LDFE: Laplacian Decoupled Feature Enhancement Block for Dual-Stream CNN-based RGB-IR Object Detection

2026-07-08

AI Summary

This work addresses the insufficient feature fusion in existing RGB-infrared multimodal object detection methods under extreme conditions by proposing a Laplacian Decoupled Feature Enhancement (LDFE) module. The LDFE module integrates Laplacian pyramids into different stages of a dual-stream CNN backbone to decouple features into global and local components, which are then separately denoised, enhanced, and reconstructed. Furthermore, it innovatively combines state space models with local convolutional correlations to enable dynamic cross-modal noise suppression and bidirectional interaction. To the best of our knowledge, this is the first study to introduce Laplacian-based feature decoupling into RGB-IR object detection. Extensive experiments on six benchmark datasets—including M3FD and DroneVehicle—demonstrate consistent improvements, with mean average precision (mAP) gains ranging from 2.0% to 6.2% over current state-of-the-art methods.

Abstract

The complementary information between RGB and IR images can significantly enhance object detection performance under extreme conditions. Existing methods prefer dual-stream CNN backbones built upon YOLO for feature extraction and focus on the design of feature fusion. In this paper, we introduce the Laplacian Decoupled Feature Enhancement block (LDFE) to fuse features from different stages of the dual-stream CNN backbone. By design, LDFE simultaneously considers the characteristics of modalities and structures for feature fusion by employing global-local decomposition, denoising, fusion, and reconstruction, sequentially. The LDFE first separates features into global and local components based on Laplacian Pyramid, and then performs denoising and fusion based on Global State Space Enhancement module (GS2E) and Local Convolutional Correlation Enhancement module (LC2E) separately. Specifically, the GS2E conducts a two-branch architecture for the main and auxiliary modalities. It dynamically suppresses noise in the main modality through cross-modal attention derived from the auxiliary modality, while employing a State Space Model to capture long-range dependencies within the global feature representations of the main modality. To obtain bidirectional interaction, the two modalities systematically alternate their main/auxiliary roles. Moreover, the LC2E suppresses noise in local features and leverages spatial and channel dimension along with triple convolution to extract fine-grained details for fusion. These innovative designs achieve a significant performance improvement, with mAP surpassing the SOTA methods 6.2%, 3.7%, 4.7%, 2.3%, 4.1% and 2.0% on M3FD, DroneVehicle, LLVIP, FLIR-Aligned, KAIST and VEDAI datasets,respectively.

Problem

• RGB-IR object detection
• feature fusion
• dual-stream CNN
• modality complementarity
• extreme conditions

Innovation

• Laplacian Pyramid
• State Space Model
• Cross-modal Attention
• Dual-stream CNN
• Feature Decoupling