Over the last few years, the title of Illumination normalization and face recognition has been particularly interesting. Many researches have focused on reducing the level of lighting by developing advanced algorithms and techniques to perform this task. Illumination normalization are an important factor in image processing because there are negative factors that need to be eliminated to identify images. The algorithms currently used as dual face recognition algorithms find difficulties in different lighting conditions. This study proposed a framework in which a single input image that has gone through an arbitrary pixel brightness transformation (PBT) generates normalized images thus reducing the illumination normalization. Numerous algorithms will be assessed comprising gamma correction, histogram equalization, blind estimation through convex optimization and sigmoid stretching. Image recognition of the new illumination normalization algorithm’s implementation & evaluation is analysed in MATLAB for results. Furthermore, the performance of suggested method will be assessed by measurements of root means square error amidst the corresponding estimated PBT, the original PBT and in minimizing the optimization problem. The study covers the effects of these image normalization techniques on an image recognition task. Experimental assessment will be accompanied on Hateren database’s benchmark image recognition. In the result part we prove the enhancement of arbitrary intensity transformation in the Illumination normalization of image by enhancement the methods of increase image visibility and details. We consider the anticipated work can serve significant roles in the allied fields.
Volume 11 | 01-Special Issue
Pages: 1741-1747