Image Recognition from Face Feature Descriptor Using Counter Propagation Neural Network

K. Anandhi
2017 International Journal of Advanced Research in Computer Science and Software Engineering  
This paper presents the artificial neural network approach namely counter propagation neural network (CPN). It proposes a novel local feature descriptor using Counter propagation method number pattern (CPMN) for face analysis, i.e., face and expression recognition. CPMN encodes the directional information of the face's textures (i.e., the texture's structure) in a compact way; it is used to produce a more discriminative code than current methods. It is used to compute the structure of each
more » ... -pattern with the aid of a compass mask that extracts directional information, and it encodes such information using the prominent direction indices (directional numbers) and signwhich allows distinguishing among similar structural patterns that have different intensity transitions. The image processing techniques have been used to divide the face into several regions, and extract the distribution of the CPMN features from them. Then, this project concatenates these features into a feature vector, and it is used as a face descriptor. It will perform several experiments in which the descriptor performs consistently under illumination, noise, expression, and time lapse variations. Moreover, by testing descriptor with different masks to analyze its performance in different face analysis tasks. I.e. getting a relevant image from a small region or part of the face. The proposed method that provides good classification efficiency.
doi:10.23956/ijarcsse/sv7i5/0105 fatcat:nhtlbh7renhorp2fptqwrfhlxu