Mamba-based multi-branch cost aggregation for stereo matching

IF 6.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Applied Soft Computing Pub Date : 2025-09-27 DOI:10.1016/j.asoc.2025.113973

Xingyuan Lu , Yanbing Xue , Leida Li , Shiyin Li , Zan Gao

{"title":"Mamba-based multi-branch cost aggregation for stereo matching","authors":"Xingyuan Lu , Yanbing Xue , Leida Li , Shiyin Li , Zan Gao","doi":"10.1016/j.asoc.2025.113973","DOIUrl":null,"url":null,"abstract":"<div><div>This study presents Mamba-Based Multi-Branch Cost Aggregation for Stereo Matching (MMBStereo), an innovative real-time stereo matching framework with high performance. The core innovation lies in the Mamba-based multi-branch cost aggregation network, which uses a unique three-branch aggregation strategy. The Mamba Aggregation Branch integrates the State Space Model from the Mamba structure, replacing conventional convolution and Transformer methods, significantly enhancing network performance and efficiency. The Spatial Aggregation Branch addresses the loss of spatial texture information, improving the scene’s contextual representation. Meanwhile, the Edge Aggregation Branch enhances edge responses, improving object boundary detection accuracy. Through a carefully designed multi-branch fusion strategy, the framework improves disparity prediction accuracy while maintaining real-time inference. Our method achieves competitive accuracy with non-real-time stereo matching frameworks, surpassing existing lightweight solutions in the widely recognized KITTI benchmark tests.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"185 ","pages":"Article 113973"},"PeriodicalIF":6.6000,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Applied Soft Computing","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1568494625012864","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}

引用次数: 0

Abstract

This study presents Mamba-Based Multi-Branch Cost Aggregation for Stereo Matching (MMBStereo), an innovative real-time stereo matching framework with high performance. The core innovation lies in the Mamba-based multi-branch cost aggregation network, which uses a unique three-branch aggregation strategy. The Mamba Aggregation Branch integrates the State Space Model from the Mamba structure, replacing conventional convolution and Transformer methods, significantly enhancing network performance and efficiency. The Spatial Aggregation Branch addresses the loss of spatial texture information, improving the scene’s contextual representation. Meanwhile, the Edge Aggregation Branch enhances edge responses, improving object boundary detection accuracy. Through a carefully designed multi-branch fusion strategy, the framework improves disparity prediction accuracy while maintaining real-time inference. Our method achieves competitive accuracy with non-real-time stereo matching frameworks, surpassing existing lightweight solutions in the widely recognized KITTI benchmark tests.

查看原文本刊更多论文

基于mamba的多分支成本聚合立体匹配

本文提出了一种基于mamba的多分支成本聚合立体匹配（MMBStereo）框架，该框架是一种新型的、高性能的实时立体匹配框架。核心创新点在于基于曼巴的多分支成本聚合网络，采用独特的三分支聚合策略。Mamba聚合分支集成了来自Mamba结构的状态空间模型，取代了传统的卷积和Transformer方法，显著提高了网络的性能和效率。空间聚合分支解决了空间纹理信息的丢失，改善了场景的上下文表示。同时，边缘聚合分支增强了边缘响应，提高了目标边界检测精度。通过精心设计的多分支融合策略，该框架在保持实时推理的同时提高了视差预测精度。我们的方法达到了与非实时立体匹配框架竞争的精度，在广泛认可的KITTI基准测试中超越了现有的轻量级解决方案。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Applied Soft Computing 工程技术-计算机：跨学科应用

CiteScore

15.80

自引率

6.90%

发文量

874

审稿时长

10.9 months

期刊介绍： Applied Soft Computing is an international journal promoting an integrated view of soft computing to solve real life problems.The focus is to publish the highest quality research in application and convergence of the areas of Fuzzy Logic, Neural Networks, Evolutionary Computing, Rough Sets and other similar techniques to address real world complexities. Applied Soft Computing is a rolling publication: articles are published as soon as the editor-in-chief has accepted them. Therefore, the web site will continuously be updated with new articles and the publication time will be short.