Exploiting Color Information for Better Scene Text Recognition

Muhammad Fraz, Muhammad Sarfraz and Eran Edirisinghe

In Proceedings British Machine Vision Conference 2014


This paper presents an approach to text recognition in natural scene images. The main contribution of this paper is to show that the colour information within the images if efficiently exploited is good enough to identify text regions from the surrounding noise. In the same way, the colour information in character and word images can be used to significantly improve character and eventually word recognition accuracy. We propose simple operations that use colour information and low level processing to enhance relevant text information in the images that eventually benefits to increase the word recognition performance. We show that our method provides significantly enhanced candidate text regions, so that, even a simple off the shelf feature representation and classification method can achieve state-of-the-art performance. We performed extensive experiments to evaluate our method on challenging datasets (Chars74K, ICDAR�03, ICDAR�11 and SVT) and results show that our method outperforms the current state-of-the-art on scene character and word recognition.


Poster Session


Extended Abstract (PDF, 1 page, 142K)
Paper (PDF, 12 pages, 557K)
Bibtex File


Muhammad Fraz, Muhammad Sarfraz, and Eran Edirisinghe. Exploiting Color Information for Better Scene Text Recognition. Proceedings of the British Machine Vision Conference. BMVA Press, September 2014.


	title = {Exploiting Color Information for Better Scene Text Recognition},
	author = {Fraz, Muhammad and Sarfraz, Muhammad and Edirisinghe, Eran},
	year = {2014},
	booktitle = {Proceedings of the British Machine Vision Conference},
	publisher = {BMVA Press},
	editors = {Valstar, Michel and French, Andrew and Pridmore, Tony}
	doi = { http://dx.doi.org/10.5244/C.28.84 }