Edge Aware Geometric Filter for Ultrasound Image Enhancement
TL;DR: The proposed filter requires almost no parameter tuning and provides good quality outputs for synthetic as well as real ultrasound images and is compared with the state-of-the-art speckle reducing filters.
Abstract: Despeckling of ultrasound images is essential for subsequent computational analysis. In this paper, an edge aware geometric filter (GF) is proposed for speckle reduction. The behaviour of conventional GF is approximated using commonly used functions like unit step. These approximations help in identifying the natural relationship between GF and other existing spatially adaptive filters. Subsequently, the modifications in GF framework are proposed to take the advantage of edge characteristics. The proposed filter requires almost no parameter tuning and provides good quality outputs for synthetic as well as real ultrasound images. It is compared with the state-of-the-art speckle reducing filters. Improvements of 10.46% and 42% are noticed in mean square error and figure of merit, respectively.
...read more
Citations
25 citations
Cites background from "Edge Aware Geometric Filter for Ult..."
...Speckle reduction in such images is required to improve the image contrast and diagnostic quality [2]....
[...]
3 citations
Cites background or methods from "Edge Aware Geometric Filter for Ult..."
...DsCNN outputs are compared with the state-of-the-art approaches including SRAD [7], DPAD [8], OBNLM [12], SBF [14], EAGF [6] and DnCNN [17]....
[...]
...For example, EAGF and OBNLM give good output for the image shown in the third column, however, for other two images, result in oversmooth and low-contrast outputs....
[...]
...The first set of values is obtained from the original articles of the approaches and the second set is obtained from the recent works [1], [3], [6]....
[...]
...Other wellknown approaches use weighted local and non-local averaging, for example optimized Bayesian non-local mean (OBNLM) filter [12], speckle reducing bilateral filter [13], squeeze box filter [14], and edge aware geometric filtering (EAGF) [6]....
[...]
...Four approaches, DPAD, OBNLM, ADMSS, and EAGF are selected for this purpose....
[...]
References
30,333 citations
2,093 citations
1,959 citations
1,672 citations
1,043 citations