基于神经网络的红外图像实时去斑方法

在分析红外图像斑纹产生机理基础上,提出利用人眼视网膜自适应调节原理的去斑方法。去斑时,根据距离不同作用不同原则,对中心及其周围像素分别予以不同权重;将斑纹信息加入求解过程,能有效快速去斑;去斑后,采用基于阈值的直方图规定化方法,以弥补对比度可能降低情况。研制出的DSP去斑系统实现了实时去斑功能。

A real-time IR image speckle reducing algorithm based on neural network

A speckle reduction algorithm using adaptive correcting principle for human eye retinal is proposed on the basis of analyzing speckle creating mechanism in IR image. During speckle reducing, the different weightings are given to center and its surrounding pixels according to the rules of distance determining actions. A quick speckle removing can be effectively carried out when adding speckle information into solving process. After speckle reducing, a histogram defined method based on the threshold is employed to compensate for possible contrast reduction. A real-time speckle removing function is performed by using the developed DSP speckle reduction system.