An optimal wavelet filter for despeckling echocardiograph images

Abstract

The major difficulty concerning echocardiograph image processing is speckle noise. In this paper, we present wavelet based algorithm for denoising echocardiograph images and then determine an optimal wavelet filter that yields good quality images. The log transformed input is represented in multiscale wavelet domain. Global hard thresholding of subband coefficients is performed to achieve lossy compression of image. The inverse wavelet transformed image is free from speckle noise. The image quality is assessed in terms of objective and statistical measures, namely, peak-signal-to-noise ratio (PSNR), compression ratio, variance, mean square error (MSE). The experiments are performed using different wavelets at various levels of decomposition. The results show that bior6.8 with 2- level decomposition yields better quality despeckled images. © 2007 IEEE.