ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술대회 Adaptive Self-Similarity based Image Super-Resolution using Non Local Means
Cited 2 time in scopus Download 1 time Share share facebook twitter linkedin kakaostory
저자
최지훈, 박세진, 이대열, 임성창, 최진수, 김종옥
발행일
201412
출처
Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) 2014, pp.1-5
DOI
https://dx.doi.org/10.1109/APSIPA.2014.7041605
협약과제
14MR2500, 클라우드 기반 대용량 실감미디어 제작 기술 개발, 최진수
초록
Self-similarity based SR method on the LF-HF domain basically relies on the assumption that LF patch is highly correlated with HF patch. However, this assumption is content-specific significantly, and we can occasionally observe little correlation between LF and HF patch especially for texture region of natural images. In this paper we propose a new self-similarity based SR method to reflect this observation. There are two differences between the proposed and existing methods. First, HF details of target HR image are recovered by finding the similar patches on the MF domain in case of texture region. Second, the proposed method performs pixel-based reconstruction by adopting the concept of non-local means. Experimental results show that the proposed method can reconstruct more realistic image details.
KSP 제안 키워드
Image super resolution, Pixel-based, natural images, non-local means(NL-means), self-similarity, similar patches, texture region