Design and Development of a Real Time Vision Enhancement System using Image Fusion - an Algorithmic Approach

Abstract

The images enhanced in different viewpoints can be registered and fused within surveillance systems. Most of the existing systems are not capable of providing a real-time image fusion. This paper embodies the design and application development of a real-time vision enhancement system using real-time image fusion. In feature-based image registration, corners are extracted, and a suitable transformation matrix is derived with which the unregistered frame is transformed. The transformed register- and reference frames are fused with discrete wavelet transform (DWT) based on maximum selection image fusion algorithm. These algorithms are implemented and validated using MATLAB/Simulink. The developed vision enhancement system provides 30 fps, and it is jitter free. The response time of the developed system is 155ms. The execution time of an un-optimized and implemented real-time image fusion algorithm on the DM642 processor stretch to 770ms. A unique experimental setup designed has enabled us to achieve an optimum vision enhancement for security and surveillance applications. The algorithm is further optimized to provide us an average execution time of 740ms. The development of the system is extended to two dissimilar cameras moving in different directions. In the current research, the feasibility and applicability of the real-time vision and fusion systems are explored in various security and surveillance application scenarios.