Image mosaic of bionic compound eye imaging system based on image overlap rate prior

Abstract

An image Mosaic algorithm utilizing image overlap rate prior is proposed for bionic compound eye imaging system based on micro-surface fiber faceplate in this paper. Firstly, the two images to be spliced together are both divided into overlapping regions and non-overlapping regions using the prior of relative position and overlap rate of the sub-eye images. Then, Feature points in overlapping regions are extracted using Speeded Up Robust Features (SURF) detector and described by Binary Robust Independent Elementary Features (BRIEF) descriptor. The initial matching of the feature points is made with the hamming distance matching. Random Sample Consensus (RANSAC) and angular consistency of the pairing feature points are used to further purify the feature point pairs. Finally, the weighted mean method is used on the images after registration to get the blended image. The sub-eye images are spliced in each layer and then the spliced images of each layer are stitched together successively to get the final panoramic image. Experimental results showed that the splicing speed of the proposed algorithm is 2 to 3 times higher than that of Scale Invariant Feature Transform (SIFT) algorithm. Compared with SURF algorithm, the splicing speed also increased by about 50%. In addition, more correct matching can be remained, so that the results of image registration and splicing can be more reliable. Thus, the proposed algorithm can promote the images real-time processing of the compound eye imaging system.