一种噪声近似图像的质量区分方法

文中针对图像间质量差别微小而需评价的难题,提出了一种新的基于偏微分方程放大模型的全参考图像质量客观评价方法.算法利用偏微分方程在图像放大过程中能够较好地保持图像的边缘特征和细节信息的优点,通过比较放大后的图像加权峰值信噪比变化的情况,得出全参考图像质量评价指标-M_HPSNR.实验结果为类似的需要区分质量近似图像的需求提供了技术参考.

A Method to Differentiate Quality of Noise Approximation Image

Based on the difficulty of the evaluation on the tiny difference between image quality （ quality approxima- tion）, this paper proposes a full reference Image Quality Assessment evaluation method, which is based on the Partial Differential Equation. The algorithm bases on the advantage that the Partial Differential E- quation keeps good image edge character and the detail information through the process. It respectively enlarges the reference image in proportion to the evaluation image. In this way, it increases the quality distance between image quality approximations. Through the amplification image and the M_HPSNR, it can be concluded the full reference image quality evaluation index M_HPSNR. The experimental result shows that the evaluation of tiny quality difference is the same with the predicted one. Meanwhile, it applies the technical reference for the need of differentiating the image approximation.