Psychologically inspired visual information storage and retrieval modeling for multiclass image classification

The computer vision research aims to enable the computers to recognize visual images as easily as human. Studies have shown that human could segregate target from its surrounding environment, which is intimately associated with the brain memory mechanism. However, it is not quite clear about how the visual images are stored and retrieved in the human brain. In this paper, we propose a psychologically visual information storage and retrieval model (PVISRM) based on sparse coding and probabilistic decision theory. First, the dense scale invariant feature transform (SIFT) algorithm is applied to extract the features of visual images and then the extracted features are represented by sparse coding. In the storage procedure, each component of the feature vector is correctly copied with certain probability generated by an exponential distribution. For retrieval, the likelihood ratio between the probe image feature vector and that of each studied image is calculated based on probabilistic theory. Then the category likelihood ratio between the probe image and each category is obtained by adding the ratio values of all images belonging to the same category. Finally, the Bayesian decision rule for image classification is presented. Experimental results show that the proposed PVISRM model can obtain good classification performance and outperforms the SVM approach.