一种无参考型模糊图像质量评价方法

A No-Reference Quality Assessment Method for Blurred Image

图像的边缘信息是人眼观察和识别物体的重要特征,根据模糊图像相对于清晰图像其边缘特征发生较大变化的特点,提出了一种基于边缘锐度的无参考模糊图像质量评价方法。首先,通过文中所示方法寻找图像中的所有阶跃边缘;其次,根据一些原则选择合适的部分边缘;最终,计算这些合适边缘的锐利程度作为图像的模糊度评价依据。实验结果表明,该方法相比于全参考模型SSIM能够更好地评价高斯模糊、离焦模糊等模糊类型图像,与主观评价结构相关性强,更符合人眼视觉系统特性,并且易于实现。

The edge imformation of images is an important feature for the human eyes to observe and recognise objects. According to the feature that the characteristics of the edges of blurred image will change more sharpnessly compared to the ed- ges of clear image, a no-reference quality assessment method for blurred image based on the edge sharpness is presented in this paper. Firstly, search all step edges of the image with the approach which is presented in the paper. In addition, accord- ing to some principles the proper step edges are chosen. Finally, the sharpness of these appropriate edges are computed as the metric to access blurred images. Experimental results show that this method is able to access blurred images better than SSIM such as gauss fuzzy, defocus fuzzy and so on, and it has strong correlation with subjective evaluation method, more consistent with human visual system and easy to realize.