标题:Blur Detection via Phase Spectrum
作者:Zhang, Renyan
通讯作者:Zhang, Renyan;Zhang, RY
作者机构:[Zhang, Renyan] Shandong Univ Sci & Technol, Coll Elect Engn & Automat, Qingdao, Shandong, Peoples R China.
会议名称:14th Asian Conference on Computer Vision (ACCV)
会议日期:DEC 02-06, 2018
来源:COMPUTER VISION - ACCV 2018, PT VI
出版年:2019
卷:11366
页码:733-748
DOI:10.1007/978-3-030-20876-9_46
关键词:Blur detection; Phase spectrum
摘要:The effectiveness of blur features is very important in blur detection from a single image and the most existing blur features are sensitive to the strong edges in the blurred image region which degrades the detection methods. We analyze the information carried by the reconstruction of an image from the phase spectrum alone (RIPS) and the influences of blurring on RIPS. We find that a clear image region has more intensity changes than a blurred one because the former has more high frequency components. And the local maxima of RIPS are at where these image components occur, which make the RIPS of the clear image regions are obviously bigger than that of the blurred ones. Based on this finding, we proposed a simple blur feature, called Phase Map (PM), generated by thresholding RIPS adaptively. And our blur detection method propagates PM to the final blur map only by filtering PM using the relative total variation (RTV) filter. Our proposed method is evaluated on challenging blur image datasets. The evaluation demonstrates that PM feature is effective for different blur types and our detection method performs better than the state-of-the-art algorithms quantitatively and qualitatively.
收录类别:CPCI-S;EI;SCOPUS
资源类型:会议论文;期刊论文
原文链接:https://www.scopus.com/inward/record.uri?eid=2-s2.0-85066931289&doi=10.1007%2f978-3-030-20876-9_46&partnerID=40&md5=d5bf3f7b67146edea9d7e4a57c0a808f
TOP