De Noising Digital Images using Modified Ridgelet Transform

Abstract

The work aims at designing Ridgelet transform based denoising newlinealgorithms using Radon projections. The Ridgelet coefficients are newlineaveraged to achieve denoising. While wavelet and ridgelet transforms newlinehave recently received a lot of attention for their usage in denoising, their newlineemployment in pattern recognition is still in its infancy. These two newlinesignificant issues are the focus of this thesis. Perform multiwavelet newlineorthonormal shell expansion on the contour to obtain the average and a newlinenumber of resolution levels. newlineThe denoising is performed using one dimensional Discrete Wavelet newlinetransform (DWT) algorithm for varying levels of White Gaussian Noise newline(WGN) and Speckle noise for variety of images. We initially investigate newlinemultiwavelet thresholding in the denoising field by adding nearby newlinecoefficients. newlineAccording to experimental findings, this method outperforms newlineneighbouring single wavelet denoising for various common test signals newlineand real-world photos. Then, using neighbouring coefficients, we provide newlinea thresholding method for wavelet images. newlineAccording to experimental findings, VisuShrink and the TI denoising newlineapproach created by Yu et al. are inferior to translation invariant (TI) newlinedenoising with neighbour dependency. newlineThe work also deals with denoising using cycle spinning techniques with newlinewavelet shrinkage. The analysis is done using Mean square error and newlinePeak signal to noise ratio of the original image, noisy image and newlinereconstructed image with and without cycle spinning algorithm