{"title":"二维不可分四元子准酉滤波器组","authors":"N. Petrovsky, Eugene V. Rybenkov, A. Petrovsky","doi":"10.23919/SPA.2018.8563311","DOIUrl":null,"url":null,"abstract":"This paper presents a novel technique of factorization for 2-D non-separable quaternionic paraunitary filter banks (2-D NSQ-PUFB). Two-dimensional factorization structures called “16in-16out” and “64in-64out” respectively for 4-channel and 8-channel <tex>$\\boldsymbol{Q}$</tex>-PUFB based on the proposed technique are shown. The given structures can be mapped to parallel-pipeline processor architecture with a minimum latency time <tex>$2 (\\boldsymbol{N}+1)$</tex> quaternion multiplication operations, where <tex>$\\boldsymbol{N}$</tex> is transform order of the <tex>$Q$</tex>-PUFB. The latency of parallel-pipeline processing does not depend on the size of the original image in contrast to the conventional 2-D transform. The coding gains <tex>$\\boldsymbol{CG_{MD}}$</tex> of 2-D non-separable Q-PUFBs for the isotropic auto-correlation function model with the correlation factor <tex>$\\boldsymbol{\\rho}=0.95$</tex> are the following: <tex>$\\boldsymbol{C}\\boldsymbol{G}_{MD}=13.4\\ \\text{dB}$</tex> for “16in-16out” structure and <tex>$\\boldsymbol{C}\\boldsymbol{G}_{MD}=15.6\\ \\text{dB}$</tex> for “64in-64out” structure.","PeriodicalId":265587,"journal":{"name":"2018 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)","volume":"66 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"15","resultStr":"{\"title\":\"Two-dimensional non-separable quaternionic paraunitary filter banks\",\"authors\":\"N. Petrovsky, Eugene V. Rybenkov, A. Petrovsky\",\"doi\":\"10.23919/SPA.2018.8563311\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This paper presents a novel technique of factorization for 2-D non-separable quaternionic paraunitary filter banks (2-D NSQ-PUFB). Two-dimensional factorization structures called “16in-16out” and “64in-64out” respectively for 4-channel and 8-channel <tex>$\\\\boldsymbol{Q}$</tex>-PUFB based on the proposed technique are shown. The given structures can be mapped to parallel-pipeline processor architecture with a minimum latency time <tex>$2 (\\\\boldsymbol{N}+1)$</tex> quaternion multiplication operations, where <tex>$\\\\boldsymbol{N}$</tex> is transform order of the <tex>$Q$</tex>-PUFB. The latency of parallel-pipeline processing does not depend on the size of the original image in contrast to the conventional 2-D transform. The coding gains <tex>$\\\\boldsymbol{CG_{MD}}$</tex> of 2-D non-separable Q-PUFBs for the isotropic auto-correlation function model with the correlation factor <tex>$\\\\boldsymbol{\\\\rho}=0.95$</tex> are the following: <tex>$\\\\boldsymbol{C}\\\\boldsymbol{G}_{MD}=13.4\\\\ \\\\text{dB}$</tex> for “16in-16out” structure and <tex>$\\\\boldsymbol{C}\\\\boldsymbol{G}_{MD}=15.6\\\\ \\\\text{dB}$</tex> for “64in-64out” structure.\",\"PeriodicalId\":265587,\"journal\":{\"name\":\"2018 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)\",\"volume\":\"66 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2018-09-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"15\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2018 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.23919/SPA.2018.8563311\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2018 Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.23919/SPA.2018.8563311","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
This paper presents a novel technique of factorization for 2-D non-separable quaternionic paraunitary filter banks (2-D NSQ-PUFB). Two-dimensional factorization structures called “16in-16out” and “64in-64out” respectively for 4-channel and 8-channel $\boldsymbol{Q}$-PUFB based on the proposed technique are shown. The given structures can be mapped to parallel-pipeline processor architecture with a minimum latency time $2 (\boldsymbol{N}+1)$ quaternion multiplication operations, where $\boldsymbol{N}$ is transform order of the $Q$-PUFB. The latency of parallel-pipeline processing does not depend on the size of the original image in contrast to the conventional 2-D transform. The coding gains $\boldsymbol{CG_{MD}}$ of 2-D non-separable Q-PUFBs for the isotropic auto-correlation function model with the correlation factor $\boldsymbol{\rho}=0.95$ are the following: $\boldsymbol{C}\boldsymbol{G}_{MD}=13.4\ \text{dB}$ for “16in-16out” structure and $\boldsymbol{C}\boldsymbol{G}_{MD}=15.6\ \text{dB}$ for “64in-64out” structure.