Abstract:Intrinsic image decomposition is a fundamental problem for many computer vision applications. While recent deep learning based methods have achieved very promising results on the synthetic densely labeled datasets, the results on the real-world dataset are still far from human level performance. This is mostly because collecting dense supervision on a real-world dataset is impossible. Only a sparse set of pairwise judgement from human is often used. It's very difficult for models to learn in such settings. In this paper, we investigate the possibilities of only using reflectance images for supervision during training. In this way, the demand for labeled data is greatly reduced. In order to achieve this goal, we take a deep investigation into the reflectance images. We find that reflectance images are actually comprised of two components: the flat surfaces with low frequency information, and the boundaries with high frequency details. Then, we propose to disentangle the learning process of the two components of the reflectance images. We argue that through this procedure, the reflectance images can be better modeled, and in the meantime, the shading images, though not supervised, can also achieve decent result. Extensive experiments show that our proposed network outperforms current state-of-the-art results by a large margin on the most challenging real-world IIW dataset. We also surprisingly find that on the densely labeled datasets (MIT and MPI-Sintel), our network can also achieve state-of-the-art results on both reflectance and shading images, when we only apply supervision on the reflectance images during training.

Extracting intrinsic images from multi-spectral

Recovering Facial Intrinsic Images From A Single Input

Intrinsic Image Transformation via Scale Space Decomposition

Correlation-Based Intrinsic Image Extraction From A Single Image

Sparse Intrinsic Decomposition and Applications.

Intrinsic Decomposition from a Single Spectral Image.

Towards High-Quality Intrinsic Images in the Wild

Hyperspectral Image Reconstruction from RGB Input Through Highlighting Intrinsic Properties

Face synthesis from near-infrared to visual light spectrum using quotient image and kernel-based multifactor analysis

Intrinsic Image Decomposition: A Comprehensive Review

Reconstructing hyperspectral images from rgb inputs based on intrinsic image decomposition

Multispectral Image Intrinsic Decomposition via Subspace Constraint

Intrinsic Image Decomposition Using Color Invariant Edge

Intrinsic Image Decomposition by Pursuing Reflectance Image.

Intrinsic Image Decomposition Using Multi-Scale Measurements and Sparsity.

L0 co-intrinsic images decomposition

Intrinsic Face Image Decomposition With Human Face Priors

Intrinsic images using optimization

Intrinsic image extraction based on deconvolutional neural networks

Multispectral Image Intrinsic Decomposition via Low Rank Constraint

Face relighting based on multi-spectral quotient image and illumination tensorfaces