偏置场下光栅图像的目标轮廓特征点检测方法

The Approach of Feature Point Detection for Illumination Pattern-based Images under Bias Field

文章目标是解决有偏场环境下带有光栅图像的目标轮廓特征点检测问题。针对目标轮廓特征检测中存在的有偏光照环境和光栅模式,提出了一个两步解决方案。首先采用一种新的基于有偏场估计的图像模糊聚类迭代算法,对原灰度图像进行分割;接着,利用Harris特征检测器提取分割后目标图像的候选特征点,并在Harris特征检测算法中提出了基于特征响应函数直方图的罚值选择方法。实验结果表明,在光栅纹理和偏置场并存情况下,该文提出的方法优于传统Harris角检测器,解决了传统Harris角检测在该特定环境下所面临的精度下降问题。文章提出的算法可用于偏置场环境下光栅图像目标形状分析。

The objective of presented work is to provide approaches of interest point detection for the illumination images under bias fields.With respect to two of main difficulties：the image with changeable illuminations and strong pattern textures,a two-step approach is provided to tackling the problems in the paper.That is,a novel iterative algorithm of fuzzy clustering based image segmentation is firstly illustrated to achieve accurate object partitions.And then the Harris corner detector is used for extracting the points as candidate interest ones in the segmented images where a proposed scheme of threshold selection is given.Experimental results show that the performance of proposed algorithms overweight conventional Harris corner detector in accurately locating feature points of object images under co-xisting illumination pattern and bias fields.Furthermore,it can be used in real applications as a basis for the shape analysis under bias field with illumination patterns.