An Enhanced Character Centric Approach for Text Spotting Using Deep Learning Techniques

Srikanth Bhyrapuneni, P. Swaminathan and R. Anandan

Text identification is considered as an imperative issue in the couple of years. Importance in the field of computer vision and machine learning just as increment in the applications dependent on Text identification and acknowledgment has brought about this pattern. Creating Text spotting frameworks, fit for checking and in this way better translating the visual world, is a challenging task however a helpful undertaking to understand. Deep learning based models, comprising of a huge number of trainable parameters, require a great deal of information to prepare adequately. In the character-driven methodology, various character classifier models are produced, fortifying each other through a component sharing system. These character models are utilized to create Text pixel maps to drive identification of text. In this part we present a Text spotting framework Image Text Pixel Recognition (ITPR) adopting a character-driven strategy. Here characters are considered as the building blocks for Text, displaying them clearly, and developing Text lines and words by gathering character location together. The proposed method is compared with Discrete Cosine Transformation (DCT) method and the results exhibit that the proposed method is more accurate.

Volume 11 | 04-Special Issue

Pages: 690-699