Libmv

libmv: Open Source Computer Vision Library for Structure from Motion and Multiview Geometry

libmv is an open source computer vision library focused on structure from motion and multiview geometry, estimating 3D coordinates from 2D image correspondences and camera motion and calibration parameters.

What is Libmv?

libmv is an open source computer vision library developed by the Blender project for reconstructing 3D scenes from images or video. It focuses on structure from motion (SfM) and multiview geometry techniques to estimate 3D structure and camera motion from 2D image correspondences.

Some key capabilities and features of libmv include:

• Estimating camera calibration parameters like focal length, principal point, distortion model, etc.
• Robust matching of feature points between images using techniques like SIFT and optical flow
• Estimating the 3D position of points seen from multiple camera viewpoints
• Reconstructing sparse and dense 3D point clouds from images
• Estimating camera motion including rotation and translation for each input image
• Bundle adjustment optimization to jointly refine structure and motion estimates
• Handling unordered image collections and variable numbers of features per image
• CUDA support for GPU-accelerated SfM computation

libmv is designed to be modular and reusable. It focuses specifically on the narrow set of algorithms required for multiview reconstruction tasks. This allows it to be integrated as a library into larger applications that require 3D reconstruction capabilities.

Some uses cases of libmv include photogrammetry, scanning objects by moving around them, background replacement in video, camera tracking, and 3D modeling from images. Since it is open source, libmv enables developers to easily build on top of state-of-the-art SfM techniques.

Libmv Features

Official Links