Studies On Passive Auto-Focusing Algorithms

Due to the complexity of imaging targets and conditions, there are still many shortcomings in auto-focusing techniques used in digital imaging systems at present. Many problems, such as how to select focus measures and focus windows, how to perform feedback control, how to improve the auto-focusing velocity, and how to reduce manufacture cost etc., are far from completely solved.Given these problems, this paper fulfilled a great deal of studies on the crucial technologies of passive auto-focusing algorithms, and attained achievements in these aspects: a. Put forward pre-processing projects to focus images, which mainly includes eliminating the influence of light and removing the ordinary noises and effectively improves the qualities of focus images, b. Carry through imitation and contrast researches on various focusing evaluation functions, and the research results indicate that Brenner gradient evaluation function is able to focus rapidly and has a good focusing effect. It is applied to the PC-based auto-focusing system research platform, c. Bring forward amelioration to golden section point based focus windows selection plan, which acquires perfect focusing effect to scenery images and therefore is used in PC-based auto-focusing system research platform as one of the three focus windows selection plans. d. Pattern recognition methods are ushered into the layout of focus windows, present a method of skin-detect based focus windows selection plan, this plan is suitable to screen the character images and can acquire very good focusing effects. Skin-detect based focus windows selection plan is also used in the PC-based auto-focusing system research platform as one of the three focus windows selection plans, e. Aiming at the problems of traditional hill-climbing searching (HCS) algorithm, put forward an optimized HCS model. There are twice searches in all to fix on the best imaging position, firstly cursorily searches in a wild bound with inconstant step length, secondly elaborately searches in a narrow bound with the minimal step length. The optimized HCS model is ushered into PC-based auto-focusing system research platform and is able to acquire a rapid focusing speed and an accurate focusing effect. f. Frequency domain band-pass filtering method is as follows: Video signals from CCD image sensor pass through a band-pass analog filter to pick up high frequencies, which are used to construct a simple focus evaluation function. Auto-focusing systems based on frequency domain band-pass filtering method are simply calculated and inexpensively cost. This paper does researches on frequency domain band-pass filtering method by digital emulation. Sampled digital images are filtered line-by-line to attain high frequencies in computer. Proper evaluation function is constructed with high frequencies and suitable filter frequencies are ascertained, thereby the theory foundations and parameter supports are provided for realizing the equivalent hardware circuit and designing auto-focusing systems based on frequency domain hand-pass filtering method. g. Obtain high frequencies by 2D filtering to a frame of image, accordingly traditional frequency domain band-pass filtering method is enlarged to 2D space by digital emulation. The focusing effect of 2D frequency domain band-pass filtering method is discussed.This dissertation builds a PC-based auto-focusing system on the bases of a great deal of studies on all sorts of key problems of auto-focusing algorithms. In this system, video signals are sampled by digital sampling card and delivered into PC computer, computer analyses the degrees of defocus, confirms the control plans and sends out control orders to single-chip. The single-chip will control the lens to be in focus by moving the step motor. According to the studies before, the evaluation function of the system chooses Brenner gradient evaluation function. There are three modes of focus windows to be chose: center selection method, improved golden section point selection method, skin-detect method. Besides, Optimized HCS algonthm is used as searching algorithm. Focus experiment results indicate that this auto-focusing system is able to obtain perfect focusing effect and can finish once focus by 1.3s.