[Paper Review] An Optimal Method For Wake Detection In SAR Images Using Radon Transformation Combined With Wavelet Filters

M. Krishnaveni, Suresh Kumar Thakur+1 more|ArXiv.org|Nov 3, 2009

Image and Signal Denoising Methods10 references8 citations

TL;DR

This paper proposes an optimal ship wake detection method in SAR images by combining Radon Transform (RT) with wavelet-based shrinkage denoising to enhance signal-to-noise ratio. The approach improves line detection in noisy SAR imagery by preprocessing with discrete wavelet transform (DWT) followed by thresholding, then applying RT to detect linear wake structures, achieving superior performance over conventional methods in both simulated and real SAR data.

ABSTRACT

A new fangled method for ship wake detection in synthetic aperture radar (SAR) images is explored here. Most of the detection procedure applies the Radon transform as its properties outfit more than any other transformation for the detection purpose. But still it holds problems when the transform is applied to an image with a high level of noise. Here this paper articulates the combination between the radon transformation and the shrinkage methods which increase the mode of wake detection process. The latter shrinkage method with RT maximize the signal to noise ratio hence it leads to most optimal detection of lines in the SAR images. The originality mainly works on the denoising segment of the proposed algorithm. Experimental work outs are carried over both in simulated and real SAR images. The detection process is more adequate with the proposed method and improves better than the conventional methods.

Motivation & Objective

• To address the challenge of detecting ship wakes in noisy SAR images where conventional Radon Transform fails due to high noise levels.
• To improve the signal-to-noise ratio (SNR) in SAR images prior to wake detection using wavelet-based denoising techniques.
• To develop a hybrid method combining Radon Transform and wavelet shrinkage for optimal detection of linear structures (wakes) in SAR imagery.
• To validate the method’s effectiveness on both simulated and real SAR image datasets, demonstrating improved detection accuracy.

Proposed method

• Apply discrete wavelet transform (DWT) to decompose SAR images into approximation and detail coefficients to separate noise from signal components.
• Use soft-thresholding on wavelet detail coefficients to suppress noise while preserving significant edge and line features.
• Reconstruct the denoised image from the thresholded coefficients to produce a cleaner input for Radon Transform.
• Apply Radon Transform to the denoised image to detect linear structures by identifying peaks in the Radon domain corresponding to wake lines.
• Optimize the thresholding level in the wavelet domain to maximize SNR and detection accuracy.
• Combine the denoising and Radon-based detection pipeline into a single, coherent algorithm for automated wake detection.

Experimental results

Research questions

• RQ1 Can wavelet-based denoising effectively improve the signal-to-noise ratio in SAR images before Radon Transform application?
• RQ2 How does the integration of wavelet shrinkage with Radon Transform enhance the detection of linear ship wake patterns in noisy SAR data?
• RQ3 Does the proposed method outperform conventional Radon Transform-based detection in terms of accuracy and robustness on real and simulated SAR images?
• RQ4 What is the optimal thresholding strategy in the wavelet domain to maximize wake detection performance?

Key findings

• The proposed method significantly improves wake detection performance in SAR images with high noise levels compared to conventional Radon Transform alone.
• Wavelet-based denoising effectively suppresses speckle and random noise, enhancing the SNR before Radon Transform application.
• The combination of wavelet shrinkage and Radon Transform leads to more accurate and reliable detection of linear wake structures in both simulated and real SAR images.
• The method achieves better detection results than standard approaches, as demonstrated through comparative evaluation on test datasets.