An Optimized Process of Multiscale Retinex Algorithm Based on Hardware System

This paper introduces an optimized process of Multiscale Retinex (MSR) algorithm that apply to the hardware system. Retinex theory is a constant color model derived from the Human Visual System (HVS) and can obtain more stable result for image enhancement than classic algorithms. Generally, the MSR model in Retinex is more suitable for enhancing images on hardware since it has less computation than other models. On the basis of ensuring the stability of the MSR algorithm results, we further optimize its calculation process. And the key issue in the restructuring model process is how to use the simplest Gaussian Convolution Kernel (GCK) to accomplish the Multiscale Gaussian Convolution (MSGC) assignment. Moreover, we briefly introduce the video display and the algorithm processing mechanisms in the hardware application, which includes a sophisticated Image Processing IP (IP-IP) based on the rearranged calculation process we designed. Under the real-time condition, the hardware system based on upgraded MSR can enhance smoothly color image and embrace a perfect performance in terms of visual intuition.