DCT-DWT-FFT Based Method for Text Detection in Underwater Images

Document Type

Conference Article

Publication Title

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)


Text detection in underwater images is an open challenge because of the distortions caused by refraction, absorption of light, particles, and variations depending on depth, color, and nature of water. Unlike existing methods aimed at text detection in natural scene images, in this paper, we have proposed a novel method for text detection in underwater images through a new enhancement model. Based on observations that fine details of text in image share with high energy, spatial resolution, and brightness, we consider Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), and Fast Fourier Transform (FFT) for image enhancement to highlight the text features. The enhanced image is fed to a modified Character Region Awareness for Text Detection (CRAFT) model to detect text in underwater images. To explore enhancement methods, we evaluate six combinations of image enhancement techniques, namely, DCT-DWT-FFT, DCT-FFT-DWT, DWT-DCT-FFT, DWT-FFT-DCT, FFT-DCT-DWT, FFT-DWT-DCT. Experimental results on our dataset of underwater images and benchmark datasets of natural scene text detection, namely, MSRA-TD500, ICDAR 2019 MLT, ICDAR 2019 ArT, Total-Text, CTW1500, and COCO Text show that the proposed method performs well for both underwater and natural scene images and outperforms the existing methods on all the datasets.

First Page


Last Page




Publication Date


This document is currently not available here.