Research On Digital Processing Technology Of Low-light Image Based On FPGA

With the development of the low-light level (LLL) night vision technique, the object's brightness can be enhanced more than one hundred thousand times, which effectively compensates for the human visual system in low-light environments. However the features of LLL image with low signal-to-noise ratio and low-contrast seriously affect the visual effect of the LLL image. Therefore, the noise characteristics and the gray level distribution of LLL image are both analyzed. Combining with the digital image processing technique, the quality of LLL image is greatly improved. At the same time, a LLL image digital processing system based on FPGA is designed. The experiment proves that the system can improve the video visual effects successfully.This paper focus on the research on the digital image processing technique for LLL image. Firstly, a denoising algorithm base on Principal Component Analysis and an improved Non-local Means image processing algorithm on the basis of previous digital image processing algorithm are proposed. Secondly, an image enhancing algorithm in mixing space is proposed. Finally, an adaptive Gauss-channel pseudo-color processing algorithm based on the human visual characteristics is introduced. These algorithms are all evaluated subjectively and objectively by software simulation.In order to design the LLL image digital processing system, Xilinx FPGA has been selected as the control chip of the system, in which a recursive frame-accumulated denoising module, an image detail enhancing module based on the Gaussian filter, and a color mapping module based on the "look-up table" are designed. With these three modules, the real-time video processing system based on FPGA is realized successfully.