Enhanced Video Mosaic Generation: Efficient ORB Feature Extraction and Hamming Distance Matching

Abstract

Video mosaicing is a technique for creating expansive panoramic views by seamlessly combining multiple video frames. This paper introduces an innovative algorithm for video mosaic generation, emphasizing the alignment and blending of non-overlapping frames. The proposed algorithm utilizes the ORB feature extraction technique to effectively address challenges related to camera motion and content variations across frames, ensuring a smooth and continuous mosaic output. Key aspects of the algorithm include optimizing the number of feature matches for frame stitching, which critically impacts the final mosaic's quality. Experimental results demonstrate the algorithm's ability to produce visually coherent and high-quality video mosaics. The model has been tested on real-time video frames, showing improved performance of 95% of accuracy with lower RMSE and contributing to the advancement of video mosaicing techniques for applications in surveillance and other fields requiring comprehensive scene views. © 2024 IEEE.