A5036002238- Advanced Vision and Imaging
- Image Enhancement Techniques
- Computer Graphics and Visualization Techniques
- Color Science and Applications
- Video Surveillance and Tracking Methods
- Generative Adversarial Networks and Image Synthesis
- Advanced Image and Video Retrieval Techniques
- Human Pose and Action Recognition
- Optical measurement and interference techniques
- Digital Media Forensic Detection
- Image Retrieval and Classification Techniques
- Advanced Image Processing Techniques
Northwestern Polytechnical University
2018-2024
Image relighting is attracting increasing interest due to its various applications. From a research perspective, im-age can be exploited conduct both image normalization for domain adaptation, and also data augmentation. It has multiple direct uses photo montage aesthetic enhancement. In this paper, we review the NTIRE 2021 depth guided challenge.We rely on VIDIT dataset each of our two challenge tracks, including information. The first track one-to-one where goal transform illumination...
Existing any-to-any relighting methods suffer from the task-aliasing effects and loss of local details in image generation process, such as shading attached-shadow. In this paper, we present PNRNet, a novel neural architecture that decomposes task into three simpler sub-tasks, i.e. lighting estimation, color temperature transfer, direction to avoid effects. These sub-tasks are easy learn can be trained with direct supervisions independently. To better preserve attached-shadow details,...
We review the AIM 2020 challenge on virtual image relighting and illumination estimation. This paper presents novel VIDIT dataset used in different proposed solutions final evaluation results over 3 tracks. The first track considered one-to-one relighting; objective was to relight an input photo of a scene with color temperature illuminant orientation (i.e., light source position). goal second estimate settings, namely orientation, from given image. Lastly, third dealt any-to-any relighting,...
Image harmonization aims at adjusting the appearance of foreground to make it more compatible with background. Without exploring background illumination and its effects on elements, existing works are incapable generating a realistic shading. In this paper, we decompose image task into two sub-problems: 1) estimation 2) re-rendering objects under illumination. Before solving these sub-problems, first learn shading-aware descriptor via well-designed neural rendering framework, which key is...
Shadows bring great challenges but also play essential roles in image understanding. Most recent shadow detection methods are based on patches, then further reasoning method is required for the obtaining of a completed result an image. In this paper, injection network proposed to detect regions whole directly. order maintain as many details, skip structure applied directly inject details from convolutional layers de-convolutional layers. Meanwhile, weighted loss function training. With...
Shadows actually play an important role in image understanding. But even for the same object, intensity and shape of shadows can vary with environment. Thus it is a quite changeling issue to detect remove from images. Recent studies have been trying solve these two tasks independently, but they are closely related each other actually. Therefore, we propose multi-task adversarial generative networks (mtGAN) that simultaneously. For proposed mtGAN, cross-stitch unit applied learn optimal ways...
Image harmonization aims at adjusting the appearance of foreground to make it more compatible with background. Without exploring background illumination and its effects on elements, existing works are incapable generating a realistic shading. In this paper, we decompose image task into two sub-problems: 1) estimation 2) re-rendering objects under illumination. Before solving these sub-problems, first learn shading-aware descriptor via well-designed neural rendering framework, which key is...
Accurate reconstruction of real-world materials' appearance from a very limited number samples is still huge challenge in computer vision and graphics. In this paper, we present novel deep architecture, Disentangled Generative Adversarial Network (DGAN), which performs anisotropic Bidirectional Reflectance Distribution Function (BRDF) single BRDF subspace with the maximum entropy. contrast to previous approaches that directly map known full using CNN, disentangled representation learning...