Image quality enhancement method based on scene category classification and its evaluation
Masato Tsukada, Chisato Funayama, Masato Toda, R. Oami, Masatoshi Arai, Akira Inoue · 2009
This paper presents a method of enhancing image quality based on the analysis of scene categories, and its practical software implementation. The proposed method analyzes the scene of input images and calculates the probabilities for five predetermined scene categories; i.e. “Portraits,” “Landscapes,” “Night scenes,” “Flowers (close-up)” and “Others.” The quality of the input image is improved by using multiple image-processing functions with correction parameters, which take the probabilities of the scene categories into consideration. Subjective experiments on image quality show the reliability of the proposed method.