Orthorectification Tool

Orthorectifies individual images (or all images) from an existing ODM reconstruction. It reads the camera model, DEM, and shot positions produced by ODM and reprojects each pixel onto the ground plane.

Prerequisites

Tool	Minimum version
CMake	3.21
Git	any (for vcpkg submodule)
Windows — MSVC (Visual Studio 2022)	14.x
Linux — GCC or Clang + Ninja	GCC 10 / Clang 12

All C++ dependencies (GDAL, Eigen3, nlohmann-json, cxxopts, plog) are managed automatically by vcpkg in manifest mode. No manual library installation is needed.

Building

Clone the repository with its submodules (vcpkg is included as a shallow submodule):

git clone --recurse-submodules https://github.com/yourorg/orthorectify.git
cd orthorectify

Using the helper scripts (recommended)

Windows (Release / Debug):

build_release.bat
build_debug.bat

Linux (Release / Debug):

./build_release.sh
./build_debug.sh

The scripts bootstrap vcpkg on first run, configure via cmake --preset, and build. The binary is placed in build/<preset>/Release/Orthorectify (or Debug/).

Manual CMake workflow

# Windows
cmake --preset windows-release
cmake --build --preset windows-release

# Linux
cmake --preset linux-release
cmake --build --preset linux-release

Available presets: windows-release, windows-debug, linux-release, linux-debug.

Docker

docker build -t digipa/orthorectify .

The Docker image uses system GDAL/Eigen3 from Debian Bookworm packages; vcpkg is used only for the header-only dependencies (cxxopts, plog).

Usage

After running a reconstruction with ODM, point the tool at the dataset directory:

Orthorectify /dataset-path

With Docker:

docker run -it -v /dataset-path:/dataset digipa/orthorectify /dataset

Options

Usage:
  Orthorectify [OPTION...] <dataset-path>

  -e, --dem arg               Absolute path to DEM (default: odm_dem/dsm.tif)
      --no-alpha              Omit the alpha channel from output
  -i, --interpolation arg     Pixel sampling method: nearest | bilinear (default: bilinear)
  -o, --outdir arg            Output directory (default: orthorectified)
  -l, --image-list arg        File listing image filenames to process
                              (default: img_list.txt)
      --images arg            Comma-separated filenames to process
                              (alternative to --image-list)
  -s, --skip-visibility-test  Skip visibility test — faster but may leave
                              relief-displacement artifacts
  -t, --threads arg           Thread count (-1 = all available) (default: -1)
  -v, --verbose               Enable debug logging
  -h, --help                  Print this help

Project Structure

orthorectify/
├── src/                  # C++ source & headers
│   ├── main.cpp          # Entry point
│   ├── parameters.hpp    # CLI argument parsing (cxxopts)
│   ├── dataset.hpp       # ODM dataset loading (shots, camera models)
│   ├── rawimage.hpp/cpp  # Image I/O via GDAL
│   ├── processing.hpp    # Core orthorectification pipeline
│   ├── transform.hpp     # Coordinate transform helpers (Eigen)
│   ├── types.hpp         # Shared type aliases
│   └── utils.hpp/cpp     # Miscellaneous utilities
├── vcpkg/                # vcpkg submodule (shallow clone)
├── vcpkg.json            # vcpkg manifest — dependency list
├── CMakeLists.txt        # Root build definition
├── CMakePresets.json     # Build presets (windows/linux, debug/release)
├── version.h.in          # Version header template
└── Dockerfile            # Multi-stage container build

CI

GitHub Actions builds both linux-release and windows-release on every push/PR to main. See .github/workflows/build.yml.

---

Roadmap

• Merging of multiple orthorectified images (blending, seam leveling)
• Faster visibility test
• Direct orthorectification method
• GPU support