Statistical Properties in the Visual Decision Process

本研究的目的是通过比较理想观察者与人类观察者的可辨别性来计算统计效率，明确视觉决策过程的统计特性。本研究的结果如下：1我们测量了矢状面三维镜像对称检测的统计效率。这种三维镜像对称是在以观察者为中心的坐标系中。为了测量理想观测器的性能，我们在双眼视差中引入高斯噪声我们测量了立体重叠的随机点表面之间的深度区分的统计效率和并排排列的两个随机点表面之间的深度区分的统计效率。为了测量理想观测器的性能，我们在双眼视差中引入高斯噪声我们测量了两组垂直振荡随机点的运动相位差判别的统计效率。为了测量理想观测器的性能，我们在振荡点的运动相位引入高斯噪声我们测量了两组正交振荡随机点之间运动相位差判别的统计效率。为了测量理想观测器的性能，我们在振荡点的运动相位引入高斯噪声我们测量了从点运动中恢复的两个刚体物体或刚体和非刚体物体的区分的统计效率。为了测量理想观测器的性能，我们在运动点的三维位置上引入高斯噪声。基于这些结果，我们建立了视觉信息处理的路径和层次结构模型。

The purpose of this research was to calculate the statistical efficiency based on the comparison of the discriminability of Ideal Observer with that of human observer, and to make clear statistical properties of the visual dicision process.The results of this research are as follows.1 We measured the statistical efficiency of detection of three-dimensional mirror symmetry about a sagittal plane. This 3-D mirror symmetry was in the observer centered coordinate system. In order to measure the performance of the Ideal Observer, we introduces a Gaussian noise to the binocular disparity.2 We measured both the statistical efficiency of depth-discrimination between stereoscopically overlapped random-dot surfaces and that of depth-discrimination between two random dot surfaces arrayd side by side. In order to measure the performance of the Ideal Observer, we introduces a Gaussian noise to the binocular disparity.3 We measured the statistical efficiency of discrimination of the motion phase difference between two groups of vertically oscillating random dots. In order to measure the performance of the Ideal Observer, we introduces a Gaussian noise to the motion phase of the oscillating dots.4 We measured the statistical efficiency of discrimination of the motion phase difference between two groups of orthgonally oscillating random dots. In order to measure the performance of the Ideal Observer, we introduces a Gaussian noise to the motion phase of the oscillating dots.5 We measured the statistical efficiency of discrimination between two rigid objects or rigid and nonrigid objects which would be recovered from dot motion. In order to measure the performance of the Ideal Observer, we introduces a Gaussian noise to 3-D position of the moving dots.From these results, we modeled pathway and hierarchical Structure of Visual information Processing.