一种基于像素域与变换域联合估计的JND改进模型

Improved pixel-based JND model with the hybrid approach

建立一种基于联合估计的像素域JND(just noticeable distortion)改进模型.对纹理区域采用DCT方法的JND模型能够方便地引入CSF函数,使得纹理区域的对比度掩蔽估计更精确;对于平坦区域、高亮度或者较暗区域则使用基于像素域直接计算的JND模型,充分利用了亮度自适应因素.仿真结果表明,与目前其他几种像素域JND模型相比,在相同的主观质量下,本文方法可以进一步减少图像的视觉冗余.

In this paper, an algorithm is studied for estimating just - noticeable difference （JND） in the pixel domain with explicit image formulation. Through experiments we find that the pixel - based JND model can make full use of the background luminance adaptation while the subband - based JND model can accurately estimate the JND in the texture regions. The proposed profile is designed based on a hybrid method which can take the advantages of two types of JND models. The dominant effect of texture masking or luminance adaptation in different regions is selected adaptively with different JND models, thus the accurate JND in pixel domain is obtained. Experimental results show that the pro-posed algorithm can reduce more data redundancy than the relevant existing JND estimators.