Abstract:
In today's authentication biometric technologies, face recognition is the most popular and widely used technique. It is the most acceptable since it could be used without requiring object coordination. However, during the acquisition of the photograph, face recognition is vulnerable to variations in lighting and changes in the orientation of the face. The grayscale representation of the facial image is commonly used as an input function of face authentication systems. However, we suggest a new approach for improving the efficiency of these schemes based on one-dimensional statistics of the face picture and the use of color representation. For the transformation of the RGB colorimetric components of the original images, we evaluated many colour spaces. The findings from the various spaces and colorimetric components are coupled using a nonlinear fusion for classification with a single RBF type neuron network. This novel facial authentication strategy that achieves a 95.85% success rate using nonlinear fusion of the colorimetric components of the YCrCb color space and a 91.75% success rate while using grayscale. As opposed to the use of grey scale photographs, the rate of effectiveness has increased by 5%. The findings reveal much interest in developing a new solution that reduces processing time due to its flexibility and robustness when dealing with a vast database and that color information improves the efficiency of this face authentication scheme. We checked these methods on frontal photographs from the XM2VTS collection using the Lausanne protocol to back up our findings.
