虽然基于结构信息的图像质量评价方法——结构相似度(SSIM)模型结构简单、评价性能优于峰值信噪比(PSNR)或均方误差(MSE)模型,但SSIM模型不能较好地评价严重模糊的降质图像.文中提出了一种基于梯度的结构相似度(GSSIM)图像质量评价方法,该方法将梯度作为图像的主要结构信息.实验结果表明,GSSIM模型比SSIM和PSNR(MSE)模型更符合人眼视觉系统特性,能较好地评价模糊图像的质量.
Although the SSIM (Structural Similarity) model, an assessment model of image quality based on the structural information, has been proved to be better than the PSNR ( Peak Signal to Noise Ratio) or the MSE (Mean Square Error) model, there still remain some deficiencies in assessing badly blurred images. In order to solve this problem, this paper proposes a gradient-based structural similarity (GSSIM) model that takes the gradient as the main structural information of an image. Experimental results show that the proposed GSSIM model is more consistent with human visual system and can assess the quality of blurred images more precisely than the SSIM and PSNR (MSE) models.