DenoDet V2: Phase-Amplitude Cross Denoising for SAR Object Detection

2025-08-12

AI Summary

SAR target detection suffers from severe degradation due to coherent speckle noise. To address this, we propose a novel transform-domain feature disentanglement and modulation paradigm that fully exploits the complementarity between phase and magnitude spectra. Specifically, we design a band-level intermodulation mechanism to enable bidirectional enhancement between these two spectral components. Our method integrates deep attention architectures with a phase–magnitude cross-denoising strategy, effectively suppressing speckle while preserving structural target information. Evaluated on multi-source SAR datasets—including SARDet-100K—our approach achieves state-of-the-art performance: +0.8% mAP improvement, alongside ~50% reduction in both model parameters and computational complexity. The core contribution lies in the first introduction of band-level intermodulation into SAR transform-domain modeling, enabling an end-to-end detector with high robustness and low inference overhead.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Feature Construction/Reformulation
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

One of the primary challenges in Synthetic Aperture Radar (SAR) object detection lies in the pervasive influence of coherent noise. As a common practice, most existing methods, whether handcrafted approaches or deep learning-based methods, employ the analysis or enhancement of object spatial-domain characteristics to achieve implicit denoising. In this paper, we propose DenoDet V2, which explores a completely novel and different perspective to deconstruct and modulate the features in the transform domain via a carefully designed attention architecture. Compared to DenoDet V1, DenoDet V2 is a major advancement that exploits the complementary nature of amplitude and phase information through a band-wise mutual modulation mechanism, which enables a reciprocal enhancement between phase and amplitude spectra. Extensive experiments on various SAR datasets demonstrate the state-of-the-art performance of DenoDet V2. Notably, DenoDet V2 achieves a significant 0.8% improvement on SARDet-100K dataset compared to DenoDet V1, while reducing the model complexity by half. The code is available at https://github.com/GrokCV/GrokSAR.

Problem

• Addresses coherent noise in SAR object detection
• Modulates transform-domain features via attention architecture
• Enhances phase and amplitude spectra complementarity

Innovation

• Phase-amplitude cross denoising via attention
• Band-wise mutual modulation mechanism
• Transform domain feature deconstruction