Abstract:In this paper, we propose to combine Active Shape Models (ASMs) and Active Appearance Models (AAMs), for reliable image interpretation. In the proposed method, ASM local searching strategy is firstly used, then, besides the statistical shape constraints as in the ASMs, global texture constraint based on the subspace reconstruction error is further exploited to evaluate the fitting degree of the currently estimated model to the novel image. And, similar to the strategy used in the AAM approach, global texture is used to predict and update the model parameters. Therefore, by such an interleave iteration of ASM and modified AAM, our method takes the advantages of both ASMs and AAMs, while avoids their shortcomings. Our experiments on the face database significantly show the effectiveness of our method.

Combining Active Shape Models and Active Appearance Models For Accurate Image Interpretation