Face recognition with unknown, partial distortion and occlusion is a practical problem, and has a wide range of applications, including security and multimedia information retrieval. The authors present a new approach to face recognition subject to unknown, partial distortion and occlusion. The new approach is based on a probabilistic decision-based neural network, enhanced by a statistical method called the posterior union model (PUM).
PUM is an approach for ignoring severely mismatched local features and focusing the recognition mainly on the reliable local features. It thereby improves the robustness while assuming no prior information about the corruption. We call the new approach the posterior union decision-based neural network (PUDBNN).The new PUDBNN model has been evaluated on three face image databases (XM2VTS, ATT and AR) using testing images subjected to various types of simulated and realistic partial distortion and occlusion. The new system has been compared to other approaches and has demonstrated improved performance.
Human face recognition has been widely explored and applied in security, human computer intelligent interaction, digital libraries and robotics. There are many methods and techniques that have been applied to facial recognition, including principal component analysis (PCA), support vector machines (SVM), linear discriminant analysis (LDA), and neural networks. However, most of the systems designed to date work mainly for images that are captured under controlled conditions. They usually lack robustness when dealing with images involving unexpected mismatches, including, for example, mismatched poses, scale, facial expression and illumination. They are also sensitive to partial distortion and occlusion. In this topic, we focus on the problem of improving the robustness against local distortion and occlusion.
A number of techniques have been developed to deal with the problem of face recognition with partial occlusion and distortion. Many of these are based on the idea of recognition by parts, also called local matching techniques.These techniques aim to focus the recognition on the parts of the images not affected by the distortion/ occlusion, thereby improving the robustness.The techniques comprise two stages
- Dividing a face image into several local parts and representing each part independently of the other parts
- Combining the local matching scores from the individual parts into an overall score to reach a recognition decision.
To combine the local matching scores into an overall decision, a common approach is to use a predefined voting space.This approach works for matched training and testing, but less so for the presence of mismatches between the training and testing features. A probabilistic approach in which each partial image is modeled by a Gaussian mixture model (GMM), and the final decision is based on the sum of the local likelihoods from the individual GMMs. Recently, this approach has been extended to include weights to emphasis those local features that are affected by facial expression changes, where the weights are estimated using an optical flow approach. Go through the attached report for design and implementation.