视觉感知的数字图像拼接痕迹可见度客观评价方法

数字图像拼接时,待拼接图像之间的亮度差异导致拼接图像出现拼接痕迹。为了对数字图像拼接痕迹的明显程度进行客观评价,研究了一种基于人眼视觉感知的数字图像拼接痕迹可见度(MAV)客观评价方法:首先,由待拼接图像获取梯度误差图像;然后,综合考虑人眼视觉频率敏感度、亮度掩模效应及纹理掩模效应对视觉分辨力的影响,导出基准图像的临界可见偏差(JND),并从梯度误差图像的小波系数中减去JND门限值,获得能够被人眼视觉感知的拼接痕迹图像;最后,利用拼接痕迹图像的均值和信息熵,构造了一种数字图像MAV的客观评价指标(MAVA).实验表明,采用MAVA对数字图像拼接痕迹的明显程度进行的客观评价,与人眼主观评价结果基本一致。

An Objective Assessment Method of Digital Image Mosaic Artifacts Visibility Based on Visual Perception

The difference of illumination among images to be mosaicked causes mosaic artifacts while the digital images are being mosaicked together. In order to objectively evaluate the visibility degree of mosaic artifacts in digital images, an objective assessment method of digital image mosaic artifacts visibility based on human visual perception was proposed . The processes of the method are as follows: 1) the gradient errors image is obtained according to the images to be mosaicked; 2) the just noticeable difference (JND) of the reference image is derived by considering the impacts from not only the human visual frequency sensitivity but also the brightness mask effects and the texture mask effects on visual resolution, then, the mosaic artifacts image which is visually perceptible can be acquired by subtracting the JND threshold values from the wavelet coefficients of gradient errors image; 3) the mosaic artifacts visibility (MAV) of digital image is customized with the aid of the average value as well as the information entropy of the mosaic artifacts image, and finally regarded as an objective index for assessing the image stitching seam visibility. The experiment indicates that the MAV related objective assessment result of digital image mosaic artifacts visibility is basically consistent with the one of the subjective perceptual method.