由于成像系统和目标的距离和方位的原因，会导致形成的图像发生透视而不再是正视图，这对于图像的特征提取、识别等后续处理非常不利。对于发生了透视畸变的图像，透视变换解决了一般仿射变换不能改变图像内部点相对位置的缺陷。文章根据透视变换的原理提出了基于改进的霍夫变换和透视变换的透视图像矫正技术，该算法根据摄像机的成像机理进行透视图像矫正，Matlab仿真实验结果表明，该算法可行性高，简单易行，稳定性好，可广泛推广至目标识别等领域。

When the images collected from camera have perspective distortion, it can affect the following work such as feature extraction and object recognition due to the influence of perspective projection. When rectifying these distorted images, perspective transformation can change the relative position of inner points, while affine transformation cannot. This article will improve the Hough transformation to find the lines that hold the four points which perspective transformation needs, and then use the perspective transformation to rectify the distortion based on the imaging principle of cameras. Emulation experiment results from Matlab reveal that the algorithm can effectively rectify the distorted images and achieve our expectation. This algorithm is easy, feasible and robust, and can be applied to many fields including object recognition.