FPGA Based Novel Architecture for Real-time Video Stitching

This paper presents a novel architecture for stitching video samples in a field programmable gate array (FPGA). Video streams of two distinct camera sensors of the same specifications are taken as input. Feature points in the video frames are uniquely identified based on Scale Invariant Feature Transform (SIFT). With the feature points as a reference, a stitching algorithm is devised and the stitched output is transmitted to a VGA display. An FPGA architecture for video frame capture, SIFT feature detection, video frame stitching, and output display driver is designed in Verilog. The designed architecture is verified by simulation and implemented in FPGA. A comprehensive analysis of the design performance is also done and is proved to be within real-time performance requirements.