Title :
No-reference image blur measurement based on wavelet high-frequency domain analysis
Author_Institution :
Department of Applied Mathematics, Zhejiang University of Technology, Hangzhou, China
Abstract :
This paper 1 discusses an objective image blur measurement in no reference based on wavelet high frequency domain analysis. It provides one novel statistical feature extraction for wavelet domain description, which is related with measuring score in a power law through NLREG. Evaluating the model by one carefully built database, it obtains correlation rate at 0.95 as human subjective scores, better than Tong´s method about 5 points.
Keywords :
Databases; Feature extraction; Frequency domain analysis; Training; Wavelet analysis; Wavelet domain; Wavelet transforms; NLREG; blur measurement; feature extraction; no-reference; non-linear model; wavelet;
Conference_Titel :
Information Science and Engineering (ICISE), 2010 2nd International Conference on
Conference_Location :
Hangzhou, China
Print_ISBN :
978-1-4244-7616-9
DOI :
10.1109/ICISE.2010.5690480
