Reassembling fragmented images is a useful technique to seize digital image evidence. Graph-based methods are one of the most popular for reassembly. However most of them use some garbage data to mistakenly calculate weights between fragment pairs which contain insufficient image data. To overcome this problem, we take special consideration on the last fragment for each reassembling image, in which the insufficient case is likely to happen. If the case comes up, then just the residual pixels are employed to compute the weight. On the images from PASCAL VOC 2010 challenge data set and USC-SIPI image database, experimental results show the effectiveness of our methods.
展开▼
