A Study On An Image Edge Detector Nased On B-spline Wavelet

The important scenery structure is embodied by the image border which carries the most important characteristic information of an image. The border detecting is main be used in image processings machine vision, and principles of pattern recognition is still one of classical technology not solved. Therefore, searching for a kind of better edge detection method which will apply to most images and have good anti-noise ability have became a urgent solution to this problem.There are several main tasks in this paper:(1) Several classical edge detection algorithms are analyzed and compared, and finally get their advantages and disadvantages by the simulation results.(2) According to the theory of wavelet transform, B-spline wavelet function is elaborated in this paper. Also comparing it with the Gaussian smoothing function, we prove this algorithm is superior to Gaussian function in anti-noise ability. Then the cubic B-spline function is simulated.(3) The images containing different noise have been detected those edges by cubic B-spline algorithm experiment, also some simulation about testing anti-noise ability have been implemented. Finally, the result images and data both proved our method is more superior to canny algorithm.(4) The cubic B-spline algorithm applied in the images containing smaller, middle and larger details and finally compared with canny algorithm.(5) We put forward a new way of using the edge detection based on cubic B-spline wavelet to apply in images fusion, and compared with the classical wavelet fusion method, we have got objective conclusion to prove its feasibility.The theories analysis and tests result have shown that:The template coefficient of cubic B-spline wavelet algorithm is brief which has reduced lots of calculation. For the images with different noise, the result about anti-noise ability is superior to canny algorithm in both the subjective vision and the objective data; To those containing different details' images, this method is more applicable to small and middle details' images, and the result is better than canny algorithm ;Cubic B-spline edge detection method applied in images fusion can embody edge details more better and make outline more clear. But to some images with more edge details, because of this method is too sensitive to edge, it will produce more edge details in the fusion pictures. So this problem needs to be improved in after studying.During lots of testings, this algorithm has shown better anti-noise ability, but that algorithm also has its own limitation, has different effect with handling different pictures. So according to different characteristic property in different pictures we need to choose correctly edge detecting algorithm, even to improve the algorithm in some degree.