This article presents a de-speckling technique for clinical ultrasound images with an aim to preserve the fine structural information and region boundaries in images. The algorithm generates restored images by minimizing the variational energy on them. To compute variational energy, a weighted total variation based method is proposed where the weights are determined from both historical (previous/earlier time stamp) as well as instantaneous oriented structural information of images. This helps in defining the anistropy at edges in the image which, in turn, helps in identifying homogenous regions on it. Moreover, the method is able to preserve the vague echo-textural differences which might be of clinical importance but may get destroyed due to smoothing operations. To elicit effectiveness, comparative analysis of the proposed approaches have been done with four state-of-the-art techniques on both in silico and in vivo ultrasound images using four standard measures (two for phantom images and two for clinical ultrasound images). Qualitative and quantitative analysis reveals the promising performance of the proposed technique.
|Title of host publication||Mining Intelligence and Knowledge Exploration - 5th International Conference, MIKE 2017, Proceedings|
|Editors||Rajarshi Pal, Ashish Ghosh, Rajendra Prasath|
|Number of pages||12|
|Publication status||Published - 2017|
|Event||5th International Conference on Mining Intelligence and Knowledge Exploration, MIKE 2017 - Hyderabad, India|
Duration: 2017 Dec 13 → 2017 Dec 15
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Other||5th International Conference on Mining Intelligence and Knowledge Exploration, MIKE 2017|
|Period||17/12/13 → 17/12/15|
Bibliographical notePublisher Copyright:
© 2017, Springer International Publishing AG.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)