Theoretical Study Of Auto-focus Algorithm Based On Image Processing

The auto-focusing method based on image is of the focusing method of opticalimaging system based on image processing technology. This focusing method couldautomatically achieve the focusing process in the optical imaging system, throughanalyzing information of grabbing images of the optical imaging system and evaluatingfocus degree of current imaging system. Because of no additional distance measurementequipment in focusing process, the automatic focusing method can effectively improveimaging quality of the optical imaging system, and has been widely used in the focusingoccasions with quick and frequent operation, such as videotape, automation production,real-time monitoring, machine vision, infrared detector.Automatic focus methods based on image are classified into DFF (Depth fromFocus) and DFD (Depth from Defocus). In order to improve the focusing efficiency inmore degree, this paper analyzes the principle and implementation of such twoautomatic focusing methods, develops a rapid automatic focusing method combinedDFF and DFD. The proposed automatic focusing method could set the searching steplength through a new estimating algorithm of defocusing amount of current focusingposition in the process of searching peak value of the image evaluation function curveto confirm the best imaging position, and finish the focusing process in a high speed.The more work and achievements in this paper include:①Proposes a rapid estimating algorithm about calculating the defocusing amountof current focusing position based on DFD. This estimating algorithm estimate the ratioof two dispersion parameters (the radius of blur) in two defocus images through theratio of two pixel values that meet some constraints in such two defocus images, andthen could give the estimated value of the defocusing amount immediately with theanalytic relation between defocusing amount and dispersion parameters in the defocusimages which would be used to set the searching step length in the focusing process.The experimental results show that the estimated value of defocusing amount indifferent focusing position has a good reference value for searching step setting.②In order to ensure the correctness of search direction judgment in the focusingprocess, avoid bad influence of the image noise, this paper design a zone dividingmethod in focusing route. This method firstly divided the focusing route into two zones:the one is a zone far from the focal plane position; the other is a zone nearby the focal plane position, then sentence which zones current position is in by the estimated valueof the defocusing amount of current focusing position. If current focusing position is inthe zone far from the focal plane position, the searching direction judgment shouldincrease the searching step length to conquer the noise influence; if current focusingposition is in the zone nearby the focal plane position, the searching direction judgmentwould do its self ignoring the noise influence.③In order to improve the efficiency of automatic focusing process, this paperdesigns a method of setting step length in the focusing process. This method adjust thestep length in real time to achieve an adaptively setting of searching step lengthaccording to a proper proportion of the estimated value of the defocusing amount ofcurrent focusing position. The experimental results show that this step-length settingmethod can reduce the times of grabbing images in focusing process with a properscale factor, and improve the focusing speed3~5times above the method with samestep length.④Compared two choosing methods of focusing window in process of automaticfocusing, and provided some suggestions for window choosing in different focusingenvironment, and designed a window choosing scheme for our microscopic visiontesting system.⑤Designs a automatic focusing method under the focus environment with morefocal plane and more targets existing. It first divide image area into more areas throughthe contour line of each target in image, then obtain the hierarchical relationships ofeach focal plane position corresponding the each area through estimating defocusingamount of each target in those areas, and combining with the inherent relative positionsrelation of the targets in the world coordinate system, complete the parallelism of thetargets in the world coordinate system and the areas in image, finally realize focusingprocess with choice of any target.