Research And Application Of Image Denoising And Edge Detection Algorithm Based On Wavelet Transform

With the continuous development of computer technology,image processing technology is more and more widely used in various industries.Noise is interference information attached to the image that affects the expression of the image content,the edge represents the area where the gray level of the image changes greatly,which is the basis for further interpretation of the image.Therefore,filtering out image noise and detecting image edges are extremely important researches content in image processing.Wavelet transform is an emerging signal processing tool in recent years,it has low entropy and good time-frequency characteristics,it is called "mathematical microscope" and has a wide range of applications in image processing.Based on wavelet theory,this paper improves wavelet threshold denoising algorithm and edge detection algorithm based on wavelet transform,and applies the improved algorithm to rail defect detection.The main work of the thesis is as follows:The paper researches the wavelet transform theory,introduces the concept and principle of multi-resolution analysis,uses Matlab to simulate the Mallat fast algorithm,and analyzes the principle and properties of one-dimensional wavelet transform and two-dimensional wavelet transform.The categories and evaluation criteria of image noise are discussed,the commonly used spatial domain filtering algorithm and wavelet domain denoising algorithm are summarized and their principles are analyzed.Aiming at the "pseudo-Gibbs" phenomenon in the hard threshold denoising function and the discontinuity at the threshold value of soft threshold denoising function,an improved wavelet threshold algorithm is proposed,the improved threshold function is better in smoothness,continuity and graduality.By comparing with the traditional wavelet threshold denoising algorithm,the effectiveness of the improved algorithm is proved.The principle of classical edge detection differential operator and edge detection algorithm based on wavelet transform is analyzed,and their advantages and disadvantages are compared.An improved edge detection algorithm combining wavelet transform and mathematical morphology is proposed.The improved algorithm designed a new anti-noise morphological edge detection operator,and proposed an adaptive weighting algorithm based on cosine distance to detect the edge in different directions.Comparing improved algorithm with other commonly used edge detection algorithms,the results show that the improved algorithm has better edge detection ability and better detection effect on noising images.Taking two kinds of rail defects(cracks and scars)as the research object,the improved denoising and edge detection algorithms are used in the process of extracting defects,the experimental results show that the improved algorithm has practicability to a certain extent.Finally,the BP neural network is designed to classify the two kinds of defects,the classification results show that the BP neural network designed by the paper is classified reasonably and meets the requirements of use.