A Novel Joint Dictionary Learning Method for Image Classification

Abstract

Abstract Image classification is an essential component in the modern computer vision field, in which dictionary learning-based classification has garnered significant attention due to its robustness. Generally, most dictionary learning algorithms can be optimized through data augmentation and regularization techniques. In terms of data augmentation, researchers often focus on how to enhance the features of specific class samples while neglecting the impact of intra-class correlations. When intra-class correlation of images is high, distinguishing between different categories can become challenging, especially when there are small differences between categories. To tackle this concern, the paper advocates a novel data augmentation approach that enhances intra-class differences. The proposed method reduces excessive similarity within class samples by randomly replacing pixel values, thereby improving classification performance. Building on this, we designed a joint dictionary learning algorithm that embeds label consistency and local consistency by combining auxiliary samples generated by the data augmentation method with original samples to create a dictionary. The basic steps of the proposed algorithm are as follows:(1) Generate specific auxiliary samples as training samples; (2) Initialize the dictionary and expression coefficients; (3) Introduce label constraints and local constraints and update the dictionary; (4) Generate a classifier and classify the test samples. Extensive experiments have demonstrated the efficiency of the proposed approach. We will provide the code and datasets on https://github.com/mjLi0403/Joint-Dictionary-Learning-Algorithm-with-Novel-Data-Enhancement-Scheme.git.