Tunable wavelet filters improve image classification and anomaly detection

Layerwise Tunable Lifting Scheme for the Convolutional Neural Network

Computer Vision and Pattern Recognition

Summary

This paper tackles a problem in computer vision where machines need to recognize patterns and detect unusual items in images. The authors introduce a way to adjust the filters inside neural networks to better capture details at different frequencies. They create three tuning strategies that keep the system stable and reversible. Testing their approach in standard image recognition and anomaly detection tasks shows that it improves performance consistently.

What this means in practice

  • For computer vision engineers: Use layerwise tunable wavelet filters to enhance feature extraction in image classification models like ResNet-18.
  • For quality control teams: Detect anomalies more reliably in industrial image inspection by integrating tunable lifting schemes into existing CNN setups.

Authors

Abdumannon Yovkochov, An Le, Sungbal Seo, You-Suk Bae, Truong Nguyen

Abstract

This work introduces a family of tunable lifting schemes for biorthogonal wavelet filter banks. We propose three lifting strategies: low-pass tuning (LS-LayLatt-LP), high-pass tuning (LS-LayLatt-HP), and a sequential lifting scheme that jointly adapts low- and high-frequency branches (LS-LayLatt-Sequential). All proposed designs are formulated using a lattice-based lifting structure, which guarantees invertibility and stability for arbitrary parameter values within the lifting functions. We evaluated the proposed methods by integrating them into a ResNet-18 backbone for image classification on the Describable Textures Dataset (DTD), as well as for anomaly detection on hazelnut images from the MVTec-AD dataset and private KRC102S dataset. Experimental results demonstrate consistent performance improvements across all evaluated tasks.