基于深度学习的无参考立体图像质量评价 - 计算机辅助设计与图形学学报.PDFVIP

第28 卷 第 6 期 计算机辅助设计与图形学学报 Vol. 28 No.6 2016 年6 月 Journal of Computer-Aided Design Computer Graphics Jun. 2016 基于深度学习的无参考立体图像质量评价 * 田维军, 邵 枫 , 蒋刚毅, 郁 梅 (宁波大学信息科学与工程学院 宁波 315211) (shaofeng@) : 立体图像质量评价是评价立体视频系统性能的有效途径, 而如何结合人类的视觉特性对立体图像质量进行 评价是目前的研究难点. 为此提出一种基于深度学习的无参考立体图像质量评价方法, 分为训练和测试2 个阶段. 在 训练阶段, 首先对左右图像分别进行Gabor 滤波, 获取不同尺度和方向的统计特征作为单目特性; 然后根据人眼视觉 系统的双目竞争特性, 将左右图像融合得到独眼图, 提取其方向梯度直方图作为双目特征; 最后通过深度信念网络 训练得到特征和主观评价值之间的回归模型. 在测试阶段, 根据已建立的回归模型, 预测得到左右图像质量并联合 得到立体图像质量. 实验结果表明, 文中方法在对称和非对称立体图像数据库都取得了较好的效果, 与人类的主观 感知保持良好的一致性. ：立体图像; 质量评价; Gabor 滤波; 独眼图; 深度学习 ：TN919.81 Blind Image Quality Assessment for Stereoscopic Images via Deep Learning * Tian Weijun, Shao Feng , Jiang Gangyi, and Yu Mei (Faculty of Information Science and Engineering, Ningbo University, Ningbo 315211) Abstract: Stereoscopic image quality assessment is an effective way to evaluate the performance of stereoscopic video systems, but how to utilize human visual characteristics effectively is still a research focus in stereoscopic image quality assessment. In this paper, a blind image quality assessment method for stereoscopic images is pro- posed via deep learning. The proposed method is composed of two stages: training and testing. In the training stage, Gabor filter is applied to the left and right distorted images respectively, and natural statistical features un- der different scales and directions are extracted to act as monocular features. Then, left and right images are fused to construct a cyclopean map, and histograms of oriented gradient features are extra