Research On Image Reconstruction And Color Correction Technology Of Image Processing

Since the 90s of the 20th century, with the very-large scale integrated circuits (VLSI) high-speed development, CMOS image sensors also get a good momentum of development. Research on image processing algorithms and study of SoC design are also developing rapidly. The next few years, CMOS sensors will extend occupation ratio of consumer digital camera market. Extensive use of CMOS image sensors will bring a long period of prosperity of image signal processing (ISP) key technology, the image reconstruction and color correction technology in this paper is the key technology of digital image signal processing, which focus on the color processing.This color reconstruction and correction system includes three kinds of image processing: auto white balance (AWB), color interpolation (CI) and auto color correction (ACC). Color correction and color interpolation have relatively fixed position, generally after the interpolation of Bayer format image, then color correction. After interpolation, the amount of data will be 3 times as before. In order to save hardware, in this paper, the automatic white balance processing is before the color interpolation; on the other hand, the proposed color correction method requires the image must be white balanced. Based on these considerations, the order of the modules in this system is automatic white balance, color interpolation, and finally auto color correction.The main innovation of this paper is the color correction method based on the Lab color space. Discuss the reasons of color shift, and also the common problems of the color correction. Based on the Lab color space, using mathematics methods, such as derivative, solving mean square error minimization problem, correction matrix is derived. The color correction can be directly operated on the RGB image. Hardware design can meet the requirement of real-time image processing. Then Auto white balance, color interpolation and color correction modules are integrated together, implemented on Altera's FPGA (Cyclone family EP1C12Q240C8). PSNR of interpolated G component up to 45.8463, after color correction, average reduction of MSE is about 15%. While the color system is based on the Lab color space, the image is more suitable to human visual characteristics.