Single Image Super-Resolution of Noisy 3D Dental CT Images Using Tucker Decomposition

TL;DR

This paper compares Tucker and CPD tensor decompositions for single image super-resolution of noisy 3D dental CT images, demonstrating Tucker's advantages in noise robustness and accuracy.

Contribution

The study introduces a Tucker decomposition-based SISR method, highlighting its improved noise handling and segmentation accuracy over CPD in dental CT imaging.

Findings

Tucker decomposition improves noise robustness in SISR.

The proposed method achieves higher PSNR and SSIM scores.

Tucker-based approach reduces runtime compared to CPD.

Abstract

Tensor decomposition has proven to be a strong tool in various 3D image processing tasks such as denoising and super-resolution. In this context, we recently proposed a canonical polyadic decomposition (CPD) based algorithm for single image super-resolution (SISR). The algorithm has shown to be an order of magnitude faster than popular optimization-based techniques. In this work, we investigated the added value brought by Tucker decomposition. While CPD allows a joint implementation of the denoising and deconvolution steps of the SISR model, with Tucker decomposition the denoising is realized first, followed by deconvolution. This way the ill-posedness of the deconvolution caused by noise is partially mitigated. The results achieved using the two different tensor decomposition techniques were compared, and the robustness against noise was investigated. For validation, we used dental images. The superiority of the proposed method is shown in terms of peak signal-to-ratio, structural similarity index, the canal segmentation accuracy, and runtime.