Studies On Definition Evaluation And Window Construction Algorithms In Auto-focusing System

Auto-focusing is one of the key technologies for all kinds of optical imaging systems to obtain clear,high quality images,is an important foundation for the subsequent image analysis and processing.In the fast-growing information society,auto-focusing technology covers various fields of production and life,from common digital cameras in our daily life,to the space remote sensing system,biomedical imaging field,etc.All of them are inseparable from the auto-focusing technology.Auto-focusing has experienced the development from the early active mode to the passive mode.The auto-focusing method based on digital image processing technology has the advantages of high accuracy,high integration,strong space and time applicability.so it has been widely used and developed rapidly.Without too much manual intervention,auto-focusing system can identify the clear picture.It solves many problems in manual focusing,such as operation difficulty,low efficiency,and poor imaging quality and so on.It provides a guarantee for the subsequent image analysis,processing and application of various imaging systems.Depth from focus(DFF)which based on image processing technology is a more obvious method.It has the advantages of high precision,low system complexity and good versatility.The process of DFF is as follows:first,construct focusing window to a series of collected images with different degrees of fuzziness;then,calculate the image sharpness evaluation values in the window;according to the evaluation results,the driving motor moves the position of camera lens based on the established search strategy,until the best picture is found.In this paper.the auto-focusing algorithm is based on DFF.The definition evaluation function is the core of the focusing process.The purpose of evaluation is to judge the difference between the focus and the defocus image,and provide the basis for the following search steps.Window construction is the key to the target orientation in auto-focusing.By analyzing and processing the region of interest in the window,it can effectively improve the real-time performance and accuracy of auto-focusing.In this paper,we mainly study the evaluation function and the focusing window construction,propose a gradient threshold evaluation function based on Kirsch operator and a window construction method based on saliency feature to improve the performance of evaluation function and the accuracy of window construction.In addition,according to the complete process of DFF,by choosing the proposed evaluation function and the proposed window construction method in this paper,and selecting an appropriate focusing search strategy,we propose an improved auto-focusing strategy,and give the specific implementation steps.The main innovations of this paper are as follows:(1)Aiming at the problem that the traditional gradient evaluation function is sensitive to noise and can not guarantee the stability of the evaluation results,a gradient threshold evaluation function based on Kirsch operator is proposed.In order to reduce the amount of computation and ensure the real-time performance,the Kirsch gradient is approximated by the sum of absolute values;a new threshold is proposed to remove the small gradient values in the image,which effectively increases the difference between the blurred image and the clear image;the gradient value of the isolated noise point in the image is suppressed,so the noise immunity of the evaluation function is improved.In this paper,the evaluation function has obvious advantages in terms of accuracy,anti noise stability and other performance.(2)In the aspect of focusing window construction,a window construction method based on saliency feature is proposed.According to the human visual system's attention degree to the salient region of the image,the saliency information of the image is innovatively selected as the basis for the construction of the focusing window.The LC algorithm is used to construct the saliency map,which has low computational complexity.Image resolution is reduced by calculating the sum of the salient values in the image partition domain.In the same time,it increases the effect of some regions with a larger salient value.Then,achieve the binary image by setting the threshold to separate the target and the background.Finally,the "barycenter" of the segmented image is calculated,mapping the rectangular window to the original image,we will get the focusing window.The window construction method of this paper has high accuracy and low computational cost for the target object,can ensure the real time performance of auto-focusing.(3)There are simulation experiments for various types of evaluation functions.Except for the subjective qualitative evaluation indexes,the objective quantitative evaluation indexes are also added to evaluate the performance of evaluation function.The detailed data can reflect the performance of the evaluation function more objectively and more comprehensively.In this paper,the quantitative evaluation result which shows that the proposed function has higher noise immunity and stability.