Main Article Content
Ultrasonic image is one of the imaging techniques that widely used and safe for medical diagnostic, due to its noninvasive, low cost and real time forming. However, the qualities of ultrasonic images are typically degraded mainly due to the presence of signal independent known as speckle noise. In this paper, speckle noise suppression in wavelet domain and feature extraction technique is studied. In particular, speckle noise reduction is a preprocessing step before applying a feature extraction process. In speckle noise suppression process, the logarithmic transform is firstly applied to the original image in order to convert multiplicative noise to additive one. 2D Stationary Wavelet Transform (SWT) is used to decompose the logarithmic image into four subbands. Next, 2D adaptive Wiener Filter is applied all over areas only in detail subbands. Finally, the 2D Inverse SWT is computed and it is followed by the exponential transform to get the reconstructed image. To evaluate the studied method for speckle noise reduction, some classical well-know methods, such as Median filter, Wiener filter, Discrete Wavelet Transform (DWT) based on Soft thresholding and DWT along with Wiener filter are compared. For feature extraction process, Haar wavelet filter is used to extract the ultrasonic features compared with Sobel and Canny operator. Moreover, the nonmaxima suppression technique is adopted to get the edge localization. Finally, the hysteresis thresholding is applied to get the final result in binary format. The results have clearly demonstrated that the studied method outperforms several existing methods for speckle noise reduction in terms of signal to mse ratio (S/mse) and edge preservation (β). Moreover, the studied method can detect well-localized and thin edges.