VFX I

4.3 Lens Distortion and Overscan


A brief look at the concept of Lens Distortion and the Overscan workflow

Lens Distortion

Every camera lens creates lens distortion. The CG camera doesn’t create such artifact, the render will have straight undistorted lines. We need to apply lens distortion to the CG to make it accurately fit the footage


osc_grid_footage-org
Original Footage

Footage has convex distortion, like a cushion

To undistort, we would need to pull the corners out

osc_grid_footage-orgcg-flat_wrong
CG without Distortion

The CG has no lens distortion, all lines are straight

This can’t fit accurately on distorted original footage

osc_grid_footage-orgcg-dist_wrong
CG with Lens Distortion

Applying lens distortion, now CG fits on footage

But lens distortion pushes corners inside

CG gets smaller and we’re missing image content

 


osc_grid_cg-overscan

Overscan

The missing image content has to be rendered in a larger canvas size called Overscan

osc_grid_final

CG with Lens Distortion and Overscan

Lens distortion pushes corners in, but CG with overscan has enough image content



osc_workflow-1_v3_full

Overscan Workflow

Integrating overscan CG on footage would go like this

  1. Lens Distortion is applied to overscan CG
  2. Then merged with overscan footage
  3. Before writing out, overscan is removed

You could remove overscan from CG before merging on normal size footage, but keeping overscan thought the comp is useful, especially when operations affecting edge pixels, such as blurs, are involved