Scholarlink

<p>This study introduces an innovative methodology that synergizes wavelet-based feature extraction with multiscale fusion and deep learning, aiming to elevate the image classification task. Initially, the Haar wavelet transform is applied to lung X-ray images, extracting approximation, horizontal, vertical, and diagonal sub-band coefficients. Following this, a multi-scale horizontal fusion technique integrates wavelet-transformed images at different resolutions with the original image. This process creates an enriched feature map with essential contextual information for the subsequent deep-learning model. The resulting image features serve as input for the deep learning models, encompassing conventional Convolutional Neural Networks and transfer learning models such as ResNet50 and VGG19. The Experimental demonstrates the proposed MHWF-CNN method performance in terms of accuracy that using the publicly accessible Chest X-ray dataset. The proposed method attains 99% accuracy with the integration of advanced methodology to ensure the power and potential of this approach in image classification domain.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

MHWF-CNN: multiscale horizontal wavelet fusion convolutional neural network with transfer learning for image classification

  • 期刊论文

  • 发表日期：

    2025年6月5日

查看全文
  • S. Kavitha,
  • H. Hannah Inbarani

摘要

This study introduces an innovative methodology that synergizes wavelet-based feature extraction with multiscale fusion and deep learning, aiming to elevate the image classification task. Initially, the Haar wavelet transform is applied to lung X-ray images, extracting approximation, horizontal, vertical, and diagonal sub-band coefficients. Following this, a multi-scale horizontal fusion technique integrates wavelet-transformed images at different resolutions with the original image. This process creates an enriched feature map with essential contextual information for the subsequent deep-learning model. The resulting image features serve as input for the deep learning models, encompassing conventional Convolutional Neural Networks and transfer learning models such as ResNet50 and VGG19. The Experimental demonstrates the proposed MHWF-CNN method performance in terms of accuracy that using the publicly accessible Chest X-ray dataset. The proposed method attains 99% accuracy with the integration of advanced methodology to ensure the power and potential of this approach in image classification domain.