Removing Shadows from Images

Abstract

We attempt to recover a 2D chromaticity intrinsic variation of a single RGB image which is independent of lighting, without requiring any prior knowledge about the camera. The proposed algorithm aims to learn an invariant direction for projecting from 2D color space to the greyscale intrinsic image. We notice that along this direction, the entropy in the derived invariant image is minimized. The experiments conducted on various inputs indicate that this method achieves intrinsic 2D chromaticity images which are free of shadows. In addition, we examined the idea to utilize projection pursuit instead of entropy minimization to find the desired direction.

Shadow Detection and Removal from Solo Natural Image Based on Retinex Theory

“…Since the illumination images represent the distribution of the lighting and shading, separating it from reflectance is useful to cancel out the effect of shadows. Land's theory could only remove the penumbrae in image since they were based on the following assumptions [12]:…”

A Bayesian approach for shadow extraction from a single image

“…The problem of shadow removal becomes particularly challenging if only a single and uncalibrated image of a complex scene is given. While previous works using a single image (e.g., [7,6]) have produced very good results on automatic shadow detection and removal, several assumptions were made about the light source and the camera. If complex textures are present in the image, they may be blurred after shadow removal.…”

Estimating Intrinsic Images from Image Sequences with Biased Illumination

“…Although this method successfully separates reflectance and shading for a given illumination direction used in training, it is difficult to create such a prior to classify edges under arbitrary lighting. Another edge-based method was proposed by Finlayson et al [11] that suppresses color derivatives along the illumination temperature direction to derive a shadow-free image of the scene. In addition to shadow edges, this approach may remove texture edges that also have a color derivative in the illumination temperature direction.…”