Improved Algorithm Based Optimal Stitching Line in Video Stitching Quality
Xuemei Bai, Yunpeng Wang, Chenjie Zhang, Hanping Hu · 2024
For the optimisation of video splicing quality in video splicing, an improved optimal stitching video splicing method is proposed. The method first extracts corner points and feature points using the improved FAST algorithm, and accurately matches these feature points by combining violent matching with GMS feature point matching; adopts a graph cutbased algorithm to transform the stitch line search problem into an energy minimisation problem to ensure the reasonableness and optimality of the stitch line position; and constructs the solution step by step in overlapping regions by means of a dynamic planning algorithm, retaining the optimal solution at each step, and finally find the globally optimal suture position. To further enhance the suture quality, the method smoothes the sutures by morphological operations (e.g., swelling and erosion) and implements fine pixel-level adjustments to eliminate artefacts and chromatic aberrations. After determining the optimal overlap region and suture line location, an advanced image fusion algorithm is used to ensure that the transition of video frames in the overlap region is natural and seamless, avoiding splice marks. The experimental results show that the method can effectively improve the quality of video stitching.