A5091921038- Advanced Vision and Imaging
- Privacy-Preserving Technologies in Data
- Imbalanced Data Classification Techniques
- Advanced Image Processing Techniques
- Machine Learning and Data Classification
- Computer Graphics and Visualization Techniques
- AI in cancer detection
- Image Enhancement Techniques
- Image Processing Techniques and Applications
- Advanced Image and Video Retrieval Techniques
Huazhong University of Science and Technology
2017-2023
Federated noisy label learning (FNLL) is emerging as a promising tool for privacy-preserving multi-source decentralized learning. Existing research, relying on the assumption of class-balanced global data, might be incapable to model complicated noise, especially in medical scenarios. In this paper, we first formulate new and more realistic federated noise problem where data class-imbalanced heterogeneous, then propose two-stage framework named FedNoRo noise-robust Specifically, stage...
The frequency estimation of the plenoptic function (POF) is an important research topic in spectral analysis for determining minimum sampling rate image-based rendering. In this paper, we mathematically derive a POF using autocorrelation theory. (ACF) studied along both spatial and image plane axes. influence scene's complexity depth on ACF analyzed. Furthermore, study error to analyze performance method. Existing techniques typically use Fourier transformation determine POF. technique...
We present a signal-processing framework for image-assisted geometry measurement in the image-based rendering (IBR). study utilized information and estimating minimum sampling rate of IBR. Our method combines decomposing complex scene into collection simpler structures on block-by-block basis. The automatic structure selection can be interactively refined by detected single salient points. In this manner, we reduce spectral analysis problem an irregular object to that structure. Predictions...
Scene surface exists some occlusions by mutual or self-occlusions, and then this phenomenon will serious influences 3D vision technique. We present a signal-processing framework to study of scene. Our method combines discontinuities establish mathematical model occlusion phenomenon. The on scene are derived using Fourier theory. In manner, spectral support is analyzed in frequency field. Predictions the content can be used control sampling rendering This extends previous work that estimated...
Federated noisy label learning (FNLL) is emerging as a promising tool for privacy-preserving multi-source decentralized learning. Existing research, relying on the assumption of class-balanced global data, might be incapable to model complicated noise, especially in medical scenarios. In this paper, we first formulate new and more realistic federated noise problem where data class-imbalanced heterogeneous, then propose two-stage framework named FedNoRo noise-robust Specifically, stage...