Medical Image Seamlessly Stitching by SIFT and GIST

Image stitching are ideally suited for applications such as image stabilization, summarization, and the creation of panoramic mosaics. This paper proposes a approach to stitch images fast and with high quality.Image alignment algorithms can discover the correspondence relationships among images with varying degrees of overlap. In this process, they uses SIFT to look for the trait points, and uses RANSAC to delete the wrong points. Image stitching algorithms take the alignment estimates produced by such registration algorithms and blend the images in a seamless manner by a kind searching method, which can deal with potential problems such as blurring or ghosting caused by parallax and scene movement as well as varying image exposures. Several real alignment images stitching experiments are carried out. The experimental results show that the algorithm can improve stitching visual quality obviously with acceptable computing cost.