HDD-Net: Haar Dual Domain Network for Ring Artifacts Correction

Ring artifacts are common artifacts in X-ray Computed Tomography (XCT) scans and have a significant impact on subsequent feature/phase extractions due to the small grayscale gradients in XCT volume data of bulk materials. This paper proposes the Haar Dual Domain Network for correcting ring artifacts. By utilizing the Haar wavelet decomposition on images containing ring artifacts in both the image and projection domains, the ring artifacts are preliminarily separated, facilitating their removal by neural networks while preserving microstructure features such as low-contrast phase boundaries. By constructing a feature fusion network, the information from both 2D slices and 3D projection volume data has been fully integrated to eliminate ring artifacts while preserving the edges of every feature. The effectiveness of the Haar wavelet transform and fusion network has been validated by ablation experiments, proving the application of HDD-Net to large volume of XCT data.

I. Introduction

Computed Tomography (CT) technology, after long-term development, is now widely applied in fields such as medicine and materials [1]. However, factors such as inconsistent detector gain errors, non-linear detector responses, and variations in temperature or beam intensity can result in ring artifacts in the 3D volume after reconstruction [2]. In 2D slices, the image shows that the deviation of its circular center from the focal point. These ring artifacts significantly degrade the quality of slice images, profoundly impacting the observation, analysis, and three-dimensional reconstruction of slice images [3]. To mitigate the occurrence of ring artifacts, hardware-based calibration of CT equipment is typically performed before scanning [6] to correct detector errors and eliminate ring artifacts. Additionally, flat field correction, as a commonly employed error correction technique, can effectively eliminate ring artifacts [7], [8]. Nevertheless, the unnecessary consumption of X-ray just for correcting the images is costly. Implementing hardware-based calibration methods and flat field correction incurs additional expenses. Furthermore, this approach necessitates the collection of detector data, making implementation relatively challenging. Moreover, different detector elements possess distinct response functions, rendering flat field correction methods incapable of completely removing ring artifacts. In contrast, post-processing methods tailored to 2D slices offer advantages such as low cost, broad applicability, and excellent correction efficacy, thus finding widespread use in ring artifacts correction.