An Application of LBF Energy in Image/Video Frame Text Detection

In the present paper, we propose a new method for detecting texts in image and video frames. Initially, texture features are extracted using wavelet transform. Further, local binary fitting energy is introduced to obtain local image intensity information. In addition to that variational level set function is integrated to detect an accurate boundary of the texts present in images/video frames. The proposed method is able to detect a region of text with high complexity as well low intensity texts and also detects a text region in varying background. Experiment is conducted on ICDAR datasets and 101 video images dataset. Performance comparison shows that the proposed method exhibits better results in terms of computational efficiency and accuracy.