Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

An Iterative Approach to Shadow Subtraction in 3D

IP.com Disclosure Number: IPCOM000193259D
Original Publication Date: 2010-Feb-16
Included in the Prior Art Database: 2010-Feb-16
Document File: 1 page(s) / 18K

Publishing Venue

IBM

Abstract

Described is a method for removing distortion in the final three-dimensional (3D) structure when the distortion is caused by dark shadows that could not be properly removed from the image using background subtraction. This approach to shadow reduction operates on the 3D structure after it has been produced, rather than working on the original two-dimensional (2D) images.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 93% of the total text.

Page 1 of 1

An Iterative Approach to Shadow Subtraction in 3D

Developers performing 3D reconstruction use 2D images from multiple perspectives to achieve the desired result. The process requires segmentation of the target from the images, most often using some form of background subtraction. If an object casts a shadow and the shadow cannot be removed from the background, then the 3D reconstruction will be inaccurate. The best known solution is to have a robust method of background subtraction to segment the target from the image and remove shadows. The problem is in the case of very dark shadows, where these approaches are lacking.

The solution to removing very dark shadows during the 3D reconstruction process includes the following steps:

1. The developer acquires the reconstructed 3D structure, and knows the locations of the light sources that are illuminating the object.

2. The shadows that would be cast by the reconstructed 3D structure are computed based on the knowledge of where light sources are located, resulting in a new 3D structure.

3. The new 3D structure is compared to the reconstructed 3D structure. If they match, then the true 3D structure of the target is verified, and no further work is required. If the two 3D structures do not match, then step 4 takes place.

4. Any parts of the new 3D structure that do not exist in the original reconstructed 3D structure are now identified as computed-shadows caused by shadows that were not segmented from the original 2D i...