A feature patch-based three-dimensional vision measuring technique for complex surface and silhouette is proposed. There are several procedures for the proposed method including detecting and matching multi-modal local features, initializing, expanding and filtering patch sets. The algorithm outputs a dense set of rectangular patches covering the surfaces visible in the input calibrated images. The first step of the proposed algorithm is implemented as a matching, expanding, and filtering procedure. It starts from a sparse set of matched key points, and repeatedly expands these to nearby pixel correspondences using the monogenic feature congruency and the epipolar geometric constraint before using visibility constraints to filter away false matches. The keys to its performance are effective techniques for enforcing local photometric consistency and global visibility constraints. A simple but effective polygonal surface extraction algorithm is then used to turn the resulting patch model into a mesh appropriate for image-based modeling. According to the multi-modal monogenic features of a patch, the color and texture information is fused into the reconstructed mesh. Thus a three-dimensional high-fidelity solid model can be obtained finally.