Search

International audience; Identifying the human face via correlation is a topic attracting widespread interest. At the heart of this technique lies the comparison of an unknown target image to a known reference database of images. However, the color information in the target image remains notoriously difficult to interpret. In this paper, we report a new technique which: (i) is robust against illumination change, (ii) offers discrimination ability to detect color change between faces having similar shape, and (iii) is specifically designed to detect red colored stains (i.e. facial bleeding). We adopt the Vanderlugt correlator (VLC) architecture with a segmented phase filter and we decompose the color target image using normalized red, green, and blue (RGB), and hue, saturation, and value (HSV) scales. We propose a new strategy to effectively utilize color information in signatures for further increasing the discrimination ability. The proposed algorithm has been found to be very efficient for discriminating face subjects with different skin colors, and those having color stains in different areas of the facial image.

International audience; Independent component analysis (ICA) models, describing a given signal as a linearcombination of various independent sources, have proven to be a fruitful endeavor. Oneprominent example deals with audio applications in order to separate the speaker’s voicefrom environmental noises disturbing it. However, very few ICA based systems areavailable for biometric encryption applications. For that specific purpose, the ICAmethod can be easily adapted to add noise to a target image in order to encrypt it. In thischapter, at first, we discuss biometric recognition systems based on the ICA andcorrelation approaches. Next, we explore an ICA-based algorithm for face recognition.Basically, it consists of building a base of independent components using a learningdatabase that contains several chosen reference images. Then, the target image (image tobe recognized) is projected on the independent component base, and the similaritybetween the target image and each of the reference images is studied. Discrimination testsbetween the proposed technique and alternate methods are conducted by using thePointing Head Pose Image Database (PHPID). In this chapter we report some of therecent developments dealing with the ICA method for face recognition applications. Aspart of our analysis, we precisely determine a set of metrics aimed at better understandingthe role of the number and choice of the reference images on the performance of theproposed technique.

International audience; We consider a new approach for enhancing the discrimination performance of the VanderLugt correlator. Instead of trying to optimize the correlation filter, or propose a new decision correlation peak detection criterion, we propose herein to denoise the correlation plane before applying the peak-to-correlation energy (PCE) criterion. For that purpose, we use a linear functional model to express a given correlation plane as a linear combination of the correlation peak, noise, and residual components. The correlation peak is modeled using an orthonormalized function and the singular value decomposition method. A set of training correlation planes is then selected to create the correlation noise components. Finally, an optimized correlation plane is reconstructed while discarding the noise components. Independently of the filter correlation used, this technique denoises the correlation plane by lowering the correlation noise magnitude in case of true correlation and decreases the false alarm rate when the target image does not belong to the desired class. Test results are presented, using a composite filter and a face recognition application, to verify the effectiveness of the proposed technique.