Research On Image-based Autofocusing Method

Autofocusing Technology is the key point of digital photography system,which is called for higher improvement in automatic equipments according to modern science development.The core algorithm of autofocus is the definition function,which evaluates the blurred extent and decides the accuracy and real-time property of focusing.Selecting the focusing window in the filmed object can reduce the number of sampled image data to speed up focusing process.What`s more, the speed and accuracy of autofocus depend on the focus searching strategy to a large extent,the best focusing position is found by the strategy.The content of the research is the main part of the crosswise project Automatic Optics Mearsuring Instrument(AOM1.0) Product Development, AOM1.0 is applied by many circuit board producers. Supported by the Innovation Foundation of Nanchang Hangkong University,the research aims at the autofocusing of circuit mearsurement,and presents innovation and improvement contrast to existing methods,the achievements are as follows:(1)The principle of image wavelet transform and Lipschitz regularity of wavelet coefficients are analyzed ,A novel definition measurement based on correlation of the wavelet transform at adjacent scales is presented,which improves the sharpness function`s performance of unimodality and noise-immunity. It takes advantage of the property that wavelet coefficients magnitude of signals increase as scale increases and wavelet coefficients magnitude of noise decrease with increasing scale, calculates the multiplication of wavelet coefficients at the adjacent scales to construct the evaluation method.It is capable to amplify signals and suppress noise.Compared with other traditional algorithms using Image Data acquired by CCD, the proposed method is found to be more robust and accuracy to estimate the focusing extent.(2) Considering the auto-selected focus window, the Visual Attention Model and Salient Feature Extraction theory are led in.Through the analysis of image texture and salient feature,the window selecting method based on orientation of the most plentiful information region is proposed, which adopts the sampling mearsures fitting for visual characteristic in image blocks, integrated with the entropy method of gray-level co-occurrence matrix,to realize the auto-selecting of interested window.The experiment results shows that the method is efficient to track the details region of the measured object, thus it`s benefit for focusing accuracy.(3)Modified interpolation location method is proposed based on present focus searching strategies, divides the focusing process to three parts:pre-focus, coarse focus,fine focus.Pre-focus stage is capable to test the fitness of initial searching step,in coarse focus stage the ordinary Hill Climbing Searching algorithm is adopted,and in fine focus phase to choose five points around the focusing ranges to locate the optimum focus point by cubic ploynomial interpolation.On the basis of the independently developed hardware platform,combined with the designed autofocus software system,the proposed focusing method is put into experiments to complete real-time test,which is proved to guarantee the autofocusing accuracy.