基于视觉心理学的物体大小恒常性计算

知觉恒常性是人类感知世界最重要、最突出的方面,它为解决计算机图像理解和物体识别等经典计算机视觉难题提供了新的思路.大小恒常性是最重要的知觉恒常性之一.正确的图像物体大小恒常性感知的关键在于准确计算物体在图像中的感知深度.本文总结了人眼使用的各种图像深度线索,提出了这些线索融合与冲突的解决方案,然后用数学方法建立了图像物体大小恒常性计算模型.实验结果表明该模型是有效的.本文是应用视觉心理学来解决计算机视觉问题的一次成功而有益的探索.

Object Size Constancy Computation Based on Visual Psychology

Perceptual constancy is the most salient aspect of the way the world looks to us. It may enlighten some new ideas for many open issues of computer vision, such as automatic image understanding and image object recognition. Size constancy is one of the most important perceptual constancy. The key to correct size constancy perception of image object is to correctly estimate the depth of objects in the image. This paper introduces most of image depth cues used by human vision, and presents a method to resolve the fusion and confliction of these cues. Then we propose a model to implement size constancy computation of image objects. Experimental results suggest that our model can work reasonably well. In addition,this paper sets a good example for applying visual psychology to resolving computer vision problems.