Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Seam Carving for Feature-aware Image Resizing

Author DengHuiPeng
Tutor LiuLiGang
School Zhejiang University
Course Applied Mathematics
Keywords Image scaling Paul image feature Line Clipping
CLC TP391.41
Type Master's thesis
Year 2008
Downloads 197
Quotes 2
Download Dissertation

Image scaling technology is a research focus in the field of digital image processing . Adjustment of the resolution of the image , the format conversion , and in the PDA and other portable electronic device in the limited high - resolution image is displayed on the screen ( s ) are required to use an image scaling technology . Traditional image scaling technology uses interpolation algorithm , often appear jagged or loss of image detail when zooming , and the contents of the image a larger scale changes . Recent years , Avidan and Shamir proposed a scaling algorithm based the line clipping holding image content , by adding or deleting the weakest of the pixel lines of the visual importance of image scaling , so that an image in the most important visual information can be in the results retained. However, since the high - level visual information without considering the image , the result of the algorithm can not maintain the characteristics of the image object contour . To this end, we propose two cropped image scaling algorithm based on line : The first method is based on partial adjustment to keep the results of image scaling feature information . The distortions of the image feature is through the feature of the image due to pixel lines result , so the characteristics of the twisted to be adjusted to maintain the features . This algorithm first compute all the pixel lines to be added or deleted through the feature of the image , if so, the pixel lines , then the pixel lines , and the intersection of the image feature is a continuous polyethylene as a class , after clustering based on the slope of the characteristic point and connectivity for each class of partial adjustment , adjust image characteristics is maintained , to overcome this issue of the original algorithm can not maintain the image features . The second method is based deletions of one line of pixels for the influence of the image feature is limited to their intersection neighborhood of the fact that , by introducing the measure of the energy function of the feature point neighborhood similarity , energy optimization the retention feature image scaling algorithm . If the lines of pixels across the image feature when using the dynamic programming algorithm to identify partial adjustment such that the adjusted characteristic point neighborhood similarity maximum path , so that the image feature in the image scaling process is not deformed , to achieve the purpose of maintaining the characteristics of the image .

Related Dissertations
More Dissertations