Image Classification Using Functional Data Analysis: A Functional Feature Representation Approach

Image classification is a fundamental task in computer vision. However, existing deep learning methods often face challenges associated with high computational complexity and time costs when processing high-dimensional image data. To address this issue, this paper proposed a novel feature
representation approach based on Functional Data Analysis (FDA). Through FDA techniques, a common two dimension Bspline basis can be applied to transform the raw image data into lower-dimensional matrices, while retaining all essential information. This lead to a more efficient feature representation of the information of high dimensional image. Experiments conducted on ResNet, WideReNet and VGG models indicate our FDA-represented input data not only improves the classification accuracy compared to traditional methods but also reduces the overall computational time. The results of this study highlight the potential of FDA techniques in image classification, providing a novel way to enhance the efficiency and performance of deep learning models.