Segmentation of melanocytic skin lesions using feature learning and dictionaries
作者：
Highlights：
• We present a feature learning scheme for skin lesion image segmentation.
• Negative Matrix Factorization is used to generate an initial dictionary and feature set.
• A subset of the dictionary atoms is selected to improve compactness and representation.
• Normalized Graph Cuts and the learned features are used to segment the input skin lesion image.
• The method potentially can be reliable based on experiments and method comparisons.
摘要
•We present a feature learning scheme for skin lesion image segmentation.•Negative Matrix Factorization is used to generate an initial dictionary and feature set.•A subset of the dictionary atoms is selected to improve compactness and representation.•Normalized Graph Cuts and the learned features are used to segment the input skin lesion image.•The method potentially can be reliable based on experiments and method comparisons.
论文关键词：Segmentation,Skin lesion,Standard camera,Dictionary learning,Feature learning,Sparse representation
论文评审过程：Received 25 December 2014, Revised 29 January 2016, Accepted 21 February 2016, Available online 2 March 2016, Version of Record 30 March 2016.
论文官网地址：https://doi.org/10.1016/j.eswa.2016.02.044
