What Is Photogrammetry?
Photogrammetry is the science of measuring the real world through photographs. By exploiting stereoscopy - matching identical points between images taken from different angles - modern algorithms can automatically reconstruct the three-dimensional geometry of a subject from a set of multi-view photographs. The result is a point cloud, later converted into a polygonal mesh with PBR textures extracted directly from the source photographs.
Photogrammetry is one of the most transformative technologies in the visual effects industry today: it captures the geometry and appearance of real objects, environments and characters with a precision and realism impossible to reach through manual modeling alone.
Its production advantages are considerable. First, near-instant photoreal realism: since textures come directly from photographs, surface appearance is authentic by definition, without the long, costly work of a texture artist. The time saved over manual modeling is also significant for complex assets - a ruined building, a cracked rock, an old vehicle with wear marks.
Applications by Industry
VFX and film - digitizing real environments to which CG extensions are added, creating hero props with a high level of detail fidelity, producing digital doubles of actors, or capturing structures meant to be virtually destroyed. Facial scanning for actor performance capture is also central to high-budget productions.
Games and real-time - in AAA games, libraries of photoreal scanned assets populate vast environments with a visual coherence and authenticity manual modeling can't match at that scale. LODs (Levels of Detail) generated from scanned meshes maintain performance while preserving foreground quality.
Classic Photogrammetry
Classic photogrammetry relies on photographing a subject from many varied angles, with 60-80% overlap between images. Processing software analyzes corresponding features between images and reconstructs each point's position in space via stereoscopic triangulation - the more photos and the greater the overlap, the more precise and complete the reconstruction.
Equipment - a professional-quality session typically uses a high-resolution DSLR or mirrorless camera, preferably with a fixed 24-50mm lens to limit distortion. Lighting must be diffuse and even, without hard shadows that would distort the geometric reconstruction. For small to medium objects, a motorized turntable automates subject rotation and ensures even angular coverage.
LiDAR (Light Detection and Ranging)
LiDAR uses laser pulses to measure distances with extreme precision, generating dense, accurate point clouds particularly suited to large scales and architectural environments. It comes in three main configurations:
- Terrestrial LiDAR - fixed, tripod-mounted scanners offering millimeter precision, particularly suited to digitizing architecture and complex interior environments. Multiple scan positions are usually needed to cover an entire space, with the resulting point clouds assembled (registration) in post-processing.
- Mobile LiDAR - LiDAR-equipped smartphones (notably iPhone Pro models since 2020) have radically democratized access to this technology. Real-time handheld scanning with immediate visual feedback lets a non-specialist operator capture spaces in minutes. Precision remains below professional scanners, but the practicality and accessibility make it valuable for previsualization and secondary assets.
- Aerial LiDAR (UAV) - LiDAR-equipped drones digitize vast territories quickly - topography, forests, complex terrain - with a precision impractical by other means. Common in cartography, landscape archaeology, and producing photoreal terrain for open-world environments.
Structured-Light and 4D Scanning
Structured-light scanners (Artec, Creaform, FARO) project coded light patterns onto the surface being scanned and analyze the pattern's deformation via stereo cameras. The resulting sub-millimeter resolution makes them the reference solution for objects requiring extreme detail - mechanical parts, faces, heritage artifacts - though usage is limited to studio conditions due to sensitivity to ambient light interference.
4D scanning adds the temporal dimension to 3D scanning: capturing the geometry and motion of an animated subject (an actor, fabric, liquid) simultaneously via a network of synchronized cameras - the technology behind productions like Unity's The Heretic and volumetric sequences in major VFX films.
Point Cloud Formats
The choice of file format for point clouds determines compatibility with processing tools and metadata retention.
- .LAS / .LAZ - the LiDAR industry standard. .LAS stores XYZ coordinates plus rich metadata (return intensity, return count, semantic point classification); .LAZ is its lossless compressed variant, drastically shrinking file sizes for very large datasets.
- .PLY / .PTS - .PLY (Polygon File Format) is universal, compatible with nearly all 3D software, storing XYZ coordinates alongside RGB color, normals and custom attributes, in both ASCII and binary encoding. .PTS is its equivalent in the terrestrial-scanner ecosystem.
- .E57 - standardized by ASTM, designed for long-term archiving of 3D scan data, natively supporting the assembly of multiple scans into a single file plus embedded panoramic images tied to scanner positions.
Mesh Formats and PBR Textures
| Format | Primary use | Notes |
|---|---|---|
| .OBJ | Universal standard | Paired with a .MTL file for materials |
| .FBX | Autodesk pipeline | Supports animation and rigs |
| .ABC | Alembic | Deformation cache for simulations |
| .STL | 3D printing | Geometry only, no textures |
A complete photogrammetry asset ships with a set of PBR (Physically Based Rendering) textures: Albedo (base diffuse color, extracted directly from photographs), Normal (a normal map encoding micro-geometry detail), Roughness (surface roughness controlling glossiness), and AO (Ambient Occlusion - contact shadows and cavities).
Processing Software
- RealityCapture (Epic Games) - the VFX and game industry standard for photogrammetry; its GPU-accelerated engine handles massive datasets.
- Agisoft Metashape - the reference in scientific and cartographic work; its precision for orthomosaics and topographic surveys makes it the tool of choice for archaeologists and cartographers.
- CloudCompare - a free, open-source tool specialized in point cloud processing: alignment, cleanup, comparison and analysis of clouds from different sources, with no commercial license needed.
- Meshroom (AliceVision) - a fully open-source photogrammetry solution; less performant than RealityCapture on large datasets, but an excellent option for learning and budget-limited projects.
The Standard Processing Workflow
- Import - loading raw photos or scan data, aligning cameras or scanners.
- Point Cloud - reconstructing the dense point cloud via multi-view correlation.
- Mesh - polygonizing the point cloud, generating surface topology.
- Textures - automatic UV mapping and baking PBR textures from the source photographs.
- Export - generating final OBJ or FBX files with associated textures.
This is the raw output most pipelines then hand off to a procedural tool like Houdini for cleanup, reconstruction and optimization - covered in the next article.