Improved Wavelet Threshold Transform for SAR Image Oil Spill Detection

EasyChair Preprint 8655

Abstract

Big data shows that offshore oil spills have been on the rise in recent years. Oil spills at sea can be monitored using SAR images, which can assist in preventing the economic damage and pollution caused by spills. Detecting offshore oil spills with SAR images is essentially a segmentation of oil spill images. However, reliably distinguishing the oil spill location from the clean sea surface area using SAR photos is a huge challenge. Considering that the SAR image itself has multiplicative noise, the traditional threshold segmentation algorithm has many defects. To overcome this challenge, methods based on a wavelet threshold transform and the Otsu segmentation algorithm were applied. Therefore, this study is devoted to enhancing the denoising effect of wavelet threshold transform, so as to further improve the segmentation accuracy of oil spill area and clean sea area. In this study, a new hierarchical adaptive threshold and a threshold function with bi-directional shrinkage are proposed to handle wavelet coefficients. While removing the SAR image noise, the edge details of the oil spill area can be retained. Experiments demonstrate that the suggested strategy enhances overall denoising and segmentation accuracy significantly.

Keyphrases: Oil spill detection, SAR images, image segmentation, threshold function, wavelet threshold transformation

@booklet{EasyChair:8655,
  author    = {Siyuan Chen and Xueyun Wei and Wei Zheng},
  title     = {Improved Wavelet Threshold Transform for SAR Image Oil Spill Detection},
  howpublished = {EasyChair Preprint 8655},
  year      = {EasyChair, 2022}}