融合注意力机制和离散小波变换的轻量化图像去雾算法

Lightweight Image Dehazing Algorithm with Attention Mechanism and Discrete Wavelet Transform Fusion

天气条件如雾霾对真实环境中采集的图像产生了对比度下降和纹理模糊等问题。现有深度学习方法在图像去雾方面取得了进展,但在不均匀雾霾情况下效果有限。此外,现有方法在性能和计算效率之间存在平衡问题。文章研究了一种轻量高效的图像去雾算法,结合了多尺度特征提取、注意力机制和离散小波变换,该算法在提升图像质量和实时处理方面表现出色。研究内容包括选择轻量的U-Net结构作为主干网络、引入特征提取块(MFE块)以增强网络辨别和提取能力,以及使用离散小波变换和注意力机制以应对复杂雾霾分布环境,实验结果显示,该算法显著提升了模型性能并解决了纹理细节损失的问题。

Weather conditions such as haze pose challenges to image quality in real-world environments,resulting in decreased contrast and blurred textures.While deep learning methods have made progress in image dehazing,their effectiveness is limited in the presence of non-uniform haze.Furthermore,striking a balance between performance and computational efficiency remains a challenge.This paper explores a lightweight and efficient image dehazing algorithm that combines multi-scale feature extraction,attention mechanisms,and discrete wavelet transform.The proposed algorithm demonstrates excellent performance in enhancing image quality and real-time processing.The research includes the adoption of a lightweight U-Net architecture as the backbone network,the introduction of Feature Extraction Blocks(MFE blocks)to enhance discriminative power and feature extraction,and the incorporation of discrete wavelet transform and attention mechanisms to address complex haze distributions,Experimental results show significant improvements in model performance and address the issue of texture detail loss.