{"title":"A high-frame-rate dense motion vector field generation processor with simplified best-match searching circuitries","authors":"Yuta Okano, T. Shibata","doi":"10.1109/ASSCC.2009.5357145","DOIUrl":null,"url":null,"abstract":"A high-frame-rate dense motion vector field generation processor employing an efficient data manipulation scheme has been developed. The computation result of detecting a motion vector (MV) at each pixel location is reused in the adjacent location. As a result, the size of a best-match searching circuitry for MV detection has been reduced to 1/10 of that without data reuse. This allows us to implement two MV calculation units on a chip without area penalty, enabling dual MV generation at every clock cycle except for the initial period of over head processing. As a result, the frame rate has been increased by 23% as compared to the previous architecture [1], while the total number of transistors has been reduced by 41%. A prototype chip was designed and fabricated in a 0.18-μm 5-metal CMOS technology. It was experimentally demonstrated that the chip can generate motion vectors from all pixel sites of 256×256-size motion images at a frame rate of 1060 frames/sec with a clock frequency of 100 MHz.","PeriodicalId":263023,"journal":{"name":"2009 IEEE Asian Solid-State Circuits Conference","volume":"28 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2009-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"6","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2009 IEEE Asian Solid-State Circuits Conference","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ASSCC.2009.5357145","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 6
Abstract
A high-frame-rate dense motion vector field generation processor employing an efficient data manipulation scheme has been developed. The computation result of detecting a motion vector (MV) at each pixel location is reused in the adjacent location. As a result, the size of a best-match searching circuitry for MV detection has been reduced to 1/10 of that without data reuse. This allows us to implement two MV calculation units on a chip without area penalty, enabling dual MV generation at every clock cycle except for the initial period of over head processing. As a result, the frame rate has been increased by 23% as compared to the previous architecture [1], while the total number of transistors has been reduced by 41%. A prototype chip was designed and fabricated in a 0.18-μm 5-metal CMOS technology. It was experimentally demonstrated that the chip can generate motion vectors from all pixel sites of 256×256-size motion images at a frame rate of 1060 frames/sec with a clock frequency of 100 MHz.