基于结构相似度的无参考遥感图像质量评价

发布时间：2018-04-22 04:26

  本文选题：遥感图像 + 无参考质量评价　； 参考：《科学技术与工程》2017年25期

【摘要】：在深入研究遥感图像与普通图像差别的基础上,对结构相似度算法进行改进,提出了一种新的无参考遥感图像质量评价方法。首先,采用非下采样轮廓波变换进行多尺度分解,将子图像进行K-means边缘检测及细化融合,最终得到边缘区域。采用一阶偏导有限差分计算遥感图像的梯度幅值,设置两个阈值,将满足区间的像素提取出来得到纹理区域。然后,使用边缘计算对比度分量,纹理计算结构相似度分量,对SSIM进行改进得到ET_SSIM方法。最后,采用"再降质"的方式构造出参考图像,将边缘纹理区域与平滑区域分开评价,ET_SSIM对边缘纹理区域评价,SSIM对平滑区域评价,根据人眼对边缘纹理及平滑区域的不同重视程度加权求平均,得到最终结果 VSSIM。实验结果证明,本文方法 VSSIM与主观评价值的线性相关度相对于MSE、PSNR、SSIM、GSSIM、BLIINDS-Ⅱ方法分别提高了22.2%,6.2%,0.8%,0.2%,1.3%。
[Abstract]:On the basis of deeply studying the difference between remote sensing image and common image, the structure similarity algorithm is improved, and a new method for evaluating the quality of remote sensing image without reference is proposed. Firstly, the non-downsampling contour wave transform is used to decompose the sub-image to K-means edge detection and thinning fusion, and finally the edge region is obtained. The gradient amplitude of remote sensing image is calculated by the first order partial derivative finite-difference method and two thresholds are set to extract the pixels satisfying the interval to obtain the texture region. Then, the contrast component is computed by edge and the structural similarity component is calculated by texture, and the ET_SSIM method is obtained by improving SSIM. Finally, the reference image is constructed by "re-degradation", and the edge texture area is evaluated separately from the smooth area. The final result of VSSIMM is obtained by weighted averaging according to the different emphasis on edge texture and smooth region. The experimental results show that the linear correlation between the method VSSIM and subjective evaluation value is higher than that of MSE PSNRN / SSIMI / GSSIMI / BLIINDS- 鈪，

本文编号：1785604