Image Classification Method Based on Non-negative Sparse Coding Based on Structural Similarity

Abstract

The image classification method based on the non-negative sparse coding of structural similarity in the present invention comprises the following steps: respectively densely extracting SIFT features from all images in the image data set to be processed; randomly selecting several SIFT features for obtaining the image data set to be processed codebook; establish a non-negative sparse coding model based on structural similarity; use randomly selected SIFT features to solve the codebook of the image data set; fix the codebook to encode all SIFT features; The encoding of each image is integrated with the spatial pyramid maximum pooling method to obtain the feature vector of each image; the image data set is divided into a training set and a test set, and the image feature vector of the spatial pyramid maximum pooling of the training set and the corresponding image are obtained. Label, train a classifier; for any image, input the feature vector after the maximum pooling of its spatial pyramid into the trained classifier, and obtain the predicted category of the image.

Description

Technical field: [0001] The invention belongs to the technical field of computer vision image classification, in particular to an image classification method based on structural similarity non-negative sparse coding. Background technique: [0002] An important function of the primary stage of the biological visual system is to remove the statistical redundancy of input stimuli as much as possible. The response of the primary visual cortex to external stimuli satisfies sparsity, that is, only a few neurons are activated, and the corresponding coding is sparse coding. Sparse coding, in layman's terms, is to represent a signal as a combination of a set of bases, and requires only a few bases to reconstruct the signal. Sparse coding has been widely used in computer vision, image signal processing and other fields, for example, signal reconstruction, signal denoising, image feature extraction, and classification applications. [0003] The structural similarity index defines str...

AI Technical Summary

Problems solved by technology

Most of the current image classification methods based on sparse coding are based on the coding model that minimizes the sum of squared errors of reconstruction. The sum of squared errors as a criterion for distortion does not conform to the visual characteristics of the human eye.

Recent studies have shown that the main function of the human visual system is to extract structured information from images and videos from the visual area, and the sum of squared errors does not fully consider the visual characteristics of the human eye, so the traditional sparse coding reconstruction cannot be very good Evaluate the structural similarity between the reconstructed image and the original image

Images