単一画像からのノイズ推定手法の評価と改善

概要

論文の詳細を見る
Estimating the noise level of a digital image is important for many applications such as noise removal, image quality assessment, and so on. For practical use in these applications, the estimation method should be achieved from a single image. Ce et al. developed a single image noise estimation method. In their method, a priori information of noise characteristics was generated by the computer simulation and used in Bayesian MAP. However, they did not consider the inherent characteristics of each camera, depressing the estimation accuracy. Therefore, in this paper, we take into account the inherent characteristics of a camera by using camera specific a priori in Bayesian MAP. Camera specific a priori is constructed from preliminarily measurements of the camera instead of a computer simulation. The experimental results show the improved accuracy of noise estimation.
社団法人映像情報メディア学会の論文
2009-01-01