Application of daisy descriptor for language identification in the wild

  • PDF / 6,309,500 Bytes
  • 22 Pages / 439.37 x 666.142 pts Page_size
  • 86 Downloads / 219 Views

DOWNLOAD

REPORT


Application of daisy descriptor for language identification in the wild Neelotpal Chakraborty 1 & Agneet Chatterjee 1 & Pawan Kumar Singh 2 Ayatullah Faruk Mollah 3 & Ram Sarkar 1

&

Received: 23 June 2019 / Revised: 7 July 2020 / Accepted: 25 August 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract

Recent years have witnessed significant development in the field of text detection in natural scene images. However, issues like poor image quality and complex background reduce the efficiency of such methods, thereby requiring a good pre-processing module for image enhancement. Also, conventional texture-based features have some limitations for classifying text and non-text components due to potential similarities between them. To this end, a new model is proposed where the image quality is first enhanced by removing noise and blur. Then, a histogram-based adaptive K-means clustering of intensity values is performed in order to extract the text candidates. These candidates are then analyzed using Daisy descriptor for text/non-text determination, and language identification of the text. The proposed model is applied on an in-house multi-lingual dataset of images with texts in Indian languages, and on standard datasets including ICDAR 2017, MLe2e and KAIST. The results indicate significant improvement in performance compared to some contemporary methods. Keywords Scene text . Multi-lingual . Text/non-text classification . Language identification . Histogram-based adaptive K-means clustering . Daisy descriptor

* Pawan Kumar Singh [email protected] Neelotpal Chakraborty [email protected] Agneet Chatterjee [email protected] Ayatullah Faruk Mollah [email protected] Ram Sarkar [email protected] Extended author information available on the last page of the article

Multimedia Tools and Applications

1 Introduction In the recent times, retrieval of textual information from image data has garnered significant interest among researchers thereby resulting in development of several state-of-the art techniques to detect text in the wild. This rapidly growing interest on such a challenging research problem is seen due to potential applications like tour guide assistance, information retrieval, query processing, image to text conversions, etc. [25]. The processes involved in scene text detection [25], require management of several factors related to image quality improvement like resolution adjustment, noise removal, blur reduction, dealing with intensity variation [10] and so on, which are illustrated in Fig. 1. The usage of vibrant colors and fonts to illustrate texts in signboards, posters and advertisements in a culturally diverse country like India, is quite common. Hence, the complexities of text containing images get augmented and become difficult to deal with. Currently popular methods involve stability analysis with respect to intensity variation and measurement of stroke width of the stable components. However, these methods falter due to poor image quality an