低位深图像质量的客观评价

Objective Evaluation for the Quality of Deep-Reduced Image

通过实验指出现有PSNR公式对低位深图像处理效果的评价与人类视觉感受有较明显的差错,利用相关视觉原理分析其原因,并在此基础上提出改进方案:将原来的逐点计算改为逐块比较,并补充灰度差与均值的偏离度分析。新公式对图像的描述较好的与视觉的空间积分机制建立了联系,无论是不同方法处理同一幅图像还是相同方法处理不同的图像,新公式对低位深图像的客观评价与人类主观视觉感受都有着较好的一致性。

Method PSNR shows its incapability in evaluating the quality of enhanced deep-reduced image by comparison and analysis. An improvement about formula PSNR which based on the difference from formula itself and visual principle is proposed. By changing the compare region, and adding a new reference about the deviation between gray difference and average, the new formula makes connection with space integration and it has more comprehensive description on images quality and it is much more coincident with the objective evaluation and subjective perception.