Research On CMOS Image Preprocessing Method And Implementation On FPGA

CMOS image sensor will introduce various kinds of noise in the process of image acquisition,which seriously affects the visual effect and subsequent image processing.Noise is usually eliminated by image processing.Image preprocessing technology generally includes image filtering,edge extraction and other technologies.Firstly,the principle of CMOS sensor imaging is introduced,and the noise introduced in its working process is analyzed,in this paper.Then,the image smoothing filtering method and edge detection method in image preprocessing are explained,and the shortcomings are pointed out.According to the characteristics of CMOS image noise,the traditional algorithm is improved.The methods of removing bad points from CMOS images,image filtering based on edge detection,image enhancement based on gray transformation and edge detection based on human visual characteristics are proposed,in this paper.In view of the improved algorithm,MATLAB is used to simulate and verify the performance of the algorithm.In order to meet the real-time requirements of image preprocessing,this paper uses the FPGA to realize the improved image preprocessing method,which is modularized and pipelined.Finally,the full reference image evaluation method is introduced,and the experimental results are evaluated.Experiments show that the proposed method can effectively improve the quality of CMOS images.After processing,the peak signal-to-noise ratio(PSNR)of the image is increased by 1.47 dB,and the accurate position information of the image edge is obtained from the final binary image.So it has certain reference value for CMOS image preprocessing design with high real-time requirement.