基于立体舒适度的立体图像感知距离估计

提出了一种基于立体舒适度的双目图像感知距离估计方法.首先,推导双目立体图像的立体深度;然后进行主观实验获取实际人眼感知到的深度,分析其与立体深度间的差异;最后,将立体舒适度作为生理因素,建立感知距离、立体深度与视觉舒适度之间的关联关系,并得出感知距离的计算模型.在IVY公共数据集上进行实验验证,结果表明:当视觉舒适度值较高时,模型预测值的平均绝对值误差与均方根误差较主观评测值分别减少了0.0049、0.0073;而当视觉舒适度值较低时,平均绝对值误差与均方根误差分别减少了0.0721、0.0594.本模型结果更接近人类主观感知到的深度.

Perceptive Distance Estimating Based on Comfort Information for Stereo Images

A novel method of estimating perceptive distance based on human factor for binocular image is proposed.Firstly,the stereo depth of binocular images are deduced.Then,subjective experiments are carried out to obtain the distance of human visual perception,and the difference between the perceptive distance and stereo depth is analyzed.Finally,the correlation between perceptual distance,stereo depth,and visual comfort is established using stereo comfort as a physiological factor,and the computational model for perceptive distance is built.Experiments are performed on a public dataset IVY.The results show that when the value of visual comfort is high,the mean absolute error and the root mean square error of the model prediction value are reduced by 0.0049 and 0.0073 respectively compared with the subjective evaluation value;and when the value of visual comfort is low,the mean absolute error and the root mean square error are reduced by 0.0721,0.0594 respectively.It is shown that the results of the proposed model are closer to the human subjective perceived depth.