**Background** - Ultrasound imaging is sometimes limited due to speckle noise, which is formed by scatterers that are usually much smaller than the wavelength of the ultrasound wave. And since speckles are rather difficult to compress, ultrasound images are less compressible than other medical imaging modalities. **Methods** - We exploit the fact that regions with different speckle patterns do not have the same diagnostic significance and thus can be compressed with adaptive compression ratios. Accordingly, we introduce a method for distinguishing between mostly speckled blocks and mostly non-speckled blocks. Assuming that the speckles are of Fisher-Tippet (FT) distribution, for each block we measure the Kullback-Leibler divergence between its histogram and a fitted FT probability function. Based on the statistical analysis, we develop a classifier that successfully separates between speckled blocks and non-speckled blocks. In this study, we focus on liver ultrasound images and accordingly extend the classifier into three components corresponding to fibrotic, normal and hypoechoic tissues. To extract these components, the Expectation Maximization (EM) algorithm is used. EM, which is an iterative method for estimating the maximum likelihood parameters of the model, is adopted to a FT mixture model by calculating and implementing the appropriate analytical expressions. **Results and Conclusion** – Based on the extraction of the components, each pixel in the image is associated with the most likely component that corresponds with the highest probability. Our results show that the proposed method contributes to both enhanced image quality and its compressibility as needed for efficient transmission and storage.

Powered by Eventact EMS