Publications by authors named "Zunjie Zhu"
Invariant Content Representation for Generalizable Medical Image Segmentation.
J Imaging Inform Med
December 2024
Article Synopsis
- Domain generalization (DG) in medical image segmentation aims to improve model robustness when using limited source data while preserving privacy.
- Current methods primarily rely on global data augmentation, which leads to insufficient diversity and a tendency for models to overfit the source style.
- The proposed invariant content representation network (ICRN) introduces local style augmentation and invariant content learning to enhance sample diversity while suppressing style bias, resulting in significant performance improvements on cross-domain datasets.
Existing methods for single image super-resolution (SISR) model the blur kernel as spatially invariant across the entire image, and are susceptible to the adverse effects of textureless patches. To achieve improved results, adaptive estimation of the degradation kernel is necessary. We explore the synergy of joint global and local degradation modeling for spatially adaptive blind SISR.
View Article and Find Full Text PDFPlaneFusion: Real-Time Indoor Scene Reconstruction With Planar Prior.
IEEE Trans Vis Comput Graph
December 2022
Real-time dense SLAM techniques aim to reconstruct the dense three-dimensional geometry of a scene in real time with an RGB or RGB-D sensor. An indoor scene is an important type of working environment for these techniques. The planar prior can be used in this scenario to improve the reconstruction quality, especially for large low-texture regions that commonly occur in an indoor scene.
View Article and Find Full Text PDF