A Hybrid Approach for Complex Layout Detection of Newspapers in Gurumukhi Script Using Deep Learning

Abstract

Layout analysis is the crucial stage in the recognition system of newspapers. A good layout analysis results in better recognition results. The complexity of newspaper layout structures poses a formidable challenge in digitization. The intricate arrangement of text, images, and various sections within a newspaper demands sophisticated algorithms and techniques for accurate layout detection. The paper introduces a diverse set of methodologies from existing literature, highlighting the evolution of techniques for newspaper layout analysis. In this paper, we present a novel method to detect the complex layout of newspapers in the Gurumukhi script by using a hybrid approach. The method developed consists of two parts. In the first part, we proposed an algorithm to remove pictures and graphics from Punjabi newspaper images that involve various image preprocessing tasks based on binarization, finding contours, and erosion on the image to remove the graphics from the image. This method removes pictures from complex non-Manhattan layouts. We have tested this algorithm on 100 newspaper images, giving an accuracy of 96.22%. In the second part, a dataset of 500 newspapers was created with images labeled with five classes on which the model was trained. Finally, we have trained the deep-leaning model based on a convolutional network to detect the columns of text in newspapers. We have used four different architectures of CNN and compared their performance based on different metrics such as precision, recall, and F1 score. We have tested this method on a number of newspapers in the Gurumukhi script. We have achieved an accuracy of 95.53% with this approach.