Recognizing human faces under varying degree of Illumination: A comprehensive survey

Illumination variation is one of the well-known and challenging problems known for face recognition. A lot of studies have been explored to reduce the effect caused by varying illumination. We have analysed the latest state of art technique and have divided into two categories. One based on Singular Value Decomposition (SVD), Resonance and Local Binary Pattern (LBP) techniques. And the second based on the Self Quotient Image (SQI) and Histogram Based techniques. In the first category SVD technique gives 99.53% of recognition rate on Yale B face database, Principle Component Analysis (PCA) technique gives 100% recognition rate on Yale B database, LBP and Circle Threshold (CT)-LBP techniques give 94.8% and 98.12% of recognition rate using Yale B database. In the second category SQI technique gives 98.3% of recognition rate on Yale B database and Histogram technique gives 100% recognition rate using CMUPIE database. This paper aims to give a detailed survey of various face recognition algorithms. The review covers all the approaches that aim to solve all the problems that are created due to varying illumination.