url: http://code.google.com/p/ceres-solver/

Solving nonlinear least squares problems comes up in a broad range of areas across science and engineering - from fitting curves in statistics, to constructing 3D models from photographs in computer vision. Ceres Solver is a portable C++ library that allows for modeling and solving large complicated nonlinear least squares problems.

Features
* A friendly API: build your objective function one term at a time.
* Automatic differentiation.
* Robust loss functions.
* Local parameterizations.
* Threaded Jacobian evaluators and linear solvers.
* Levenberg-Marquardt and Dogleg (Powell & Subspace) solvers.
* Dense QR and Cholesky factorization (using Eigen) for small problems.
* Sparse Cholesky factorization (using SuiteSparse) for large sparse problems.
* Specialized solvers for bundle adjustment problems in computer vision.
* Iterative linear solvers for general sparse and bundle adjustment problems.
* Runs on Linux, Windows, Mac OS X and Android. An iOS port is underway.