A blind deconvolution model for scene text detection and recognition in video
作者：
Highlights：
• We explore quality metrics for blur text image/video classification.
• Proposed a deblur model explores Gaussian weighted-L1 in different ways.
• Kernel based energy minimization enhances the edge strengths.
• Experiments to evaluate and validate the proposed deblur model are presented.
• Experimental result shows that the proposed method is useful for text detection.
摘要
•We explore quality metrics for blur text image/video classification.•Proposed a deblur model explores Gaussian weighted-L1 in different ways.•Kernel based energy minimization enhances the edge strengths.•Experiments to evaluate and validate the proposed deblur model are presented.•Experimental result shows that the proposed method is useful for text detection.
论文关键词：Text detection,Text recognition,Blind deconvolution,Alternative minimization,Text restoration
论文评审过程：Received 11 May 2015, Revised 8 November 2015, Accepted 4 January 2016, Available online 18 January 2016, Version of Record 27 February 2016.
论文官网地址：https://doi.org/10.1016/j.patcog.2016.01.008
