The Research On Realization Of The Auto-exposure Algorithm Based On Entropy

Nowdays, more than eighty percent of the amount of information can be got through human vision. Image sensor technology is the most important and basical way to obtain information, which makes it dominant in the word of information. With the continuous improvement of semiconductor technology, CMOS image sensor becomes the basic device in accessing visual information. On account of the fact that different environment and objects give rise to pictures with significant differences, it becomes a burning question in how to control CMOS image sensor exposured automatically and correctly in different environment to get more image information.Through the study of the main factors that affect the exposure and the analysis of the merit and demerit of different auto-exposure algorithms, this paper presents an improved algorithm based on image entropy to adjust exposure time of CMOS image sensor. However, it is important in reducing the complexity of the algorithm to make it possible to be realized on FPGA.In this paper, we optimized the algorithm with formula deformation of image entropy, grouping gray value and piecewise linearization of the log function. The whole optimization will do behavioral level simulation by Matlab to ensure its effectiveness. What's more, the preliminary orientation of exposure time can be given by using the algorithm based on mean gray value, which saves a lot of time for following adjustment by using the algorithm based on entropy and laies a strong foundation for hardware inplementation.This paper describes the auto-exposure module and submodules in terms of their fuctions, strctures and hardware implementation. After finishing fuction simulation by Modelsim and module synthesis by Quartus, programs are burn on FPGA. The feasibility of the algorithm is verified by the logic simulation results and systhesis results, which show that after adding the auto-exposure module, the maximum operating frequency of FPGA reaches 109.15 MHZ, the power consumption increases 800 m W and the scale of circuit increases about 10%.Finally, the effectiveness of the algorithm is verified by comparing the quality index of images before and after auto-exposure. And this paper also compares the egde images and image quality index of two sets of images, which are exposured with entropy algorithm and mean gray value algorithm in different conditions. The comparing results demonstrate the entropy algorithm enhances the image's contract and its image contains more egde points. Furthermore, the image presents more details and informations, and its gray level is richness.