Detection and Localization of Seam Carving in Digital Images

Seam carving is a content-aware digital image resizing algorithm. Pixels on a digital image are assigned with energy for later connection into seams. Repeatedly removing seams that are with minimal energy, we can reduce the width (or height) of the image. This way seam carving retains important visual information while resizing an image implies that we can assign lowest energy value to some regions on the image and thus eliminate particular objects deliberately. Consequently, development of a seam carving detection method is extremely important, although challenging. In our previous granted project, we have proposed a novel method, referred to as "patch analysis", for detecting seam-carved images. This method involves some key technology, i.e., image segmentation, cosine similarity, transition probability and support vector machine. Experimental results reveal that our patch analysis method achieves the best seam carving detection accuracy in the current stage. However, this research leaves scope for further work. We therefore apply for a follow-up project here to complete the study as bellow: (1) detection of images with floating percentage of seam carving, (2) detection of images with particular objects protected or removed, (3) detection of images with post-processing after seam carving, (4) localization of the hot regions that frequently passed by carved seams, (5) detection of images with seam insertion, (6) using Hopfield networks to identify the optimal patch type, and (7) using patch analysis to detect other image retargeting algorithm. This research area deserves further attention and we believe we will gain an outstanding achievement in a term of two years.

Project ID:PB10207-1817
External Project ID:NSC102-2221-E182-062