A novel technique manages the effect of lighting on photometric-based human face recognition through a fuzzy-based illumination invariant method. The technique, named OptiFuzz, uses an extended reflectance model to adjust the effect of lighting on human faces, thereby improving face detection and recognition results under a variety of illumination conditions. Result of the illumination invariant face processing using Yale B Extended database: input images (top) and processed images (bottom). Courtesy of Toyohashi University of Technology. Developed by researchers at Toyohashi University of Technology, OptiFuzz has one variable, the illumination ratio, which is controlled by a fuzzy inference system (FIS). The researchers used a genetic algorithm (GA) to optimize the FIS rule to handle a range of illumination conditions. "To eliminate the effects of light, image contrast should be adjusted adaptively," said researcher Bima Sena Bayu Dewantara. "To produce an invariant face appearance under backlighting, for example, cheeks need to be brightened, while the eyeballs must be kept dark. Such an adaptive contrast adjustment can be performed using the developed reflectance model, and we show that a combination of FIS and GA is very effective for implementing the model." Results of illumination invariant face recognition for real implementations: (a) person #1 outdoor, (b) person #1 indoor, (c) person #2 outdoor; and (d) person #2 indoor. Small image on the bottom-right side of each image is the input image. Courtesy of Toyohashi University of Technology. The researchers tested the method using Yale B Extended and CAS-PEAL face databases to represent the offline experiments. For the online indoor and outdoor experiments, they recorded several videos. They used the Viola-Jones Face Detector and the Mutual Subspace Method to run the online face detection and face recognition experiments. The experimental results demonstrated that the researchers’ algorithm could outperform existing methods for recognizing a specific person under variable lighting conditions with a significantly improved computation time. The results also showed that illumination invariant images could improve face detection performance. Professor Jun Miura and Ph.D. candidate Bima Sena Bayu Dewantara. Courtesy of Toyohashi University of Technology. "By just adding this contrast adjustment to present face recognition systems, we can largely improve the accuracy and performance of face detection and recognition," said professor Jun Miura. "Moreover, this adjustment runs in real time, and therefore, it is appropriate for real-time applications such as robot and human-interaction systems." To date, vision-based face detection and recognition has been shown to be effective only under normal illumination conditions. In addition to a person's identity, a face provides information such as a person's focus of attention and degree of tiredness, which can be useful for maintaining a comfortable human-machine interaction. The researchers expect that their proposed contrast adjustment method will be useful in various situations, especially under severe illumination conditions. The research was published in Machine Vision and Applications (doi: 10.1007/s00138-016-0790-6).