Texture Image Classification Method of Porcelain Fragments Based on Convolutional Neural Network

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Traditional image decomposition methods struggle with complex textures and noise.
• Convolutional Neural Networks (CNNs) offer powerful tools for image analysis and feature extraction.
• Variational methods and compressed sensing are advanced techniques for image reconstruction.

Purpose of the Study:

• To develop an energy minimization-based algorithm for texture image decomposition of porcelain fragments.
• To systematically research image decomposition using variational methods and CNN-based compressed sensing reconstruction.
• To enhance the retention of image texture details and improve reconstruction quality for big data.

Main Methods:

• Layered variational image decomposition to separate structural and texture components.
• Compressed sensing reconstruction utilizing a hybrid basis and a joint sparse dictionary (wavelet-based shearlet transform and wave atoms).
• Image-based normalization and a super-resolution model for texture feature restoration.

Main Results:

• The proposed algorithm effectively decomposes images into structure, texture, and noise.
• The compressed sensing approach with a joint sparse dictionary retains more texture details at the same sampling rate.
• A 0.1 average improvement in Peak Signal-to-Noise Ratio (PSNR) was achieved with the classification-extracted dataset.

Conclusions:

• The developed method offers a robust approach for texture image decomposition and reconstruction.
• The algorithm shows promise for improving image restoration and detail preservation in big data scenarios.
• Standardized datasets and normalization techniques are crucial for future big data recognition and experimental training.