低光照强噪声背景下图像多阈值分割方法研究  

Research on Multi-Threshold Image Segmentation in Low Light and Strong Noise

作　　者：王宁[1] 韩院彬[1] WANG Ning;HAN Yuan-bin(Hebei University of Engineering,Hebei Handan 056038,China)

机构地区：[1]河北工程大学,河北邯郸056038

出　　处：《计算机仿真》2022年第3期170-173,178,共5页Computer Simulation

摘　　要：针对光照不均匀、突发噪声背景下,采用单一图像阈值分割方法兼顾所有像素,会造成图像分割效果较差、噪声平滑程度较低,导致降噪处理图像所丢失的细节较多的问题。提出低光照强噪声背景下图像多阈值分割方法。优化自适应灰度直方图多阈值,采用相应矩阵滤波处理低光照强噪声图像,根据直方图上点的界定,获取图中的背景与目标区域,根据区域对应的频率密度函数分布,得到相关概率公式与方差向量公式,依据不同灰度值的初始图像定义,构建直方图阈值等价判定准则函数,实现最优分割阈值的选取。实验结果表明,在对低光照强噪声的图像处理时,所提方法的图像分割效果较好,噪声平滑程度较高,降噪处理图像所丢失的细节较少。In the background of uneven illumination and burst noise,if a single image threshold segmentation method gives consideration to all pixels,the image segmentation effect and noise smoothness will be low,resulting in detail loss during noise reduction.Therefore,a multi-threshold image segmentation method in low illumination and strong noise was proposed.Multiple threshold values of the adaptive gray histogram were optimized.The corresponding matrix was used to filter the image with low light and strong noise.According to the point on the histogram,the background region and target region in the image could be obtained.According to the frequency density function distribution corresponding to the regions,the probability formula and variance vector formula were obtained.Based on the definition of the initial image with different gray values,the equivalent judgment criterion function of the histogram threshold was constructed to select the optimal segmentation threshold.Experimental results show that when processing images with low illumination and strong noise,the proposed method has a better image segmentation effect.In addition,its noise smoothness is higher,with fewer details lost during the noise reduction.

关 键 词：低光照 强噪声 多阈值分割 灰度等级 交叉熵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]