摘要
提出一种以人类视觉系统初级阶段多通道滤波理论为基础的光电混合纹理分割技术，每一通道用盖伯函数描述．在光学系统中采用不同取向的针孔对阵列作为盖伯滤波器，用２×２Ｄａｍａｎｎ光栅作为分束器，建立了具有４个通道的纹理图像分割处理系统．光学系统结果由计算机进行聚类或神经网络决策，得到最后的分割标定图像．若增加针孔对的数目，并改变针孔大小，用更高分束比的Ｄａｍａｎｎ光栅作为分束器件，可以实现通道的纹理图像分割。
An opto electrical hybrid texture segmentation technique was proposed based on the multi channel filtering mechanism in the early stage of human visual system.Each channel was characterized by Gabor wavelet.In the optical realization,a pinhole pair array was fabricated as the Gabor filters and a 2×2 Damann grating was adopted as the beam splitting element.The optical filtering results were transferred into a digital computer for further processing by clustering algorithm or neural network strategy.A final labeled image was obtained.The proposed system could be implemented for practical application of the number if pinhole arrays was increased and the Damann grating with a high beam splitting ratio was adopted.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
1996年第2期145-150,共6页
Journal of Infrared and Millimeter Waves
基金
国家自然科学基金
关键词
多通道
滤波
纹理分割
子波变换
图像处理
multi channel filtering,texture segmentation,wavelet transform.
