Deep Learning-Based Scene Text Image Super-Resolution Methods: A Survey
Shuo Jiang, Bo Ning, Yunpeng Wang, Haocheng Zhang, Yingying Diao, Peize Yang · 2023
Text recognition techniques have yielded positive results in identifying clear text images. However, recognition of low-resolution or fuzzy text may lead to a significant degradation of performance. The main challenge in recognizing fuzzy text lies in the absence of detailed information pertaining to the text. Unfortunately, low-resolution text images are often encountered in natural scenes, such as distant images captured by mobile phones, pictures taken under severe lighting, or in the presence of rain or fog. With the recent developments in deep learning, an increasing number of researchers have focused on computer vision, leading to greater exposure of super-resolution techniques. This paper aims to compare various models of scene text image super-resolution and elucidates on its evolution in recent years.