New fusion based enhancement for text detection in night video footage
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Text Detection in night video footage is hard due to low contrast and low resolution caused by distance variations between camera and ground under poor light. In this paper, we propose a new fusion based enhancement method for text detection especially in night video footage. The proposed method integrates the merits of color space and frequency based enhanced methods for sharpening low contrast details. Specifically, for each enhanced image, the proposed method derives weighted mean for the pixels values to widen the gap between high contrast (texts) and low contrast (background) pixels. The weighed means are further modified as dynamic weights with respect to enhanced images. These weights are convolved with pixel values of respective enhanced images to produce fused images. The proposed fusion based enhancement method is tested on images collected from night video footage to demonstrate the effectiveness of the method. For the output of each enhancement method including the proposed method, text detection rates are computed to show that the proposed enhancement method outperforms the existing enhancement methods.
Zhang, Chao; Shivakumara, Palaiahnakote; Xue, Minglong; Zhu, Liping; Lu, Tong; and Pal, Umapada, "New fusion based enhancement for text detection in night video footage" (2018). Conference Articles. 139.