A Simple Model for Intrinsic Image Decomposition with Depth Cues
@article{Chen2013ASM, title={A Simple Model for Intrinsic Image Decomposition with Depth Cues}, author={Qifeng Chen and Vladlen Koltun}, journal={2013 IEEE International Conference on Computer Vision}, year={2013}, pages={241-248} }
We present a model for intrinsic decomposition of RGB-D images. Our approach analyzes a single RGB-D image and estimates albedo and shading fields that explain the input. To disambiguate the problem, our model estimates a number of components that jointly account for the reconstructed shading. By decomposing the shading field, we can build in assumptions about image formation that help distinguish reflectance variation from shading. These assumptions are expressed as simple nonlocal…
163 Citations
Intrinsic decomposition from a single RGB-D image with sparse and non-local priors
- Computer Science2017 IEEE International Conference on Multimedia and Expo (ICME)
- 2017
A new intrinsic image decomposition method that decomposes a single RGB-D image into reflectance and shading components is proposed that yields better results and enjoys lower complexity compared with two state-of-the-art methods.
Intrinsic Decompositions for Image Editing
- Computer ScienceComput. Graph. Forum
- 2017
A new synthetic ground‐truth dataset is introduced that is used to evaluate the validity of these priors and the performance of the methods, and the performances of the different methods in the context of image‐editing applications are evaluated.
Scene Intrinsics and Depth from a Single Image
- Computer Science2015 IEEE International Conference on Computer Vision Workshop (ICCVW)
- 2015
This work demonstrates the first pipeline for full intrinsic decomposition of scenes from a single color image input alone, and regresses depth by a fully convolutional network then jointly optimizes the intrinsic factorization to recover the input image.
Learning an Intrinsic Image Decomposer Using Synthesized RGB-D Dataset
- Computer ScienceIEEE Signal Processing Letters
- 2018
This letter uses a synthesized dataset to train a deep neural network for intrinsic image decomposition and further fine-tune it for real-world images, and results verify the effectiveness of the model on realistic images.
Intrinsic Image Decomposition using Deep Convolutional Network
- Computer Science
- 2016
This paper proposes a deep convolutional network for intrinsic image decomposition from a single image, approach from the perspective of filtering, without pooling for downsampling and unpooling for recovering the original resolution.
Joint Estimation of Depth, Reflectance and Illumination for Depth Refinement
- Computer Science2015 IEEE International Conference on Computer Vision Workshop (ICCVW)
- 2015
A simple optimization based approach with smoothness constraints on depth, reflectance and illumination from a single RGB-D image is proposed to generate high quality depth without additional processes such as pre-training of stochastic models or image segmentation.
Intrinsic Image Decomposition Using Structure-Texture Separation and Surface Normals
- Computer ScienceECCV
- 2014
A novel image model for handling textures in intrinsic image decomposition and a novel constraint based on surface normals obtained from an RGB-D image are proposed, which can produce superior results to existing approaches.
A Variational Model for Intrinsic Light Field Decomposition
- Computer ScienceACCV
- 2016
A novel variational model for intrinsic light field decomposition is presented, which is performed on four-dimensional ray space instead of a traditional 2D image, and the advantages of this approach with respect to intrinsic image decomposition and specular removal are demonstrated.
Direct Intrinsics: Learning Albedo-Shading Decomposition by Convolutional Regression
- Computer Science2015 IEEE International Conference on Computer Vision (ICCV)
- 2015
The strategy is to learn a convolutional neural network that directly predicts output albedo and shading channels from an input RGB image patch, which outperforms all prior work, including methods that rely on RGB+Depth input.
Intrinsic Scene Properties from a Single RGB-D Image
- Computer ScienceIEEE Transactions on Pattern Analysis and Machine Intelligence
- 2013
In this paper, we present a technique for recovering a model of shape, illumination, reflectance, and shading from a single image taken from an RGB-D sensor. To do this, we extend the SIRFS (“shape,…
References
SHOWING 1-10 OF 40 REFERENCES
Ground truth dataset and baseline evaluations for intrinsic image algorithms
- Computer Science, Environmental Science2009 IEEE 12th International Conference on Computer Vision
- 2009
This work presents a ground-truth dataset of intrinsic image decompositions for a variety of real-world objects, and uses this dataset to quantitatively compare several existing algorithms.
Color Constancy, Intrinsic Images, and Shape Estimation
- Mathematics, Computer ScienceECCV
- 2012
SIRFS is presented, the first unified model for recovering shape, chromatic illumination, and reflectance from a single image, which outperforms all previously published algorithms for intrinsic image decomposition and shape-from-shading on the MIT intrinsic images dataset and on its own "naturally" illuminated version of that dataset.
Estimating Intrinsic Component Images using Non-Linear Regression
- Computer Science, Mathematics2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)
- 2006
A method for estimating intrinsic component images from a single image, which is based on learning estimators that predict filtered versions of the desired image, and shows better performance on the shading/albedo separation problem than previous approaches.
Intrinsic Scene Properties from a Single RGB-D Image
- Computer ScienceIEEE Transactions on Pattern Analysis and Machine Intelligence
- 2013
In this paper, we present a technique for recovering a model of shape, illumination, reflectance, and shading from a single image taken from an RGB-D sensor. To do this, we extend the SIRFS (“shape,…
Intrinsic Images by Clustering
- Computer Science, MathematicsComput. Graph. Forum
- 2012
A novel algorithm that requires no user strokes and works on a single image is presented, based on simple assumptions about its reflectance and luminance, which builds a linear system describing the connections and relations between them.
Intrinsic images using optimization
- Computer Science, MathematicsCVPR 2011
- 2011
This work presents a novel intrinsic image recovery approach using optimization based on the assumption of color characteristics in a local window in natural images, which achieves a better recovery of intrinsic reflectance and illumination components than by previous approaches.
Shading-Based Shape Refinement of RGB-D Images
- Computer Science2013 IEEE Conference on Computer Vision and Pattern Recognition
- 2013
A shading-based shape refinement algorithm which uses a noisy, incomplete depth map from Kinect to help resolve ambiguities in shape-from-shading and refinement of surface normals using a noisy depth map leads to high-quality 3D surfaces.
User-assisted intrinsic images
- Mathematics, Computer ScienceSIGGRAPH 2009
- 2009
A new approach is described that enables users to guide an optimization with simple indications such as regions of constant reflectance or illumination, and derives a new propagation energy that enables a closed form solution using linear least-squares.
Estimating Intrinsic Images from Image Sequences with Biased Illumination
- MathematicsECCV
- 2004
This paper describes how biased illumination sampling leads to biased estimates of reflectance image derivatives, and proposes a solution that explicitly models spatial and temporal constraints over the image sequence to avoid the effects of illumination bias.
Recovering Intrinsic Images with a Global Sparsity Prior on Reflectance
- Computer Science, MathematicsNIPS
- 2011
This work introduces a new prior on reflectance, that results in a Random Field model with global, latent variables (basis colors) and pixel-accurate output reflectance values, and shows that without edge information high-quality results can be achieved, that are on par with methods exploiting this source of information.