Fuzzy Histogram of Oriented Gradients to characterize Single Line Text Regions

In this work we discuss the use of the histogram of oriented gradients (HOG) descriptors as an effective tool for text description and recognition. Specifically, we propose a Fuzzy HOG-based texture descriptor (F-HOG) that uses a partition of the image into three horizontal cells with fuzzy adaptive boundaries, to characterize single-line texts in outdoor scenes and video frames. The input of our algorithm is a rectangular image presumed to contain a single line of text in latin like characters. The output is a relatively short (54-features) descriptor that provides an effective input to an SVM classifier. Tests show that F-HOG is more accurate than Dalal and Triggs original HOG-based classifier using a 54-features descriptor, and comparable to their best classifier (which uses a 108-features descriptor) while being half as long.

2011