OutCast: Outdoor Single Image Relighting with Cast Shadows

1 University College London

2 Adobe


We propose a relighting method for outdoor images. Our method mainly focuses on predicting cast shadows in arbitrary novel lighting directions from a single image while also accounting for shading and global effects such the sun light color and clouds. Previous solutions for this problem rely on reconstructing occluder geometry, e.g., using multi-view stereo, which requires many images of the scene. Instead, in this work we make use of a noisy off-the-shelf single-image depth map estimation as a source of geometry. Whilst this can be a good guide for some lighting effects, the resulting depth map quality is insufficient for directly ray-tracing the shadows. Addressing this, we propose a learned image space ray-marching layer that converts the approximate depth map into a deep 3D representation that is fused into occlusion queries using a learned traversal. Our proposed method achieves, for the first time, state-of-the-art relighting results, with only a single image as input.


If no comparisons are present please refresh your browser (CTRL/CMD R).

Interactive Timelapse

Drag the slider beneath the relit image (right) to change sun position. All results are generated using our method.


Network Overview

The main network architecture of our approach. Input is outlined in orange, output in green. Pink lines indicate the cost functions we optimize for. Blocks with learnable parameters are double-wedges, and a dotted line indicates siamese training / weight sharing. The scissor symbol is where the gradients are detached from the network graph, to prevent further back propagation of the gradient.


    title={OutCast: Single Image Relighting with Cast Shadows},
    author={Griffiths, David and Ritschel, Tobias and Philip, Julien},
    journal={Computer Graphics Forum},
    organization={Wiley Online Library}