Blind tone-mapped image quality assessment based on clustering perception

Abstract

In order to display a high dynamic range (HDR) image on a standard monitor, tone-mapping operators (TMOs) aim to compress HDR images into low dynamic range tone-mapped (TM) images. To accurately evaluate the performance of different TMOs, this paper proposes a no-reference image quality assessment (IQA) method for TM images. Firstly, the image is divided into dark area, middle area and bright area by using clustering algorithm. The entropy and area ratio features are extracted from three areas mentioned above and the saliency area that is detected by the proposed method. Then the natural scene statistics features of the luminance channel and RGB color channels of TMI are used to assess the luminance naturalness and chrominance naturalness, respectively. Finally the support vector regression module is utilized to yield a quality score of the TM images. The experimental results on the tone-mapped image database (TMID) show the effectiveness of the proposed algorithm. Compared with the existing representative IQA methods, the proposed method has better performance.