随着计算机视觉和计算机图形学的快速发展，从图像特征信息出发，探索成像对象已经成为了一个热点问题，也在立体视觉发展中担任着重要的角色。因此，本文从目标对象的图像角度出发，通过对图像边缘检测、特征点提取，并通过特征点的匹配来获得描述目标的特征信息点。 　　　为了能够获得恰当的描述目标对象的图像信息，考虑受到拍摄环境噪声等因素的影响，本文提出了一种改进的图像边缘检测算法，对模糊图像能够提取到精确的、完整的图像边缘。再通过Harris算法和SIFT算法分别提取图像中的特征点信息，然后，针对不同的图像特征点的描述信息，考虑到在灰度特征信息下的传统的灰度相关法以及在描述子信息下的描述子欧氏距离匹配都会产生较大误匹配的缺陷，本文在这方面提出了对于不同特征点描述信息的改进的匹配算法，能够获得更多的、正确的匹配点。最后，通过对立体视觉理论的研究，在摄像机成像原理的基础上，通过对对极几何约束法和最小平方中值法两种剔除误匹配算法研究，比较其性能，本文采用了最小平方中值法来对已获得匹配特征点集进一步精确匹配，剔除了误匹配点后，极大地减少了匹配点集当中误匹配点的数量。最后，通过图像的特征信息点提取到了立体空间中的特征信息点。 本文对于图像特征信息提取算法的仿真实验主要是采用Matlab工具完成的，实验结果显示本文算法能够取得大量地、可靠地描述图像信息的特征点，由此也证实了，本文在提取图像特征信息的过程中提出的方法是切实可行。

With the rapid development of computer vision and computer graphics, based on image feature information, exploring the imaging object has become a hot topic and has been also important role in the development of stereo vision. Therefore, describing the characteristical information points of the target based on the angle of the target image has been extracted through edge detection of the image, feature point extraction, and feature points matching. 　　　Considering effect from factors like environmental noise, in order to be able to obtain the appropriate description information of the target object image, the paper proposed improved algorithm of image edge detection, the algorithm can extract accurate the image edge for the fuzzy image. Respectively, Harris algorithm and SIFT algorithm extract the feature points in the image information, according to different description information of image feature points: the traditional gray correlation method under gray information and Euclidean distance matching of the descriptor information will be considered to produce larger defects of mismatch. Therefore, the improvement different feature point description information of matching algorithm is put forward, more and correct matching points is able to be get. Based on the research of the theory of Stereopsis, on the basis of camera imaging principle, two eliminate false matching algorithm research: epipolar geometry method and the least square median method, after comparing, the least-square-median method is adopted to further exact match for obtained matching feature points, after eliminating the false matching points, the number of the points of mismatch points is greatly reduced. Finally, the features in the solid space information point is extracted by the characteristics of the image Information. Simulation experiments for image feature information extraction algorithm is completed by using Matlab tools, through the experiment, large quantities of describing the image information of feature points in a reliable way are obtained in the experimental results.Thus, the feasibility of proposed method is confirmed.