A Non-Blind Deconvolution Semi Pipelined Approach to Understand Text in Blurry Natural Images for Edge Intelligence
作者：
Highlights：
• The text images are synthetically blurred as per degradation model described above to generate blurry dataset.
• To recover a clear text image from blurry dataset, a non-blind deconvolution approach is adapted.
• Text localization and edge-based connected component text extraction is proposed.
• Separate foreground from background followed by proposed CL-CNN for deep feature extraction.
• Classification of recovered images into text class and non-text class from unseen data.
摘要
•The text images are synthetically blurred as per degradation model described above to generate blurry dataset.•To recover a clear text image from blurry dataset, a non-blind deconvolution approach is adapted.•Text localization and edge-based connected component text extraction is proposed.•Separate foreground from background followed by proposed CL-CNN for deep feature extraction.•Classification of recovered images into text class and non-text class from unseen data.
论文关键词：Synthetic blur,Non-blind deconvolution,Edge Intelligence,2D Radix-4 DIT
论文评审过程：Received 15 April 2021, Revised 14 June 2021, Accepted 25 June 2021, Available online 30 July 2021, Version of Record 30 July 2021.
论文官网地址：https://doi.org/10.1016/j.ipm.2021.102675
