论文部分内容阅读
本文提出了一种数字图象边缘检测的非线性代数方法。克服了常用方法对不同强度边缘进行检测时易丢掉弱边缘的弱点,并有效地抑制了噪声的影响。首先对中心象元与其δ邻城内各象元灰度差△T_i按其大小分成两相等的集合,再据一非线性加权系数A(△T_i),求得各集合的平均灰度差△T_i(i=1,2),定义边缘值为|△T_1-△T_2|。另外在本文的后一部分,提出了一个用来评价各种方法检测效果的指标一优质系数R,它比较客观、准确。
This paper presents a method of nonlinear algebra for edge detection of digital images. The method overcomes the weakness that the commonly used method is easy to lose the weak edge when detecting the edge with different intensities and effectively restrain the influence of the noise. First of all, the central pixel and its adjacent neighborhood within the gray-level pixel △ T_i according to their size divided into two equal set, and then a non-linear weighted coefficient A (△ T_i), the average gray-level difference of each set △ T_i (i = 1, 2), define the edge value as | △ T_1 - △ T_2 |. In addition, in the latter part of this paper, we propose a high quality coefficient R which is used to evaluate the detection effect of various methods. It is more objective and accurate.