Calibration of blurred text in images using deep learning techniques

Text in an image carries a high level of information for humanity. These texts have an
important role in the computer vision application. In recent years, the rapid development
of machine learning and deep learning, in
uence the applications of computer vision and
document analysis topics. All the previously proposed methods use dierent algorithms
to detect text in images; however, they suer from poor performance while performing
detection in blurred images. The proposed method capable of handling blurred text
detection and recognition in images. It is an automatic end to end system to recognize the
blurred text in images. It has two stages, rst is the detection of text in an image using
an object detection method. In the second step, it segments the text area and recognizes
it using hybrid CNN and LSTM method. It acquires 92% accuracy in detection and 93%
accuracy in the recognition phase.

MPhil CS 2018-2019 INT Dr Elizabeth Sherly