摘要： 根据人眼对图像的背景比细节具有更大的感觉容量的特点，提出一种图像质量评价算法。将原始图像和被测试图像进行小波变换，其低频系数的误差总和反映图像背景的改变程度，而2幅图像的拉普拉斯均方误差(LMSE)定义图像细节部分的降质，并将两者相乘的结果用于评价图像质量。实验结果表明，该算法优于广泛应用的PSNR和LMSE等传统模型，其评价结果与HVS判断结果相符，可以评价不同的图像处理过程。

关键词: 图像质量评价, 感觉容量, 拉普拉斯均方误差, 小波变换

Abstract: Based on different human sense capacity to background and detail, an algorithm of image quality assessment is proposed. Change of background is weighed by error summation of low frequency coefficients of wavelet transform of the original and the test images. Change of detail is defined by Laplacian Mean Square Error(LMSE), and quality of image is valued by combination of the two factors. Experiments indicate that it performs better than the widely used PSNR, LMSE, etc, and it is consistent with HVS, applicable to various image processing applications.

Key words: image quality assessment, sense capacity, Laplacian Mean Square Error(LMSE), wavelet transform

中图分类号: 

• TP391