A Novel Approach for Speckle Noise Reduction in Ultrasound Images using Multistage Filtering

The image quality and diagnostic accuracy of medical ultrasound imaging is greatly disrupted by speckle noise from various sources. Here, effective multi-stage filtering approach is introduced to eliminate maximum speckle noise from ultrasound images. In this method, the ultrasound image corrupted by speckle noise is initially subjected to homomorphic filtering to improve contrast. Afterwards, wavelet decomposition is applied to the image using a DAUB-32 wavelet to split it into approximation and detail sub-bands. In the next stage, bilateral filtering is performed on the approximation sub-band, while Bayes filtering is performed on the horizontal and vertical detail sub-bands, and Susan filtering on the diagonal details. Finally, inverse of the homomorphic filtering and wavelet transform are performed to obtain the despeckled image. The experimental results demonstrate the effectiveness of the proposed method in significantly reducing speckle noise while preserving important image features thereby enhancing diagnostic quality over other state-of-the-art methods.