Study And Impementation Of Automatic White Balance Algorithm For Image Processing

With the rapid development of image processing and network technology,the era of network camera is coming.Network camera is an intelligent equipment which combines traditional camera and video technology.As the core module of network camera,image signal processor plays a decisive role in its overall performance.Automatic white balance technology is a key step in the image processing process,which has a great influence on the accurate correction of image.Based on the ARM Cortex M4 platform,this paper studies the automatic white balance technology in the network camera,proposes a new improved algorithm and realizes it.The improved algorithm first USES the white point detection algorithm to judge the white pixels according to the threshold range,and then calculates the mean value of each channel and the number of white points.Then the gray world algorithm is used to get the initial gain.Then the initial gain is adjusted based on the gain polygon and the final gain is obtained.Finally,the new pixel value is calculated according to the Von Kries hypothesis,and the color compensation is carried out.In addition,FPGA development board is adopted in this paper,and the algorithm is realized by combining software and hardware.Among them,the data information of the white pixel points in the image is judged and counted by the hardware design,the gain value is calculated and adjusted by the software,and the overall color correction of the image is finally completed by the hardware.Finally,the paper adopted the Imatest(?)test tools to evaluate image after correction.In addition,the classical algorithm is simulated by MATLAB,and both the classical algorithm and other improved algorithms are evaluated and analyzed subjectively and objectively.Simulation and test results show that the saturation errors of the algorithm are 0.001 and 0.039,respectively,which are smaller than those of the three classical algorithms.At the same time,compared with the results of the network camera algorithm in the past two years,first of all,from the saturation error index,the error of this algorithm is smaller at high color temperature;From the RGB channel mean of the white block,the test data of this algorithm is closer to 255.The test results all meet the design requirements.At present,the network camera with this algorithm has been successfully listed on the market.