The Research Of Image Denoising And Edge Detection Based On Wavelet Transformation

Edge detection and denoising are important image preprocessing technologies, which have been used widely in many application fields, such as extraction of image contour, character and texture analysis. Wavelet analysis possesses good localization features in time and frequency, as well as multiresolution character, which has been adopted widely in image processing applications. This thesis deeply investigated wavelet analysis theory and its applications in the fields of image edge detection and image noise removing.In this thesis, the principle and implementation approach of signal and image edge detection based on wavelet analysis theory were firstly studied in depth, where the relationship between Lipschitz index and signal edge was mainly discussed. Construction of wavelet edge detector was probed next, followed by research on edge extraction method using modulus maximum value of wavelet coefficients. Motivated by the redundant wavelet transform, a new image edge detection method based on un-downsampling wavelet transform and context-filtering was proposed. The experimental results demonstrated the proposed method could be validity in detecting the edge of image spurred by the stronger additive noise.Finally, the traditional wavelet denoising methods based on the hard threshold and soft threshold were investigated. Aiming to the defects of discontinuity and non-adaptive of the classical threshold functions, three novel threshold functions, one with the characters of continuity and high-order derivative, another with local context adaptive and the other with local context adaptive and the characters of continuity and high-order derivative, were explored and used in the image denoising. The experimental results indicated the validity of these novel threshold functions in improving denoising performance.