Abstract
This study presents an effective approach based on Schur decomposition in frequency domain for removing speckle noise from ultrasound images. The proposed scheme has been tested on both simulated and real ultrasound images, and is compared with different benchmark schemes including the Schur Regular, PNLM and Lee. The theme of proposed approach is to segment a speckle noise corrupted image into various overlapping blocks of a small size, and generate the global covariance matrix by averaging the covariances of individual blocks. The Fourier transform of global covariance matrix is taken and Schur decomposition is applied to generate eigenvectors. These frequency domain orthogonal vectors are arranged in a descending order, and a subset is used to create the Feature matrix, which is used for speckle noise removal. The proposed approach shows better performance than benchmark techniques in terms of both Peak Signal to Noise Ratio (PSNR) and Signal to Noise Ratio (SNR), which are regarded as key parameters in the despeckling area.