An ultrasound image enhancement method using local gradient based fuzzy similarity

• A novel noise suppression method based on fuzzy logic rules is proposed.
• Using two sets of fuzzy rules each pixel is categorized by its local gradients.
• Fuzzy similarity measures determine the non-local weighted restoring framework.
• The results are investigated using simulated and clinical medical ultrasound data.

In this paper, a fuzzy rule based filter is proposed for speckle reduction in Ultrasound (US) images. Considering a relevant US noise model this filter uses local gradients of the image and fuzzy inference to categorize image regions regarding their characteristics due to noise and structural information. Then in the restoration step each pixel is restored using fuzzy similarity criteria to weight its similar neighborhood pixels. Quantitative results on synthetic data show the performance of the proposed method compared to state-of-the-art methods. Results on real clinical images demonstrate that the proposed method is able to preserve accurately edges and structural details. As an application, this filter is used as a preprocessing step for a well-established US segmentation method known as Disk Expansion (DE). The results show an improved true diagnosis of lesion in both simulated and clinical US images.