An Image Mosaic Algorithm Based on Improved ORB

Due to the lack of scale invariance in ORB (Oriented FAST and Rotated BRIEF) algorithm, image mosaic in complex scenes, high mismatch rate, and easy to produce mosaic seams in synthetic images, etc. In this paper, an improved ORB algorithm for image mosaic based on BRISK feature extraction is proposed. Firstly, the FAST9-16 of BRISK algorithm is introduced to detect corner points and solve the scale invariance problem. Then ORB algorithm is used to describe feature points, and Hamming distance is used to roughly match image features. Then KNN algorithm and PROSAC (Progressive Sample Consensus) algorithm are used to refine and purify the matching points to further improve the accuracy. Finally, image seamless fusion and mosaic are completed by gradual-in-gradual-out weighted average fusion. The experimental results show that the proposed algorithm has good performance in image scaling, rotation and illumination intensity. It is a real-time image mosaic method with high accuracy and good mosaic effect.