Medical Image Classification Algorithm Based on Frequency Domain Perception

EasyChair Preprint 15872

Abstract

In the field of medical image classification, early research used machine learning algorithms to classify medical images based on artificially extracted features, but the development was slow. In order to solve the above existing problems, this paper proposes two methods that combine frequency domain learning with deep learning. The main research contents are as follows. First, this paper proposes a medical image classification method based on wavelet transform and transfer learning, combining frequency domain learning with deep learning. In order to comprehensively consider the color feature infor-mation and frequency domain feature information in RGB images, a fre-quency domain feature extraction module is proposed: first, the input image is separated along the channel to obtain three color components of FR, FG, and FB, and then the wavelet transform method is used to transform the three components of FR, FG, and FB into frequency domain features. At the same time, in order to reduce the amount of information processing, the low-frequency features that have the greatest impact on the classification results are selected from the three components for splicing to obtain the frequency domain feature matrix.

Keyphrases: Attention Mechanism, Convolutional Neural Network, Transfer Learning, medical image classifica-tion, wavelet transform

@booklet{EasyChair:15872,
  author    = {Linzheng Huang and Jiaxin Zhou},
  title     = {Medical Image Classification Algorithm Based on Frequency Domain Perception},
  howpublished = {EasyChair Preprint 15872},
  year      = {EasyChair, 2025}}