Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 99 === Recently, images display on different device with various resolutions. To keep main objects in an image without distortions is more and more important. In this thesis, we present a faster method for content aware image retargeting, using additional information of high level tasks, such as saliency map, results of edge detection, and results of face detection, with original seam carving. After these features are combined into a map, the input image is resized by modified seam carving and then resized by linear scaling interactively until this image is resized to target size for preserving artificial structure in the image. By resizing images with modified seam carving and linear scaling, the proposed approach generates better results than previous works.
|