Ultrasound image despeckling based on local binary pattern and local homogeneity

Abstract

Speckle pattern is a form of multiplicative noise which blurs the ultrasound images. It reduces the contrast and resolution of ultrasound images which results in poor interpretation of image features. Hence speckle reduction is an important preprocessing step in many image processing tasks such as segmentation, classification and pattern recognition. Bilateral Filter has been proven to be very effective in denoising as it makes use of spatial averaging with out smoothing edges. The bilateral filter reappropriate the pixel intensity with a weighted sum of the pixels in its local neighborhood. The weights are computed based on intensity similarity and spatial proximity. The bilateral filter blurs the image structure when the intensity variations of the underlying images are very poor [1], [2], [3]. This work is an attempt to improve the structure preserving capability of bilateral filter by incorporating local structure of images such as local homogeneity and local binary pattern in computing weights.