融合自适应采样与全局感知的图像深度估计算法  

Image Depth Estimation Algorithm Incorporating Adaptive Sampling and Context-Aware Module

作　　者：王国相 李昌隆 宋俊锋[1,2] 叶振 金恒[4] WANG Guoxiang;LI Changlong;SONG Junfeng;YE Zhen;JIN Heng(Lishui University,Lishui,Zhejiang 323000,China;Key Laboratory of Digital Design and Intelligent Manufacture in Culture and Creativity Product of Zhejiang Province,Lishui,Zhejiang 323000,China;School of Information and Communication Engineering,BeijingUniversity of Posts andTelecommunications,Beijing 100876,China;Zhejiang Dahua Technology Company Limited,Hangzhou 310051,China)

机构地区：[1]丽水学院,浙江丽水323000 [2]浙江省特色文创产品数字化设计与智能制造重点实验室,浙江丽水323000 [3]北京邮电大学信息与通信工程学院,北京100876 [4]浙江大华技术股份有限公司,杭州310051

出　　处：《计算机工程与应用》2025年第5期261-268,共8页Computer Engineering and Applications

基　　金：浙江省“尖兵”重点研发计划项目(2023C01004);装发共用基金(61405180409);浙江省自然科学基金探索项目(LTGN23F020001);丽水市公益性技术应用研究计划项目(2022GYX06)。

摘　　要：深度估计旨在通过少量稀疏深度样本点预测场景的稠密深度图,现有方法通常直接从稀疏深度样本生成最终的深度预测图,没有充分挖掘稀疏深度图包含的几何信息,导致深度估计算法的预测精度不够高。针对上述问题,提出一种融合自适应采样与全局感知的图像深度估计算法,由粗粒度到细粒度逐步预测深度图。通过引入预训练的深度补全网络预测粗粒度的稠密深度图,获取丰富的场景结构信息和语义信息。设计自适应深度采样方法,引导算法模型对远处的区域施加更多关注,缓解深度数据的长尾分布问题。同时通过新设计的全局感知模块,捕获并融合多尺度特征,从而获取更多的场景上下文信息。在NYU-Depth-v2数据集上的实验结果表明,算法在整体性能上超越了其他方法;消融实验的结果验证了提出的各个模块的有效性;Zero-shot实验的结果表明算法有较好的泛化性能,其中在ScanNet数据集上的阈值精度指标δ<1.25相比P3D方法提升了42个百分点,相比S2D方法则提升了3.8个百分点。Depth estimation aims to predict dense depth maps of the scene from a few sparse depth samples.Existing works directly generate the final depth prediction but not sufficiently exploit the geometric information in sparse depth maps,which results in the prediction accuracy of the depth estimation algorithm not being high enough.To solve this problem,an image depth estimation algorithm incorporating adaptive sampling and context-aware module is proposed to progressively predict depth maps from coarse-level to fine-level.Firstly,a pre-trained depth completion network is introduced to predict coarse-level dense depth maps and obtain rich scene structures and semantic information.Then,the adaptive sampling is designed to guide the model to pay more attention to distant regions which can alleviate the long-tail problem of depth data.Meanwhile,the newly designed context-aware module captures and fuses multi-scale features to obtain more context information of the scene.Experimental results on NYU-Depth-v2 dataset show that the heuristic depth estimation network surpasses compared with methods in several indicators.Results of ablation study demonstrate the effectiveness of the proposed modules.Zero-shot experiments verify the generalization ability of the proposed algorithm,and the accuracy indicatorδ<1.25 improves 42 percentage points over P3D and 3.8 percentage points over S2D,respectively.

关 键 词：深度估计 深度补全 稠密深度图 多尺度特征融合 自适应采样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]