Summary
International Technical Conference on Circuits/Systems, Computers and Communications
2016
Session Number:M1-5
Session:
Number:M1-5-3
Adaptive Multi-scale Self-similarity-based Enhancement Algorithm for Effective Up-scaling of Thermal Infrared Images
Yong-Jun Kim, Byung Cheol Song ,
pp.79-80
Publication Date:2016/7/10
Online ISSN:2188-5079
DOI:10.34385/proc.61.M1-5-3
PDF download (1.1MB)
Summary:
Infrared (IR) images usually have blurred edges and weak details in comparison with visible light images. Due to such phenomenon, up-scaled IR images do not provide acceptable visual quality, either. So, we propose an edge enhancement algorithm as a pre-processing for effective up-scaling of IR images. The proposed algorithm utilizes the block-based self-similarity which indicates local similarity between a current image and its scaled ones. First, for each block in the input IR image its edge strength is computed. Then, high-frequency information proper for the computed edge strength is derived. Here, the best self-example of the current block is found from a down-scale determined by the computed edge strength. By adding the extracted high-frequency information to the current block, the definition of the current block is enhanced. Experimental results show that the proposed algorithm provides better edge enhancement than state-of-the-art algorithms. Also, the proposed algorithm outperforms the other algorithm up to 0.07 even in terms of a quantitative metric, i.e., just noticeable blur metric (JNBM).