A New Bilinear Supervised Neighborhood Discrete Discriminant Hashing

Feature extraction is an important part of perceptual hashing. How to compress the robust features images into hash codes has become a hot research topic. Converting two-dimensional image one-dimensional descriptor requires higher computational cost and not optimal. In order maintain internal feature structure original image, new Bilinear Supervised Neighborhood Discrete Discriminant Hashing (BNDDH) algorithm proposed in this paper. Firstly, constructs two neighborhood graphs geometric relationship between samples reduces quantization loss by directly constraining codes. Secondly, small rotation matrices are used realize bilinear projection descriptor. Finally, experiment verifies performance BNDDH under different types, such as pixels Convolutional Neural Network (CNN)-based AlexConv5 feature. The experimental results discussion clearly show that better than existing traditional hashing can represent more efficiently