Intrinsic Image Transformation via Scale Space Decomposition
This addresses the problem of separating albedo and shading components in images for computer vision applications, representing an incremental improvement with a novel network architecture.
The paper tackles intrinsic image decomposition by proposing a network structure that learns image-to-image transformation in frequency bands via Laplacian pyramid expansion, achieving state-of-the-art performance on MPI-Sintel and MIT Intrinsic Images datasets with clear progression over existing methods.
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.