A5077375508- Image and Signal Denoising Methods
- Advanced Image Processing Techniques
- Image Processing Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
- Image Retrieval and Classification Techniques
- Handwritten Text Recognition Techniques
- Advanced Image Fusion Techniques
- Computer Graphics and Visualization Techniques
Southeast University
2024
Scene text image super-resolution (STISR) aims to simultaneously increase the resolution and legibility of images, resulting images will significantly affect performance downstream tasks. Although numerous progress has been made, existing approaches raise two crucial issues: (1) They neglect global structure text, which bounds semantic determinism scene text. (2) The priors, e.g., prior or stroke prior, employed in works, are extracted from pre-trained recognizers. That said, such priors...
Real-world text can be damaged by corrosion issues caused environmental or human factors, which hinder the preservation of complete styles texts, e.g., texture and structure. These issues, such as graffiti signs incomplete signatures, bring difficulties in understanding thereby posing significant challenges to downstream applications, scene recognition signature identification. Notably, current inpainting techniques often fail adequately address this problem have restoring accurate images...
Real-world text can be damaged by corrosion issues caused environmental or human factors, which hinder the preservation of complete styles texts, e.g., texture and structure. These issues, such as graffiti signs incomplete signatures, bring difficulties in understanding thereby posing significant challenges to downstream applications, scene recognition signature identification. Notably, current inpainting techniques often fail adequately address this problem have restoring accurate images...
Scene text image super-resolution (STISR) aims to simultaneously increase the resolution and legibility of images, resulting images will significantly affect performance downstream tasks. Although numerous progress has been made, existing approaches raise two crucial issues: (1) They neglect global structure text, which bounds semantic determinism scene text. (2) The priors, e.g., prior or stroke prior, employed in works, are extracted from pre-trained recognizers. That said, such priors...
Scene text image super-resolution (STISR) aims at simultaneously increasing the resolution and readability of low-resolution scene images, thus boosting performance downstream recognition task. Two factors in semantic information visual structure, affect significantly. To mitigate effects from these factors, this paper proposes a Prior-Enhanced Attention Network (PEAN). Specifically, diffusion-based module is developed to enhance prior, hence offering better guidance for SR network generate...