Year: 2024
Type: article
Abstract: Non-uniform illumination underwater image occasionally produced during image acquisition. Underwater image has low visibility due to light absorption and dispersion, therefore by using an artificial light source to increase visibility can occasionally result in different illumination regions in the image. In order to generate uniform illumination image and restore the loss details, image enhancement must be performed. Despite the fact that Histogram Equalization (HE) is well known and widely used in image enhancement, current HE-based approaches usually produce washed-out effects and have an unnatural appearance due to the over-enhancement problem, limiting their ability to achieve image illumination uniformity. Therefore, this study proposes a modified HE approach for enhancing underwater image. Firstly, the underwater image is divided into three illumination regions. Then, the histogram of each region is modified and the modified HE transformation function is performed. 100 non-uniform illumination underwater images were utilised to assess the performance of the proposed modified HE approach. The proposed approach is compared with three state-of-the-art HE approaches. Qualitatively, the proposed approach produces more uniform illumination image, able to preserve the naturalness and retrieve the details on the image. Meanwhile, the proposed approach obtained the best scores in three quantitative measurements namely as Discrete Entropy (DE), Image Contrast Function (ICF), Absolute Mean Brightness Error (AMBE) and Peak Signal to Noise Ratio (PSNR). From both analyses, the proposed approach able to produce more uniform illumination image, enhancing the image details, and preserving the image naturalness.
Authors: Nor Hidayah Saad, Nor Ashidi Mat Isa, Hariyanti Mohd Saleh, Farah Hanim Mohd Fauzi, Shivam Gangwar
Language: English
Topic: Image Enhancement Techniques
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4408481949