The ultrasound images contain significant concentration of speckle noise. Speckles denoising is of prime importance in order to make the correct diagnosis. This paper provides an orthognalization based technique with projection in the frequency domain that has proved to be effective in speckles denoising. The performance of proposed scheme has been verified on both simulated and real ultrasound images. A comparative analysis between proposed scheme and benchmark schemes such as Orthogonal regular, Frost and Total Variation Filtering (TVF) is presented in this research. The proposed algorithm starts with segmenting a noise corrupted ultrasound image into small sized blocks in an overlapping manner. These blocks contribute in creation of an overall covariance matrix. Fourier transform of the overall covariance matrix is taken to project the image into frequency domain. A subset of the orthonormal vectors are secured by applying the orthogonalization scheme, and a features matrix is formed that is used to clean the speckle noise. The outcomes show that proposed approach is more effective in speckle denoising as compared to benchmark schemes with reference to Signal to Noise Ratio (SNR), Contrast to Noise Ratio (CNR) and Peak Signal to Noise Ratio (PSNR). The aforementioned parameters are of significant importance in assessing performance of any despeckling scheme.