A novel Non-local means image denoising method based on grey theory. (January 2016)

• Record Type: Journal Article
• Authors: Li, Hongjun; Suen, Ching Y.
• Is Part Of: Pattern recognition. Volume 49(2016:Jan.)
• Page Start: 237
• Page End: 248
• Publication Date: 2016-01
• Subjects: Image denoising -- Non-local means -- Grey theory -- Weight function; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2015.05.028
• Languages: English
• ISSNs: 0031-3203

Abstract

Abstract: In this paper, a novel Non-local means image denoising method, called Grey theory applied in Non-local Means (GNLM) is proposed. Different from previous works, our method is based on grey theory. The advantage of grey theory is its flexibility in handling complex scenes. The grey relational analysis needs fewer testing samples, but can achieve a better performance. Therefore, we analyze the structure similarity by grey relation of coefficients, set similar weight function accordingly, and propose an efficient Non-local means. This new method solves the parameter setting problem encountered by traditional Non-local means methods and reduces the computational complexity. Experimental results validate our proposed method. It removes noise well and is efficient in capturing details, especially edges and corners. This leads to a state-of-the-art denoising performance. The performance is equivalent and sometimes surpasses recently published leading alternative denoising methods. Highlights: New method solves the parameter setting problem encountered by traditional methods. The grey relational analysis needs fewer testing samples, but can achieve a better performance. Efficient in capturing details, especially edges and corners.