仿生物视觉的非均匀采样方法及其硬件设计  

Non-Uniform Sampling and Hardware Design of the Bionic Vision Method

作　　者：张斌[1] 梅魁志[1] 郭青[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所,西安710049

出　　处：《西安交通大学学报》2010年第6期99-103,共5页Journal of Xi'an Jiaotong University

基　　金：国家自然科学青年基金资助项目(60905007);国家高技术研究发展计划资助项目(2009AA01Z307)

摘　　要：利用生物的选择注意机制可减少计算量的优势,提出了仿生物视觉的非均匀采样方法并实现了硬件设计.利用基于高斯函数的模型模拟了生物视觉的选择性响应特性,由此得到采样层优先和采样密度优先的2种非均匀采样表达式.在视觉处理芯片中,根据人眼特性并利用采样密度优先的方法,将图像分为4层进行非均匀采样和数据传送.实验表明:较之已有的非均匀采样方法,所提方法更具生物视觉特性和物理可实现性,并具有较高的压缩率;当注视点为图像中心时,数据压缩率达到了1/28;利用任意选取注视点得到的非均匀采样数据恢复的图像效果符合人眼视觉特性.The usage of biological vision selective attention mechanism（VSAM） is important in the vision research,and can reduce the computing data.A non-uniform sampling method of bionic vision is designed and implemented.A Gaussian function-based model is proposed from the analysis of the selective response characteristics of biological vision,and two non-uniform sampling expressions are obtained based on the priority of sampling layer and the priority of sampling density.According to characteristics of human eye,four non-uniform sample layers are carried out in images by the second expression,and the proposed sampling method is implemented in our self-designed vision chip.Experiment results and comparisons with existing non-uniform sampling methods show that the proposed method has the biological visual characteristics and hardware realization with a high compression ratio.The compression ratio reaches 1/28 when the focus point is in the image center,and the images restored from the non-uniform sampling data are in line with the VSAM of human eye.

关 键 词：仿生物视觉 视觉处理芯片 非均匀采样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]