Mr Jhuntu and Tathagata Roy Chowdhury*

Department of Computer Science and Engineering, St. Mary’s Technical Campus Kolkata, Kolkata, West Bengal, India

*Corresponding Author: Tathagata Roy Chowdhury, Department of Computer Science and Engineering, St. Mary’s Technical Campus Kolkata, Kolkata, West Bengal, India.

Received: September 23, 2022; Published: October 13, 2022

Abstract

Images of nature scenes that contain language contain useful information, such as text-based landmarks, etc. There are several steps involved in text extraction from scene photos. To get effective results, each stage is equally vital. The steps of text detection, localization, segmentation, and recognition are crucial. Due to differences in size, position, and alignment from one image to the next, extracting text from scene photographs is exceedingly challenging. The challenge of extracting text from scene photos is difficult because of all these issues. However, the placement of text in such images is arbitrary and is not restricted to any particular page layout. We used the dataset made available in conjunction with IIIT5K to assess how well the planned "segmentation of scene text images using connected component analysis" performed. Since each image comprises roughly four characters in various font styles and font sizes. A disjunct categorization does not seem to be possible, nevertheless, as even two segmentation procedures that are utterly unrelated can share traits that defy singular categorization 1. As a result, the classification being presented is one that takes into account an approach's emphasis rather than being a perfect split.

Keywords: Image Processing; Connected Components; Image; Binary Image; Segmentation; Grey Scale