Texture segmentation for seam carving

This paper analyzes the relationships between texture characteristics of an image and seam-carving (SC) algorithm for image retargeting (resizing). SC is a smart image content aware algorithm. The algorithm, changes the aspect ratio of an image while keeping the "important" objects untouched. Based on a subjective test, we combine between an image's characteristics and a saliency map which provides best results. Furthermore, in order to determine the significance of these results, we use principal component analysis (PCA) statistical method, which identifies those relationships in the subjective test's results. Finally, we present a novel saliency map for SC algorithm. This saliency map is based on a new classification model, which uses statistical connections between neighboring pixels and significantly reduces the level of distortion along the process, providing optimal preservation of important objects within the image.