关键词: 计算机视觉, 模式识别, 局部影像特征匹配, 仿射不变特征

Abstract: A novel algorithm based on local affine invariants was proposed for searching correspondences between images in the most general case. Image point features were used as anchor points and the local space-variant regions around the anchor point were used as local window region to generate geometric and brightness affine invariants for image matching. The method is efficient because only the local image features are used in the process. The matching results are reliable and robust because its feature matching region is relatively large, and the shape of local window region was transformed according to possible projective deformation. In addition, overlapped sub-regions were used in local feature extraction to reduce partially occluding problems.

Key words: computer vision, pattern recognition, local image feature matching, affine invariants