Contrast Enhancement Evaluation Database (CEED2016)

The CEED2016 is newly developed image database dedicated to contrast enhancement evaluation. The database contains 30 original color images and 180 enhanced images obtained using six different CE methods. The database is built with our own captured images and some common pictures used by the image processing community. The subjective experiments were performed at Universite Paris 13, Sorbonne Paris Cité at Laboratoire de Traitement et Transport de l’Information (L2TI). The images were displayed on a calibrated LCD monitor in a dark room environment to avoid any problem with the illumination adaptation of background. Twenty-three observers, 10 experts, and 13 non-experts, from different age groups, gender, and background participated in the experiments. To obtain the ranking scores, we adopted a balanced pairwise preference based ranking protocol. The interface for the subjective experiments was developed in Matlab, where, for each original image, we randomly displayed all possible pair combinations of enhanced images to the observers. We also showed the original image in the center of the screen (a pair of enhanced images are displayed to its left and right), to facilitate the analysis of after effects of CE. The observers had the choice to rank equally similar stimuli. In the PC ranking protocol, each enhanced image is compared with the others in pairs and ranking results are stored in a preference matrix.CEED2016是专为对比度增强评估开发的新型图像数据库。该数据库包含30幅原始彩色图像，以及通过6种不同对比度增强（Contrast Enhancement，简称CE）方法生成的180幅增强图像。该数据库的图像来源包括自主拍摄的图像，以及图像处理学界常用的部分公开图像。主观实验在巴黎第十三大学、索邦巴黎西岱大学的信息处理与传输实验室（Laboratoire de Traitement et Transport de l’Information，简称L2TI）内开展。实验于暗室环境中进行，图像通过经过校准的LCD显示器呈现，以规避背景光照适配引发的各类问题。共有23名来自不同年龄组、性别及学术背景的观察者参与本次实验，其中10名为领域专家，13名为非专家观察者。为获取图像的排序评分，我们采用了基于均衡成对偏好的排序协议。本次主观实验的交互界面基于Matlab开发：针对每幅原始图像，我们会向观察者随机呈现其所有增强图像的成对组合；同时将原始图像显示在屏幕中央，左右两侧分别展示一对增强图像，以便分析对比度增强后的视觉后效。观察者可对相似度较高的视觉刺激物进行同等排名。在该成对排序协议中，每幅增强图像均会与其余增强图像两两配对进行比较，最终的排序结果将存储至偏好矩阵中。