The Application Of Bandelet Transform In Image Denoise And Enhancement Processing

Wavelets is the optima bases in expressing the point singularity objective function. Because two dimensional separable wavelets is gained directly by tensor product of two independent one-dimensional wavelets,the wavelets fails to provide the optima representation of two-dimensional image with the "linear" or "face" singularity,And the effect of image denoising and enforcement with the wavelet transform will fall down.In order to overcome the insufficiency of wavelet transform in processing high dimensional signal, the Bandelet transform which provides a sort of new image representation means based on the image edge is introduced by French scholar pennec,mallat and peyre. The Bandelet bases could realize best sparsity expression for the geometry regular image. Because the researcher constructs the Bandelet bases from the angle of compressed encoding, it is insufficient when the Bandelet bases perform image denoising. Because of the effect of noise coefficient, the direction of the Bandelet bases is incapable to be oriented exactly and the image approximation capability of the Bandelet bases is reduced greatly.The noiseproof adaptive Bandelet bases Based on the Bandelet tansform idea are constructed in order to denoise in this paper. A new class of noiseproof adaptive Bandelet bases (NA-Bandelet bases) is proposed. First, employing soft-threshold function and the threshold obtained following SURE principle, the approximation of the original image in each dyadic square is computed, then the geometrical direction in each dyadic is found according to the principle of the minimal approximation error. secondly,the small dyadic square are combined based on the principle of Minimum Mean Square Error (MMSE), so the directions of NA-Bandelet bases are determined at last.Moreover, a novel denoising method based on the NA -Bandelet bases is presented in this paper. First,the image is decomposed by the NA -Bandelet bases ,then the adaptive thesholding is computed and the thesholding denoising is performed in each NA-Bandelet block. Compared with the denoising methods based on Orthonormal wavelet bases and Bandelet bases, the result of experiment shows that the denoising methods based on the NA -Bandelet bases can reduce noise more effectively with improved edge-preserving ability for the desnoising and despeckling of optical image and SAR image.A image enhancement arithmetic based on the NA -Bandelet bases is presented for the noise image in this paper. Acording to the geometry characters of image, The Bandelet blocks are classified into two sorts: the Bandelet blocks with geometrical directions and the Bandelet blocks without geometrical directions, and the different characters of the two sorts of Bandelet blocks are analyzed in order to distinguish the noise and signal,the weak contrast features and the strong contrast features. Image enhancement is realized by a new strategy that can emphasize features with low contrast and protect the strong contrast features from distortions while restraining noise. Results of experiments show that this method offers significant improved performance in restraining noise and emphasizing features with low contrast over the conventional image enhancement algorithms.The construction of the NA-Bandelet bases is the leading focus of the whole thesis. and it has gained better effect that the NA-Bandelet is applied in the image denoise and enhancement.