A novel algorithm based on wavelet-trilateral filter for de-noising medical ultrasound images

Abstract

Speckle noise causes a sequence of serious problems in medical ultrasound imaging system and limits the development of automatic diagnosis technology. In the paper, a novel approach, based on wavelet transformation and trilateral filter, was proposed. The clinical images are optimized by the integrated filter. Firstly, a dynamically varying additive model is developed to account for medical ultrasound signal with speckle noise. Secondly, an adaptive wavelet shrinkage algorithm, in accordance with the statistical property of the additive model, is developed to apply to the noisy medical signal, especially with the component of high frequency in wavelet domain. Thirdly and most importantly, the low frequency component of speckle noise is suppressed by a trilateral filter that simultaneously reduces the speckle and extreme impulse noise in real set data. Finally, plenty of experiments are taken on synthetic images. And what's more, for real clinical ultrasound images, a lot of experimental studies of the proposed method are conducted. Compared with other existing method, experimental results show that the proposed approach demonstrates an admirably de-noising performance and offers great flexibility while substantially sharpening the desirable edge.