中文摘要：

溶蚀孔洞在碳酸盐岩储层中是重要的流体储集空间,研究成像测井（FMI）图像孔洞连通域标记及信息定量拾取很有意义.全井眼微电阻率成像测井（FMI）经数据处理后可得到全井壁高分辨率的彩色图像,经图像灰度化、中值滤波处理后,通过阈值分割得到能够反映井壁溶蚀孔洞特征的二值图像,孔洞表现为黑色暗斑.基于等价对处理的图像连通域标记算法具有快速、不重复标记的优点,利用该算法,可准确地从二值图像中标记溶蚀孔洞连通域,进而可对每个连通域进行目标信息拾取,包括孔洞尺寸、连通域面积、圆度等.利用反映溶蚀孔洞发育程度的面孔率曲线对图像进行分层,在此基础上可拾取每一层段溶蚀孔洞面孔率、分选系数及溶洞密度值的分布等非均质信息,能够定量地评价溶蚀孔洞发育、非均质性强的碳酸盐岩储层,也是FMI图像应用于岩石孔洞结构信息定量表征新的尝试.

英文摘要：

Solution caves are important spaces in carbonate reservoirs,which usually have complex causes and strong heterogeneity,so it is very difficult to predict.The caves,which control the oil-gas distributions to some extent,are developed during the palaeokarst process and are usually associated with fractures.According to shape and attitude they can be divided into many types such as the spot-plaque solution cave,stratiform cave and reticular fracture-cave.Studying the connected domain of these caves with quantificational labeling and parameterextraction have important significance to effective evaluation of carbonate reservoirs.The full borehole microresistivity imaging logging（FMI）with high resolution and large amount of information can reflect wellbore section features.It has a broad application in geology and rock physics,so it is important information to studying solution caves and fractures in carbonate reservoirs.This work takes the FMI high-resolution color images from the hole-type carbonate formation of Ordovician in the Tarim basin as the research object.First of all,the original image is gray processed to reduce the data size of the image to facilitate subsequent processing.Then in order to filter noise and protect the edge information,we conduct median filtering processing on gray images to realize the nonlinear smoothing.Finally,we use cores calibrate image to select the appropriate threshold and segmentation image after the median filtering processing to get the binary image which can reflect the characteristics of the borehole wall,in which solution pores are displayed as black spots.After the binaryzation,we adopt the labeling algorithm for image connected domain based on equivalence pair processing which has the advantages of fast and no-repeat to process the image.In the same time,we combine the eight-neighborhood connect rule to labeling holes connected domain,where every hole will yield a unique number.And then,the information of every connected domain can be fetched including the hole size（