基于HVS和四元数的彩色图像质量评价方法

Assessment method for color image quality based on HVS and quaternion

彩色图像质量客观评价为后续图像处理的基础,须与人眼感觉一致,而传统的评价往往与主观视觉相差较大.为此,分析了人类视觉系统特性、四元数基础,解析了彩色图像转化为灰度图像处理等难度,提出了结合人眼视觉特性和四元数的全参考型彩色图像质量评价方法 HVS-QSVD.其通过空间位置函数、局部方差函数、纹理边缘复杂度函数、颜色信息等,构建了相应的评价算法模型.经LIVE(Laboratory for Image and Video Engineering)图像数据库与其他方法比对,结果表明该方法优越,更为符合人眼视觉感知,为高速公路视频事件检测、工业视觉瓷砖分类检测的应用,奠定了基础.

As the base of further image processing,color image quality assessment is assumed to be identical to the subjective assessment of human vision systems（HVS）.However,the traditional assessment methods may result in evaluations quite different from the subjective visual assessment.To solve this problem,this paper analyzed the characteristics of human vision systems and the quaternion,discussed difficulties in converting color images into gray ones and finally proposed a full-reference color image quality assessment method called HVS-QSVD,which integrated the characteristics of human vision systems and the quaternion.Also,this paper constructed evaluations of spatial locations,local variance,the complexity of the texture edge and the color information based on the human vision characteristics.By utilizing the above four evaluations as the elements of quaternions,a model is finally proposed to assessment the quality of color images.Experimental results on LIVE image database and other methods confirm that the proposed model is more effective to get an evaluation result identical to human vision.The proposed assessment model could also be used in applications such as Shanghai-Nanjing expressway video detection and industrial tile classification detection.