|
Colour in Context
Research group Computer Vision Center |
 |
Object Recoloring based on Intrinsic Image Estimation:
Watch the updated video describing our work here: (please right click on the link and choose "save as...")
Here's the high resolution poster presented in ICCV 2011:
And the full resolution paper published in ICCV 2011 for further reference:
Shida Beigpour, Joost van de Weijer, Object Recoloring based on Intrinsic Image Estimation, Proc. ICCV, Barcelona, Spain, 2011.
Recoloring or color-alteration is a term referring to the modification and adjustment of the image color appearance. Color modification methods are applied to photo montage, color correction, visual effects in movies, and also in industrial and commercial applications (Computer-Aided Design) as a technique to visualize the final color appearance of the 3D object products before actual production in order to improve and facilitate their design.
The problem of object recoloring is highly under-constrained, and graphics methods often fail to properly generate a realistic perception of the recolored object (e.g. the case of non-white illuminant) as they suffer from a lack of knowledge about the physics rules applied on the light interaction with the object surface. Application of these methods to real-world scenes lit by colored illuminants, multiple illuminants, or interreflections, results in unrealistic recoloring of objects.
The main idea is that the incident light not only affects the brightness of the pixels we perceive in the image of an object, but also affects their chromaticy. Therefore we believe by making a distinction between different regions of the object surface (e.g. shading, and highlights), and decomposing the object chromaticy into the natural color of the object surface as well as the color of the illuminant, we are capable of performing the correct color modification.
Using this idea, we developed a physic-based framework for modeling the change in the object color as well as the chromaticy of its illuminant light. The experimental results on natural images taken with non-calibrated cameras indicate that a realistic recoloring of an object with complex specularity and shading have been achieved by our framework. We use the single-colored object constraint to fit a more comprehensive physical model to the object pixels color distribution. This permits us to perform realistic recoloring of objects lit by colored lights, and multiple illuminants. Moreover, the model allows for more realistic scene relighting. Recoloring results on images captured by uncalibrated cameras demonstrate that the proposed framework obtains realistic recoloring for complex natural images. Furthermore we use the model to transfer color between objects and show that the results are more realistic than existing color transfer methods. Below is an example of the performance of the method as well as a video describing the paper.
Above is an example of multi-illuminant object recoloring: (a) Original image; (b) Object mask as the initial mask for the primary illuminant; (c) The Mask after 1st iteration; (d) The Mask at 3rd (final) iteration; (e) estimated body (diffuse) reflectance cause by the primary illuminant; (f) estimated body (diffuse) reflectance cause by the secondary illuminant (the interreflection area has been correctly detected); (g) estimated specular reflectance; (h) An example recoloring (the green interreflection is preserved).
[1] Shida Beigpour, Joost van de Weijer, Object Recoloring based on Intrinsic Image Estimation, Proc. ICCV, Barcelona, Spain, 2011. - Suplementary Video (please right click on the link and choose "save as...")- Poster
|
||
|
Contact
webmaster | Last updated: Monday 11
May 2009
|
