Image Processing Based Auto-focus Technology

Compared with the traditional auto-focus methods, auto-focus method based on image processing does not need the information of the target position, the focal length of the optical system, which directly processes the image information to complete automatic focusing, so it has simple structure and has been widely applied to many fields, such as medicine, military, industry etc. This paper mainly carries on research on two basic technical problems of image focusing: image definition evaluation and focusing searching algorithm, and sets up the hardware platform of automatic focusing system. Theoretical analysis, MATLAB simulation and experiments of automatic focusing system verify that theautomatic focusing system is feasible and effective.Firstly, the paper conducts deep research on automatic focusing technology. From imaging theory of optical system standpoint, this paper discusses operational principle and advantages of image focusing for the limitations of traditional automatic focusing methods and analyzes the purpose and significance of using image focusing.This paper focuses on image definition evaluation algorithm. This paper summarizes several common image definition evaluation functions and proposes the image definition evaluation function based on lifting wavelet transform for the disadvantage traditional wavelet transform has large amount of computation and is difficult to achieve real time. Through MATLAB simulation and experiments of automatic focusing system, this paper analyzes real time, sensitivity and noise immunity. The experimental results show that the evaluation function based on lifting wavelet transform has a small amount of calculation, high sensitivity, and strong anti-noise characteristic and is an optimum algorithm for the target object.This paper studies thefocusing searching algorithm. This paper discusses several common focusing searching algorithms and proposes an algorithm of rapid climbing searching strategy combining roughly focusing and fine focusing for the disadvantage traditional climbing method is easy to fall into local extreme value. The phase of roughly focusing uses Brenner function with a large step size to near the focus position. The phase of fine focusing uses lifting wavelet transform with a small step size to locate the best focus position. By comparing three pictures, the method greatly improves accuracy, speed and noise immunity of focusing searching algorithm.This paper sets up experimental platform of automatic focusing system. For the feature image focusing has a large scale of computing, this paper chooses an automatic focusing hardware platform of TMS320DM6467 T as a core processor and Virtex-4 FPGA chip as a coprocessor, and verifies the algorithms used in automatic focusing system. The experimental results show that the methods used in this paper can meet the requirements of automatic focusing system based on image processing, and the system has a good stability and achieves good focusing effect.