Abstract: Noise distortions introduced in natural images generally break the initial probability distributions by dispersing image pixels randomly. It finds that there exists a big difference between the distributions of Discrete Wavelet Transform（DWT） coefficients of natural images and noisy images: for natural images, their distributions are sharp with highpeakedness and slight tail; for noisy images, the shapes are much flatter with lower peakedness and heavier tail. Kurtosisis able to measure and differentiate the probability distributions of noisy images with various noise levels. Moreover, the kurtosis values of DWT coefficients are stable for varyingfrequency filters. This paper proposes a Blind NoisyImage Quality Assessment model using Kurtosis（BNIQAK）. Five types of noisy images in the three biggest databases are taken for testing BNIQAK. Experimental results show that BNIQAK has better evaluation performance compared with existing blind noisy models, as well as some general blindand Full-Reference（FR） methods.

Key words: blind noisy image quality assessment, discrete wavelet transform, kurtosis

摘要： 随机散布在自然图像里的噪声失真一般会破坏图像的原始概率密度分布。研究发现，无失真自然图像和它对应的噪声图像在离散小波变换（Discrete Wavelet Transform，DWT）系数分布上有很大区别：对于自然图像，其DWT系数分布比较尖锐，峰值高，拖尾短；对于噪声图像，其系数分布则比较扁平，峰值低，拖尾长。作为一种常用的统计特征描述，峰态值可以度量和区分不同失真程度的噪声图像的DWT系数分布，而且，DWT系数分布的峰态值具有很好的频率尺度不变性。基于以上特性，提出了一种无参考噪声图像质量评价模型（Blind Noisy Image Quality Assessment model using Kurtosis，BNIQAK）。实验测试了三个最大的质量评价图像库中的五种噪声失真图像，结果表明，和现有无参考噪声评价模型、一般无参考评价模型和全参考（Full-Reference，FR）评价模型相比，BNIQAK具有更好的评价效果。

关键词: 无参考噪声图像质量评价, 离散小波变换, 峰态值