Pyramid-Based Shadow Removal

Yael Shor
The Hebrew University
The Hebrew University


Abstract:

In this paper we propose a novel method for detecting and removing shadows from a single image thereby obtaining a high-quality shadow-free image.With minimal user assistance, we first identify shadowed and lit areas on the same surface in the scene using an illumination-invariant distance measure. These areas are used to estimate the parameters of a novel affine shadow formation model. A new pyramid-based restoration process is then applied to produce a shadow-free image, while avoiding loss of texture contrast and introduction of noise. Unlike previous approaches, we account for varying shadow intensity inside the shadowed region by processing it from the interior towards the boundaries. Finally, to ensure a seamless transition between the original and the recovered regions we apply image inpainting along a thin border. We demonstrate that our approach produces results that are in most cases superior in quality to those of previous shadow removal methods. We also show that it is possible to easily composite the extracted shadow onto a new background or modify its size and direction in the original image.

 

    Additional Results:

    Comparison:  Two of our results compared  to state-of-the art  methods. (a) Input image (b) Our result (c) Results from Finlayson et al "Intrinsic images by entropy minimization"  (d) Results from Wu et al  "Natural shadow matting". The results in (c) exhibit some smoothing artifacts on the boundaries and the removed shadow area appears slightly darker than its surrounding areas. The results in (d) show some added noise (top) and some loss of contrast and color (bottom). Our results are free from boundary artifacts, color and/or contrast mismatches between the recovered shadow areas and the surrounding illuminated areas, do not introduce noise into the recovered areas, and handle well shadows of non-uniform intensity.

    Warping and Compositing the shadow mask back to the shadow-free image may be used to fake changes in the direction of illumination:

     

    The shadow mask can be also simply composited onto a novel background image: