Intrinsic Image Transformation via Scale Space Decomposition

TL;DR

This paper presents a novel multi-channel neural network architecture that decomposes images into albedo and shading by leveraging scale space and Laplacian pyramids, achieving state-of-the-art results.

Contribution

Introduces a scalable, multi-channel CNN that processes intrinsic image components across frequency bands using Laplacian pyramids, improving decomposition performance.

Findings

Outperforms existing methods on MPI-Sintel and MIT datasets.

Effectively captures multi-scale features for intrinsic image decomposition.

Demonstrates both quantitative and qualitative improvements.

Abstract

We introduce a new network structure for decomposing an image into its intrinsic albedo and shading. We treat this as an image-to-image transformation problem and explore the scale space of the input and output. By expanding the output images (albedo and shading) into their Laplacian pyramid components, we develop a multi-channel network structure that learns the image-to-image transformation function in successive frequency bands in parallel, within each channel is a fully convolutional neural network with skip connections. This network structure is general and extensible, and has demonstrated excellent performance on the intrinsic image decomposition problem. We evaluate the network on two benchmark datasets: the MPI-Sintel dataset and the MIT Intrinsic Images dataset. Both quantitative and qualitative results show our model delivers a clear progression over state-of-the-art.