基于有序NPsim矩阵的颜色近邻搜索

doi:10.3964/j.issn.1000-0593(2018)02-0377-09

Abstract
References
Related Citation (15)

Download: PDF (2633 KB)
Export: BibTeX | EndNote (RIS)

Abstract The core of color nearest neighbor search based on spectral representation is high dimensional nearest neighbor searching, which is affected by equidistance of similarity measurement and low query efficiency of indexed tree seriously. To solve this problem, a color nearest neighbor search algorithm based on sequential NPsim matrix was proposed. First of all, the NPsim values between every two colors in color space were calculated and stored into the corresponding position of the NPsim matrix. After that, the elements in each raw of NPsim matrix were sorted in descending order to represent the similarity degree from strong to weak. Finally, the Top-K neighbors of given color can be found directly according to its raw number in the sequential NPsim matrix. To validate this algorithm, the nearest neighbor search algorithms based on KD-tree or SR-tree were compared on Munsell spectral data set. Experimental results indicated that the precise was better than that of other algorithms, the construction time of sequential NPsim matrix was longer than that of them, but the searching speed was more than thousands times of others and independent of K value. In addition, the construction of sequential NPsim matrix was easy to be paralleled to speed up, but the parallelization of construction of KD-tree or SR-tree was difficult.

Key words：Color; Nearest neighbor search; Spectrum; Similarity; Indexing tree; Sequential NPsim matrix; Munsell

Received: 2017-03-18 Accepted: 2017-09-14

ZTFLH:

O433.4

Corresponding Authors: WANG Gong-ming E-mail: gongmingwang@126.com

Cite this article:

URL:

https://www.gpxygpfx.com/EN/10.3964/j.issn.1000-0593(2018)02-0377-09 OR https://www.gpxygpfx.com/EN/Y2018/V38/I02/377