基于边缘相似度的全参考图像质量评价算法

客观图像质量评价(IQA)方法旨在利用计算模型自动实现与主观感知一致的图像质量评估,是当前计算机视觉领域和图像处理领域的研究热点.由于图像的边缘信息对理解图像语义信息至关重要,提出一种基于边缘信息的图像质量客观评价方法.通过比较参考图像和失真图像的多方向边缘强度、边缘强度对比度及亮度信息等特征的相似性,实现图像质量的客观评价.利用国外的LIVE2、TID2008及CSIQ等多个图形数据库计算结果证明,本算法能实现对多种失真类型的图像的质量预测,并且在预测准确性、鲁棒性和主观评价一致性方面与当前大多数方法相比都有明显优势.

Full-Reference Image Quality Assessment Based on Edge Similarity

The objective image quality assessment (IQA) devotes to automatically evaluate the image quality consistent with subjective perceptions by using computational models,and has been a research hotspot in the field of computer vision and image processing.Because of the importance of image edge semantic information,a new IQA method based on edge information is proposed,which considers the similarity of multi-direction edge intensity,edge intensity contrast and luminance information between the reference image and the distorted image.The extensive experiments performed on three benchmark IQA databases (LIVE2,TID2008 and CSIQ) demonstrate that the proposed method not only can realize the quality prediction of various distortion types of images,but also can achieve much higher accuracy,stronger robustness and higher consistency with the subjective evaluations than mostly state-of-the-art IQA metrics.