Deep Denoising Method for Side Scan Sonar Images without High-quality Reference Data

Zhou, Xiaoteng; Yu, Changli; Yuan, Xin; Luo, Citong

doi:10.1109/icccr54399.2022.9790164

Cited by 5 publications

(4 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Chen et al [14] proposed an ANLResNet model for sonar image despeckling by combining SRResNet with non-local blocks of asymmetrical pyramids for speckle noise in sonar images. Zhou et al [15] proposed a self-supervised denoising method for sonar images without high-quality reference images since obtaining such references during the sonar image denoising process was often difficult. Perera et al [24] proposed a transformer-based model for sonar image despeckling.…”

Section: Deep Learning Methods For Speckle Denoisingmentioning

confidence: 99%

“…Second, they inevitably lead to losses in the resolution, textures, and many other details of sonar images after denoising. Therefore, more recently, researchers have turned their attention to deep neural network-based methods [10][11][12][13][14][15][16][17] for sonar image despeckling. In particular, convolutional neural networks (CNNs) are the most widely adopted due to their strong abilities for spatial feature extraction, which is important to reduce speckle noise from sonar images since such noise is inherently spatial-aware.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SID-TGAN: A Transformer-Based Generative Adversarial Network for Sonar Image Despeckling

Zhou,

Tian,

Zhou

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

Sonar images are inherently affected by speckle noise, which degrades image quality and hinders image exploitation. Despeckling is an important pre-processing task that aims to remove such noise so as to improve the accuracy of analysis tasks on sonar images. In this paper, we propose a novel transformer-based generative adversarial network named SID-TGAN for sonar image despeckling. In the SID-TGAN framework, transformer and convolutional blocks are used to extract global and local features, which are further integrated into the generator and discriminator networks for feature fusion and enhancement. By leveraging adversarial training, SID-TGAN learns more comprehensive representations of sonar images and shows outstanding performance in speckle denoising. Meanwhile, SID-TGAN introduces a new adversarial loss function that combines image content, local texture style, and global similarity to reduce image distortion and information loss during training. Finally, we compare SID-TGAN with state-of-the-art despeckling methods on one image dataset with synthetic optical noise and four real sonar image datasets. The results show that it achieves significantly better despeckling performance than existing methods on all five datasets.

show abstract

Section: Deep Learning Methods For Speckle Denoisingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

SID-TGAN: A Transformer-Based Generative Adversarial Network for Sonar Image Despeckling

Zhou,

Tian,

Zhou

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…However, these methods are limited when dealing with complex seabed environments, as Side-scan sonar images often suffer from low resolution, insufficient features, high noise, and deformation. The advancements in deep learning technologies in the field of computer vision have significantly improved the performance of target detection and are thus widely used in the field of underwater intelligent detection [8][9][10]. Models based on Deep Convolutional Neural Networks (DCNN) are effective but require high-quality training data, which are often scarce and limited in representativeness for Side-scan sonar images [11][12][13][14].…”

Section: Introductionmentioning

confidence: 99%

Sample Augmentation Method for Side-Scan Sonar Underwater Target Images Based on CBL-sinGAN

Peng,

Jin,

Bian

et al. 2024

JMSE

View full text Add to dashboard Cite

The scarcity and difficulty in acquiring Side-scan sonar target images limit the application of deep learning algorithms in Side-scan sonar target detection. At present, there are few amplification methods for Side-scan sonar images, and the amplification image quality is not ideal, which is not suitable for the characteristics of Side-scan sonar images. Addressing the current shortage of sample augmentation methods for Side-scan sonar, this paper proposes a method for augmenting single underwater target images using the CBL-sinGAN network. Firstly, considering the low resolution and monochromatic nature of Side-scan sonar images while balancing training efficiency and image diversity, a sinGAN network is introduced and designed as an eight-layer pyramid structure. Secondly, the Convolutional Block Attention Module (CBAM) is integrated into the network generator to enhance target learning in images while reducing information diffusion. Finally, an L1 loss function is introduced in the network discriminator to ensure training stability and improve the realism of generated images. Experimental results show that the accuracy of shipwreck target detection increased by 4.9% after training with the Side-scan sonar sample dataset augmented by the proposed network. This method effectively retains the style of the images while achieving diversity augmentation of small-sample underwater target images, providing a new approach to improving the construction of underwater target detection models.

show abstract

“…This creates a domain gap between synthetic data and real sonar images, leading to artifacts or loss of potential information [36]. Self-supervised learning methods have shown promise as an effective denoising solution [37]. These methods employ a blind convolutional network structure to efficiently process noisy versions of each image in the dataset and reconstruct clean pixels from adjacent pixels [38], [39].…”

Section: Introductionmentioning

confidence: 99%

LPMsDE: Multi-Scale Denoising and Enhancement Method Based on Laplacian Pyramid Framework for Forward-Looking Sonar Image

Wang,

Li,

Teng

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Forward-looking sonar (FLS) images present various challenges in interpretation, recognition, and segmentation due to limitations like low resolution, speckle noise, and low contrast, making them more complex than optical images. Existing methods often focus solely on denoising or enhancement, neglecting the potential benefits of utilizing multi-scale features to create an integrated image processing approach. This paper introduces the Laplacian pyramid-based Multi-scale Denoising and Enhancement (LPMsDE) method tailored for FLS images. The proposed method begins by presenting a novel multiplicative speckle noise model, grounded in the Gaussian distribution, specifically designed for FLS images. Next, the Laplacian pyramid decomposition is utilized to estimate noise variance, with an improved adaptive local filter. Lastly, a combination of the Laplacian pyramid framework, the enhanced adaptive local filter, and Contrast-Limited Histogram Equalization (CLHE) is employed to denoise and enhance images at different resolution levels. Through comprehensive experiments conducted on both simulated and real sonar images, the effectiveness of the LPMsDE method is demonstrated. It surpasses other denoising and enhancement techniques, as evidenced by superior scores in Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), Equivalent Number of Looks (ENL), Natural Image Quality Evaluator (NIQE), and Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE).

show abstract

Deep Denoising Method for Side Scan Sonar Images without High-quality Reference Data

Cited by 5 publications

References 12 publications

SID-TGAN: A Transformer-Based Generative Adversarial Network for Sonar Image Despeckling

SID-TGAN: A Transformer-Based Generative Adversarial Network for Sonar Image Despeckling

Sample Augmentation Method for Side-Scan Sonar Underwater Target Images Based on CBL-sinGAN

LPMsDE: Multi-Scale Denoising and Enhancement Method Based on Laplacian Pyramid Framework for Forward-Looking Sonar Image

Contact Info

Product

Resources

About