Studies On Quality Evaluation And Window Construction Algorithms In Auto-Focusing Technology

Auto-focusing technology is an indispensable tool to obtain clear images for the modern high-speed information society. Focusing technology has developed from manual mode to automatic mode, and in the automatic mode it has developed from active mode to passive mode. In recent year, there has been an auto-focusing based on image processing as the digital information development, and it contains the depth from defocus (DFD) and the depth from focus (DFF). Auto-focusing can get high-resolution images with less manual intervention and it can track and detect moving targets. It can also overcome the disadvantages of manual mode which is difficult to operate, low efficiency and unclear.The definition evaluation function is the core part of auto-focusing technique, and the focusing curve which has better performance must be unimodality, unbiasedness, high sensitivity and anti-noise. For digital image auto-focusing, the basic theory is to find the differences between clear and blur images, and to show them by the methods of image processing. The commonly used types of differences can be summarized as the information, the gradient edges and the frequency components. The classical evaluation functions can be divided into three categories: one is the gradient functions, the two is the frequency functions and the three is the statistical and information functions. At present, the main algorithms usually use only one aspect of difference to construct the evaluation function. The sensitivity of the traditional evaluation functions decline with the noise, which leads to the focusing results inaccurately.The construction of the focusing window is an important part of the auto-focusing technology, which affects the accuracy and real-time of the focusing results. The traditional focusing window algorithms contain the fixed methods such as the center window and the multi-point window, and the dynamic methods such as the Gauss window and the first moment window. The paper introduces the principle and process of the commonly used search strategies, and proposes a new improved auto-focusing technique which gives the specific steps. Finally, the paper introduces the process of the auto-focusing. The focusing search strategy can connect the whole process to ensure the evaluation function and the search algorithm could complement each other to achieve the auto-focusing process.The main innovations of this paper are as follows:(1) In the evaluation function, a new function is proposed based on the combination of the gradient difference and the statistical correlation. It can remove the interference of additive noise without correlation and the threshold is set to remove these pixels whose contributions are small, which can improve the real-time performance and anti-noise. The main part is the gradient function to make full use of the advantages of space function that is simple to realize, which to meet the real-time requirements. On the other hand, when the sequences of pictures contain noise, the cross-correlation function can use the natural link of these pixels to distinguish the blur and clear images, which overcomes the disadvantage that the spatial domain functions are sensitive to the noise.(2) In the focusing region selection, the gray level first order moment method uses the absolute gradient difference to locate the prospect target, a new method based on region contrast is proposed to make up for the shortcomings. When the window is constructed, it is adaptive and can track the subject object better. The method calculates the relative value of gray difference in each region, and then combines with the first order moment to determine the image center of gravity position. The algorithm can eliminate the influence of the background region of high brightness. The experiments use different types of images to test these functions and the results show that the proposed function can make accurate positions whether the target is in or off the center of the image.(3) There are simulation tests for the improved algorithms and other classical algorithms using some image sequences. Except for the qualitative indexes, the experiments use some quantitative indexes to evaluate the performances of these algorithms. The result shows that the proposed algorithm has higher resolution ratio and sensitivity factor and lower fluctuation than other functions, which shows that it has higher anti-noise and stability.