An Efficient Deep Neural Network with Adaptive Galactic Swarm Optimization for Complex Image Text Extraction
Abstract
Text present in a complex image contains various details which are employed to recognize the conditions of the image and to nontext details. The details available in the complex degraded image are found to be important for people to see the entire situation. But the text in complex degraded images shows a highly dynamic form in an unrestricted situation that makes text visual recognition a challenge. Therefore, the Naïve Bayes algorithm (NBA) weighted reading method is used in this chapter to get the right textual information from the complex degraded pictures. Normally, images contain some sound as noise, which is the fact that in the initial preprocessing phase, the guided filter (GF) is introduced. A very important feature in the separation process is extracted using different techniques such as Gabor transform (GT) and stroke width transform (SWT), and along with all related features, text identification and recognition is done by Weighted Naïve Bayes Algorithms and a high-quality learning process for good practice. Subsequently, performance parameters such as accuracy, F1 scores and precision were tested using the IIIT5K database to measure the effectiveness.