Research And Design Of Image Noise Reduction And Compression Pre-Processing Circuits For CMOS Image Sensors

The noise in the CMOS image sensor reduces the image’s quality tremendouslyand affects the sensor’s imagery performance. Therefore, reducing noise is the keytechnology which must be solved in the CMOS image sensor. On purpose ofeliminating noise, this paper adopts digital image processing techniques.This article first reviewed the CMOS Image Sensor course of development,followed by analysis of a CMOS image sensor in various types of noise source; and todiscuss the classification and the noise model. According to the noise of differenttypes of different algorithms, denoising better-performing, the main issue is to bridgethe digital image in the Salt and Pepper noise and Huaidian and can be integrated toachieve the objective, so here is based on an average of four areas of law.First, MATLAB is used to verify this algorithm, then the hardware descriptionlanguage Verilog is adopted to describe the algorithm and the function simulation ispassed in the end. By the comparison to the primitive image, the image after addingnoise and the image after eliminating noise, it can be directly observed that thealgorithm is useful to eliminate noise, but also keep the image’s detail.When image data is processed in real time, it must be compressed before becauseit is huge. When Bayer image is compressed directly, image recovery-quality is notgood. It finds that high quality image recovery-quality could be got if imagepretreatment had been done at a certain compression rate.The research background of the characteristic of color CMOS image sensor andBayer image, a near-lossless compression method for Bayer image is introduced. Andit is tested in Model Sim6.0, at last, hardware circuit is created in the environment ofQuartus Ⅱ6.0. Image data are sampling down and filtered in the method. The wholearchitecture was optimized in the pipeline design, fewer line buffers instead of largememory to get higher hardware utilization and speed. When tested, it is simple andefficient. This method can get high recovery-quality image under near-losslesscompression, so it can create good condition for coding and transmission. So thismethod suits to medical or remote sensing image and other high-quality image.