刘文亮　王正勇　卿粼波　滕奇志

(四川大学电子信息学院　成都　610065)



LBP算法在岩石薄片图像拼接中的应用*

刘文亮王正勇卿粼波滕奇志

(四川大学电子信息学院成都610065)

摘要基于SIFT特征点的图像拼接算法对受消光性和干涉色影响的岩石薄片正交偏光图像进行拼接时,由于特征信息相似度极高,传统的BBF-RANSAC算法无法很好地剔除错误匹配对,从而导致错误的拼接结果。论文将LBP算法引入SIFT特征点描述中,利用SIFT算法对岩石薄片正交偏光图像进行特征点提取,在特征点描述时,采用隔点计算特征点周围像素点的旋转不变均匀LBP特征值,用其替代一阶梯度幅值,生成改进的SIFT特征点描述向量。实验结果表明,论文提出的基于旋转不变均匀局部二值模型的SIFT特征描述子提高了拼接算法的鲁棒性和速度。

关键词岩石薄片; 图像拼接; LBP; SIFT

Application of LBP Algorithm in Rock Slice Image Stitching

LIU WenliangWANG ZhengyongQING LinboTENG Qizhi

(College of Electronic and Information, Sichuan University, Chengdu610065)

AbstractThe image stitching algorithm based on SIFT features will lead to incorrect results when it is applied to orthogonal polarization core slice image stitching with high similarity of feature information, which are affected by extinction and interference, as the traditional BBF-RANSAC algorithm cannot eliminate false matches well. LBP algorithm is applied to SIFT feature descriptors. Firstly, features of orthogonal polarization rock images are calculated by using SIFT algorithm. Secondly, rotation-invariant uniform LBP values of uniformly-spaced pixels around the features are calculated instead of gradient value, when calculating feature descriptors. Finally, modified SIFT feature descriptors are generated. Experimental results show that the presented modified SIFT feature descriptors can improve the robustness and speed of image stitching algorithm.

Key Wordsrock slice, image stitching, LBP, SIFT

Class NumberTN315

1引言

在石油地质勘探中,需要将钻探设备钻井获取的岩心样品制成岩石薄片,利用显微成像技术和数字图像分析技术获得岩石样品的微观特性[1]。在对岩石薄片显微成像的过程中,受目镜视野范围和成像设备靶面的限制,需要对岩石薄片进行多次图像采集,得到多幅局部视域岩石薄片图像。通过图像拼接技术将所获得的多幅局部视域岩石薄片图像拼接成最终的岩石薄片全景图,将实物地质样本数字化,易于管理和实现永久保存[2]。David Lowe于1999年提出SIFT(Scale Invariant Feature Transform)特征提取算法[3]并于2004年整理和完善[4]。吴晓红、罗代升于2009年将SIFT特征提取算法引入岩心图像自动配准[5],实现了岩心图像的自动拼接。……