TY - GEN
T1 - Scene Text Eraser
AU - Nakamura, Toshiki
AU - Zhu, Anna
AU - Yanai, Keiji
AU - Uchida, Seiichi
N1 - Funding Information:
VI. ACKNOWLEDGMENTS This research was partially supported by MEXT-Japan (Grant No. 17H06100). Thanks to the contributors of the pictures taken from Fickr with the copyright license. Left bottom of Fig.5(a) and Left top of Fig.5(b) : alykat
Publisher Copyright:
© 2017 IEEE.
PY - 2017/7/2
Y1 - 2017/7/2
N2 - The character information in natural scene images contains various personal information, such as telephone numbers, home addresses, etc. It is a high risk of leakage the information if they are published. In this paper, we proposed a scene text erasing method to properly hide the information via an inpainting convolutional neural network (CNN) model. The input is a scene text image, and the output is expected to be text erased image with all the character regions filled up the colors of the surrounding background pixels. This work is accomplished byaCNNmodelthroughconvolutiontodeconvolutionwithinterconnection process. The training samples and the corresponding inpainting images are considered as teaching signals for training. To evaluate the text erasing performance, the output images are detected by a novel scene text detection method. Subsequently, the same measurement on text detection is utilized for testing the images in benchmark dataset ICDAR2013. Compared with direct text detection way, the scene text erasing process demonstrates a drastically decrease on the precision, recall and f-score. That proves the effectiveness of proposed method for erasing the text in natural scene images.
AB - The character information in natural scene images contains various personal information, such as telephone numbers, home addresses, etc. It is a high risk of leakage the information if they are published. In this paper, we proposed a scene text erasing method to properly hide the information via an inpainting convolutional neural network (CNN) model. The input is a scene text image, and the output is expected to be text erased image with all the character regions filled up the colors of the surrounding background pixels. This work is accomplished byaCNNmodelthroughconvolutiontodeconvolutionwithinterconnection process. The training samples and the corresponding inpainting images are considered as teaching signals for training. To evaluate the text erasing performance, the output images are detected by a novel scene text detection method. Subsequently, the same measurement on text detection is utilized for testing the images in benchmark dataset ICDAR2013. Compared with direct text detection way, the scene text erasing process demonstrates a drastically decrease on the precision, recall and f-score. That proves the effectiveness of proposed method for erasing the text in natural scene images.
UR - http://www.scopus.com/inward/record.url?scp=85045197884&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85045197884&partnerID=8YFLogxK
U2 - 10.1109/ICDAR.2017.141
DO - 10.1109/ICDAR.2017.141
M3 - Conference contribution
AN - SCOPUS:85045197884
T3 - Proceedings of the International Conference on Document Analysis and Recognition, ICDAR
SP - 832
EP - 837
BT - Proceedings - 14th IAPR International Conference on Document Analysis and Recognition, ICDAR 2017
PB - IEEE Computer Society
T2 - 14th IAPR International Conference on Document Analysis and Recognition, ICDAR 2017
Y2 - 9 November 2017 through 15 November 2017
ER -