As mentioned in Chap. 4 , template matching is the most basic and simplest form of image matching, where the template and the sub-image to be matched must have a high degree of similarity. So, how is image matching generally achieved? For example, how can we find the matching relationship between two images of a building taken from different perspectives, as shown in Fig. 6.1a? This requires detecting features in the image for matching. Speaking of features, the first thing that might come to mind are those special locations, such as the vertices of a building, as shown in Fig. 6.1b. These local features are often referred to as keypoint features or interesting point features. As keypoint features, they usually need to satisfy four properties: (1) repeatability, the same keypoint feature should be detectable in images under different geometric and photometric transformations; (2) saliency, each keypoint feature has a distinctive description that is different from other keypoint features, usually representing the image block around its location; (3) compactness, the number of keypoint features should be much less than the number of image pixels; (4) robustness, keypoint features should have a certain insensitivity to cluttered image backgrounds and occlusions. The keypoint features shown in Fig. 6.1b possess these properties, and they are all located at the intersection of two lines (edges), so they have a vivid name—corners. By matching corresponding corners in two images (Fig. 6.1c), and calculating the transformation matrix based on the matching relationship, the two images can be stitched into a larger image (Fig. 6.1d). This chapter will introduce a classic image corner detection algorithm—the Harris corner detection algorithm.