Construction of Delaunay Triangles for Face Recognition in Images

A Delaunay graph method is a proposed solution to the matching issue in facial recognition. The techniques for feature matching and feature detection are scale-invariant feature transform (SIFT) and speeded-up robust features (SURF). In the existing method, the major drawback is that each algorithm only extracts particular features, leading to many mismatches in matched features. To avoid those mismatches, this paper concatenates the SIFTSURFLBPCNN together. Input images are pre-processed and key points are detected using sift algorithm. For each Detected key point, textural, and geometrical uutilizing, characteristics are extracted scale-invariant feature transform(SIFT), speeded-up robust features (SURF), local binary pattern(LBP), Convolution neural network(CNN) and concatenated to form a feature vector. Using the lows method, the key points between the input images are filtered out. Then the Delaunay triangles are constructed for the input images from their respective filtered key points using the Delaunay triangulation method. Testing is done and each algorithm’s performance is evaluated.