Quality based classification of images for illumination invariant face recognition

Quality of an image plays a fundamental role in taking vital decisions. In various walks of life, one such decision is personal identification. Hence, it's assessment is essential prior to using it in many biometric applications such as face recognition, iris, fingerprint analysis etc. The proposed technique classifies images into four classes based on their illumination and contrast quality. Then, the proposed technique chooses the most suitable enhancement technique for particular class to get best possible image. The proposed technique has been experimented on the Yale B database and the results obtained are 97.14% accurate on an average in terms of the correct classification of images into the appropriate classes. In another experiment where 50 random images of 30 random subjects were selected and this process repeated over 10 times, the classifier was 99.17% accurate in classifying the images.