a time reduction up to 80% with a little bitrate penalty but maintaining the same PSNR.
{"title":"Reducing DVC decoder complexity in a multicore system","authors":"Alberto Corrales-García, José Luis Martínez, G. Fernández-Escribano","doi":"10.1109/MMSP.2010.5662039","DOIUrl":"https://doi.org/10.1109/MMSP.2010.5662039","url":null,"abstract":"Distributed Video Coding (DVC) provides a new coding paradigm based on lower complex encoders than decoders. On the decoder side some missed frames have to be estimated by means of available frames and a correlation noise model. In addition, parity bit chunks can be requested to the encoder across the feedback channel to correct the mismatches of these frames. This is an iterative procedure which collets most of the complexity of the decoder. In this work, a novel approach is proposed to parallelize the DVC decoding process in a multicore system. In this way, each bitplane is decoded at the same time by a different core and they exchange information to update the integration limits of the probably model, reaching a time reduction up to 80% with a little bitrate penalty but maintaining the same PSNR.","PeriodicalId":105774,"journal":{"name":"2010 IEEE International Workshop on Multimedia Signal Processing","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125802153","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2010-10-01DOI: 10.1109/MMSP.2010.5662002
A. A. Moghadam, M. Aghagolzadeh, Mrityunjay Kumar, H. Radha
A typical consumer digital camera uses a Color Filter Array (CFA) to sense only one color component per image pixel. The original three-color image is reconstructed by interpolating the missing color components. This interpolation process (known as demosaicing) corresponds to solving an under-determined system of linear equations. In this paper, we show that by replacing the traditional CFA with a random panchromatic CFA, recent results in the emerging field of Compressed Sensing (CS) can be used to solve the demosaicing problem in a novel way. Specifically, during the image reconstruction process, we exploit the fact that the multi-dimensional color of each pixel has a compressible representation in a (possibly overcomplete) color system. While adhering to the “single color per pixel sensing” constraint at the sensing stage, during the reconstruction process we utilize the inter-pixel correlation by exploiting the compressible representation of the overall image in some sparsifying bases. Depending on the CFA, sparsifying bases and the color system, we form an underdetermined system of linear equations and find the sparsest solution for the color image by utilizing a CS solver. We illustrate that, for natural images, the proposed Compressive Demosaicing (CD) framework visually outperforms leading demosaicing methods in a consistent manner; in many cases it achieves clear visible improvements in a significant way.
{"title":"Compressive demosaicing","authors":"A. A. Moghadam, M. Aghagolzadeh, Mrityunjay Kumar, H. Radha","doi":"10.1109/MMSP.2010.5662002","DOIUrl":"https://doi.org/10.1109/MMSP.2010.5662002","url":null,"abstract":"A typical consumer digital camera uses a Color Filter Array (CFA) to sense only one color component per image pixel. The original three-color image is reconstructed by interpolating the missing color components. This interpolation process (known as demosaicing) corresponds to solving an under-determined system of linear equations. In this paper, we show that by replacing the traditional CFA with a random panchromatic CFA, recent results in the emerging field of Compressed Sensing (CS) can be used to solve the demosaicing problem in a novel way. Specifically, during the image reconstruction process, we exploit the fact that the multi-dimensional color of each pixel has a compressible representation in a (possibly overcomplete) color system. While adhering to the “single color per pixel sensing” constraint at the sensing stage, during the reconstruction process we utilize the inter-pixel correlation by exploiting the compressible representation of the overall image in some sparsifying bases. Depending on the CFA, sparsifying bases and the color system, we form an underdetermined system of linear equations and find the sparsest solution for the color image by utilizing a CS solver. We illustrate that, for natural images, the proposed Compressive Demosaicing (CD) framework visually outperforms leading demosaicing methods in a consistent manner; in many cases it achieves clear visible improvements in a significant way.","PeriodicalId":105774,"journal":{"name":"2010 IEEE International Workshop on Multimedia Signal Processing","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2010-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132891003","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：