An efficient technique for speckle noise reduction in ultrasound images

Ultrasound images (US) are very prone to speckle noise, which degrades their quality and make doctor's evaluation difficult. This paper presents method that suppresses the speckle and preserves the edges. The proposed method is based on combining the results of two filters: one filter is an edge filter and another is a smoothing filter. Filters applied are Detail Preserving Anisotropic Diffusion (DPAD) and Optimized Bayesian Non-Local Mean (OBNLM). The outputs of the filters are combined with the help of homogeneity map (HM). HM differentiates homogeneous and edge region of the image. The experiments are conducted using a field II generated synthetic image and a real ultrasound image. The results are obtained for proposed method and compared with other methods based on the parameters such as Peak Signal to Noise Ratio (PSNR) and Edge Keeping Index (EKI).