Image Edge Detection Based On B-spline Wavelet Transform

One of the most basic features of the image is edge, which contains a majority of information. How to improve the accuracy, the integrity and the antinoise ability are classic problems in edge detection. In this paper, the Cubic B-spline wavelet transform and its application in edge detection were studied, and some improvements were made. The main research contents are as follows:1.Several classical edge detection algorithms were studied, and lots of experiments were simulated in the MATLAB to compare the advantages and disadvantages of each edge detection algorithm.2. The basic theory of wavelet transform was studied, and an expansion to two-dimensional wavelet transform was made. Combined with multi-resolution analysis and Mallat algorithm, the implementation method of multi-scale wavelet edge detection algorithms was studied.3. The differences between the B-spline wavelet function and the Gaussian function were studied when they used as a smooth function. By contrast, the Cubic B-spline wavelet function was used as a smooth filter operator in this paper.4. The Problems of local modulus maxima selecting and threshold setting were studied, and improved method was proposed. With the help of the canny operator's non-maxima suppression method and the Contourlet transform adaptive threshold method, the improvements are as follows:firstly, divide the image into several pieces, for every piece, find the average value of local modulus maxima. Secondly, construct a new value by multiplying the average value with a constant, and then add the new value with another constant as each piece's threshold. This method was simulated in the MATLAB, and compared with the canny operator. The experiment results show that this method has better antinoise ability.