A Deep Learning Method to Identify the Tempered Regions in a Seam-Carved Image

Seam carving is a content-aware algorithm for image resizing and tampering. This algorithm assigns an energy map to an image and removes the seams with low energy from the image. By doing this, seam carving makes it possible to reduce the image size and eliminate specific content from images. The detection of seam carving has lately been an important but challenging area of research. In past work, we had proposed a method that involved dividing the images into "mini-squares", analyzing optimal patch types to recover seams, and detecting seam-carved images by the path transition probability. This method yielded the most accurate detection results among other prevalent techniques, and had been presented in ACM SIGGRAPH conference and also published in Pattern Recognition Letters. We worked on improving the detection of seam-carved images and the location of carved seams ever since. We then found it is difficult to identify the tampered regions in a seam-carved image. The method we used in the current stage selected the best patch type and calculated the transition probability in the surrounding area of each mini-square; however, seams are paths connected across the image. This situation implies that the tampered region can be identified more precisely if we can collect inter-pixel information from a broader area. The idea of deep learning is one of the potential methods that can take this into account. We therefore attempt to use convolutional neural networks over the raw image to mark the tampering location in this project.

Project ID:PB10708-1939
External Project ID:MOST107-2221-E182-070