In this section, you will find the final results of the ICDAR2015 Competition on Text Image Super-Resolution.
Team /
Method |
RMSE |
PSNR |
SSIM |
OCR accuracy |
Bicubic |
19.04 |
23.50 |
0.879 |
60.64 |
ASRS [1] |
12.86 |
26.98 |
0.95 |
71.25 |
SRCNN-1 [2] |
7.52 |
31.75 |
0.98 |
77.19 |
SRCNN-2 [2] |
7.24 |
31.99 |
0.981 |
76.10 |
Synchromedia
Lab [3] |
62.67 |
12.66 |
0.623 |
65.93 |
Original HR |
- |
- |
- |
78.80 |
[1] based on R. Walha, F. Drira, F. Lebourgeois, C. Garcia, and A. Alimi, "Resolution enhancement of textual images via multiple coupled dictionaries and adaptive sparse representation selection," IJDAR, vol. 18, no. 1, pp. 87-107, 2015.
[2] based on C. Dong, C. C. Loy, K. He, and X. Tang, "Learning a deep convolutional network for image super-resolution," in ECCV 2014. Springer, 2014, pp. 184-199.
[3] based on R. Farrahi Moghaddam and M. Cheriet, "A multi-scale framework for adaptive binarization of degraded document images," Pattern Recogn., vol. 43, no. 6, pp. 2186-2198, 2010.