Seam Carving Algorithm For Image Retargeting Based On Saliency And Distortion Control

Advances in imaging technology have made the capture and display of digital images ubiquitous. A variety of displays are used to view them, ranging from high-resolution computer monitors to low-resolution mobile devices, and images often have to undergo changes in size and aspect ratio to adapt to different screens. Also, displaying and printing documents with embedded images frequently entail resizing of the images to comply with the overall layout.Images are typically authored once, but need to be adapted for consumption under varied conditions. This diversity of image consumption conditions introduces a new problem:images must be resized for optimal display or use in different applications. The process, also known as image retargeting or image resizing, consists of modifying the image’s aspect ratio and size in order to best satisfy the new requirements.Traditional retargeting techniques use straightforward image resizing operators, such as scaling and fix-window cropping, usually do not produce satisfactory results, since they are oblivious to image content. To overcome this limitation, a class of techniques attempt to resize the images in a content-aware fashion, which is taking the image content into consideration to preserve important regions and minimize distortions. This is a challenging problem, as it requires preserving the relevant information while maintaining an aesthetically pleasing image for the user. To tackle this challenging problem, a novel seam carving algorithm for image retargeting based on image distortion control was proposed in this thesis, characterizing with its capability of reducing the information loss and geometric distortion for the resulting image of retargeting. This method modified the traditional seam carving algorithm from two aspects, which are listed as follows:1) A saliency map combining local and global contrast saliency was employed as the energy map for seam carving, and it sorted out the seams need to be carved, with the purpose of better preserving the regions of interest(ROI) of the image, thus the major content information of the image was maintained.2) The dense SIFT flow was utilized to measure the degree of image distortion during seam carving. In the process of seam carving, each time a certain number of seams were removed, the distortion degree between the original image and the resized image was calculated. Once the distortion reached a certain threshold, seam carving was terminated and switched to scaling the image the image to the required size.Simulation of the proposed retargeting method was done and contrast experiments between the proposed method and other image retargeting algorithms were carried out to measure their performances.