Some approaches to recognize 2d facial images

Abstract

Face recognition plays a vital role in day-to-day life. The process of identifying a person from a digital image or video from the region containing face is known as face recognition. Recognising face in an image is challenging since many variations exist in the image. Variations are in the form of different backgrounds, variation in colors, size, style, and so on. The extracted face image may not match with the query image. This issue can be overcome by developing better face detection algorithms, by improving the preprocessing stage, and using dominant features with respect to the face region. Image preprocessing improves the quality of the image. Hence, new improved preprocessing chain has been developed for enhancing the uality of the face image. This work proposes a face recognition system based on image preprocessing and texture features. The existing normalization rocess results in poor-quality images and consumes time to extract features. Therefore, the face images are preprocessed using image cropping, Gamma correction, Difference of Gaussian (DoG) filtering, and histogram equalization. This preprocessing chain removes most of the unwanted effects of changing illumination and retains the necessary information for recognising the face. From the preprocessed image, the Local Binary Pattern (LBP) code is obtained. Gray Level Co-Occurrence Matrix (GLCM) is constructed for the resultant LBP code, and Haralick texture features are calculated from the GLCM matrix. So, the time required for calculating features reduces to 7.5 seconds when compared to the time required for calculating the features directly from the normalized image in 10.5 seconds. Feret, Yale-B, and FRGC-204 datasets are used in experimentation newline