Search

We present a face verification system inspired by known properties of the human visual system. In the proposed algorithm the face is normalized for geometry and luminance, and Fourier‐Bessel (FB) descriptors are extracted from three locations in the eyes region (local analysis). The resulting representations are embedded in a dissimilarity space, where each image is represented by its distance to all the other images, and a Pseudo-Fisher discriminator is built. Using the FERET database, we submitted the system to a battery of tests under a wide variation of imaging conditions, including expression, age, and illumination variations. Results showed that the system outperformed previous state-of-the-art methods in most testing conditions. To deal with partial occlusions, we implemented an occluded region detector that resulted in low performance loss under up to 50% occlusion level. Finally, we automated the registration step by implementing face and eye detection algorithms. We also showed that the local-FB analysis outperforms the global-FB version of the system and an alternative polar frequency representation. In conclusion, the intermediate-scale local analysis approach used in the proposed system resulted in state-of-the-art face verification performance and high

Motivated by a recently proposed biologically inspired face recognition approach, we investigated the relation between human behavior and a computational model based on Fourier-Bessel (FB) spatial patterns. We measured human recognition performance of FB filtered face images using an 8-alternative forced-choice method. Test stimuli were generated by converting the images from the spatial to the FB domain, filtering the resulting coefficients with a band-pass filter, and finally taking the inverse FB transformation of the filtered coefficients. The performance of the computational models was tested using a simulation of the psychophysical experiment. In the FB model, face images were first filtered by simulated V1- type neurons and later analyzed globally for their content of FB components. In general, there was a higher human contrast sensitivity to radially than to angularly filtered images, but both functions peaked at the 11.3-16 frequency interval. The FB-based model presented similar behavior with regard to peak position and relative sensitivity, but had a wider frequency band width and a narrower response range. The response pattern of two alternative models, based on local FB analysis and on raw luminance, strongly diverged from the human behavior patterns. These results suggest that human performance can be constrained by the type of information conveyed by polar patterns, and consequently that humans might use FB-like spatial patterns in face processing.

Motivated by a recently proposed biologically-inspired face recognition approach, psychophysical experiments have been carried out. We measured recognition performance of polar frequency filtered face images using an 8-alternatives forced-choice method. Test stimuli were generated by converting the images from the spatial to the polar frequency domain using the Fourier-Bessel transformation (FBT), filtering of the resulting coefficients with band-pass filters, and finally taking the inverse FBT of the filtered coefficients. We also evaluated an automatic FBT-based face recognition model. Contrast sensitivity functions of the human observers peaked in the 8-11.3 radial and angular frequency range, with higher peak sensitivity in the former case. The automatic face recognition algorithm presented similar behavior. These results suggest that polar frequency components could be used by the human face processing system and that human performance can be constrained by the polar frequency information content.

We present a novel local-based face verification system whose components are analogous to those of biological systems. In the proposed system, after global registration and normalization, three eye regions are converted from the spatial to polar frequency domain by a Fourier-Bessel Transform. The resulting representations are embedded in a dissimilarity space, where each image is represented by its distance to all the other images. In this dissimilarity space a Pseudo-Fisher discriminator is built. ROC and equal error rate verification test results on the FERET database showed that the system performed at least as state-of-the-art methods and better than a system based on polar Fourier features. The local-based system is especially robust to facial expression and age variations, but sensitive to registration errors.

We present an automatic face verification system inspired by known properties of biological systems. In the proposed algorithm the whole image is converted from the spatial to polar frequency domain by a Fourier-Bessel Transform (FBT). Using the whole image is compared to the case where only face image regions (local analysis) are considered. The resulting representations are embedded in a dissimilarity space, where each image is represented by its distance to all the other images, and a Pseudo-Fisher discriminator is built. Verification test results on the FERET database showed that the local-based algorithm outperforms the global-FBT version. The local-FBT algorithm performed as state-of-the-art methods under different testing conditions, indicating that the proposed system is highly robust for expression, age, and illumination variations. We also evaluated the performance of the proposed system under strong occlusion conditions and found that it is highly robust for up to 50% of face occlusion. Finally, we automated completely the verification system by implementing face and eye detection algorithms. Under this condition, the local approach was only slightly superior to the global approach.

A novel biologically motivated face recognition algorithm based on polar frequency is presented. Polar frequency descriptors are extracted from face images by Fourier-Bessel transform (FBT). Next, the Euclidean distance between all images is computed and each image is now represented by its dissimilarity to the other images. A Pseudo-Fisher Linear Discriminant was built on this dissimilarity space. The performance of Discrete Fourier transform (DFT) descriptors, and a combination of both feature types was also evaluated. The algorithms were tested on a 40- and 1196-subjects face database (ORL and FERET, respectively). With 5 images per subject in the training and test datasets, error rate on the ORL database was 3.8, 1.25 and 0.2% for the FBT, DFT, and the combined classifier, respectively, as compared to 2.6% achieved by the best previous algorithm. The most informative polar frequency features were concentrated at low-to-medium angular frequencies coupled to low radial frequencies. On the FERET database, where an affine normalization pre-processing was applied, the FBT algorithm outperformed only the PCA in a rank recognition test. However, it achieved performance comparable to state-of-the-art methods when evaluated by verification tests. These results indicate the high informative value of the polar frequency content of face images in relation to recognition and verification tasks, and that the Cartesian frequency content can complement information about the subjects' identity, but possibly only when the images are not pre-normalized. Possible implications for human face recognition are discussed., Comment: ACM Transactions on Applied Perception