基于QDCT和贝叶斯谱熵的盲彩色图像质量评价

图像质量评价算法在评价彩色图像质量时,往往 会因损失色彩信息或者破坏彩色图像结构的整体性, 而使得评价结果与人眼观测结果不一致。由于图像越模糊其频谱的高频分量分布越不均匀, 基于四元数离 散余弦变换(QDCT)和贝叶斯谱熵,提出了一种无参考模糊彩色图像质量评价算法。首先,利 用四元数矩阵 对彩色图像进行表示并分解成不重叠的8×8 pixel的子矩阵。其次 ,对每一个子矩阵进行QDCT后得到子 矩阵的频谱。最后,通过贝叶斯谱熵衡量频谱中高频分量分布的均匀性,实现彩色图像质量 评价。所提算 法在数据库IVC、TID2013和CSIQ上的PLCC和SRCC值均达到了 0.920以上。实 验数据表明,本文算法的评价结果准确性好、性能稳定且与人的主观观测结果的一致性高。

Blind color image quality assessment based on QDCT and Bayes spectral entropy

Image quality assessment algorithms usually lose color information or destroy the structure integrity of image,during the process of assessment,which makes the evaluation results inconsistent with the results of human observation.A no-reference blur red color image quality assessment algorithm is proposed based on QDCT and Bayes spectral entrop y,because the more blurred a image is,the more unevenly distributed these high frequency components are.Firstly,a quaternion matrix is used to represent a color image before it i s divided into non-overlapping sub-matrices of 8×8pixels.Then,the spectrum of each sub-m atrix can be obtained by QDCT.Finally,in order to evaluate the quality of a color image,Bay es spectral entropy is calculated to measure the uniformity of high frequency components in the spectrum. The PLCC and SRCC values of the proposed algorithm on database IVC,TID2013and CSIQ all reach above 0.9202.Experimental data shows that the assessment resu lts of the algorithm in this paper are accurate,stable,and consistent with human subjective observations.