ACM SIGGRAPH Asia 2017 (Transactions on Graphics) |
|
High-Quality Hyperspectral Reconstruction Using a Spectral Prior |
|
|
Inchang Choi† |
Daniel S. Jeon† |
Giljoo Nam† |
Diego Gutierrez* |
Min H. Kim† |
|
|
|
|
|
|
|
|
|
|
†KAIST |
|
*Universidad de Zaragoza, I3A |
|
|
|
|
Our novel hyperspectral reconstruction algorithm works with input from any existing compressive imaging architecture, and yields high-quality results, both in terms of spectral accuracy and spatial resolution. As the comparisons show, our results improve significantly over previous state-of-art methods. For instance, both TwIST and SpaRSA provide suboptimal spatial reconstruction in general, while sparse coding yields a noisy reconstruction of the color chart, and fails to accurately reconstruct the green border in the coffee mug. The charts on the right show how our reconstruction provides an excellent fit to the ground-truth data. In addition, we provide in this work a new high-resolution hyperspectral image dataset. |
|
|
|
|
|
|
|
Abstract |
|
|
|
We present a novel hyperspectral image reconstruction algorithm, which overcomes the long-standing tradeoff between spectral accuracy and spatial resolution in existing compressive imaging approaches. Our method consists of two steps: First, we learn nonlinear spectral representations from real-world hyperspectral datasets; for this, we build a convolutional autoencoder, which allows reconstructing its own input through its encoder and decoder networks. Second, we introduce a novel optimization method, which jointly regularizes the fidelity of the learned nonlinear spectral representations and the sparsity of gradients in the spatial domain, by means of our new fidelity prior. Our technique can be applied to any existing compressive imaging architecture, and has been thoroughly tested both in simulation, and by building a prototype hyperspectral imaging system. It outperforms the state-of-the-art methods from each architecture, both in terms of spectral accuracy and spatial resolution, while its computational complexity is reduced by two orders of magnitude with respect to sparse coding techniques. Moreover, we present two additional applications of our method: hyperspectral interpolation and demosaicing. Last, we have created a new high-resolution hyperspectral dataset containing sharper images of more spectral variety than existing ones, available through our project website. |
|
|
|
|
|
@Article{DeepCASSI:SIGA:2017,
author = {Inchang Choi and Daniel S. Jeon and Giljoo Nam
and Diego Gutierrez and Min H. Kim},
title = {High-Quality Hyperspectral Reconstruction
Using a Spectral Prior},
journal = {ACM Transactions on Graphics (Proc. SIGGRAPH Asia 2017)},
year = {2017},
volume = {36},
number = {6},
pages = {218:1-13},
doi = "10.1145/3130800.3130810",
url = "http://dx.doi.org/10.1145/3130800.3130810",
}
|
|
|
|
|
|
Hosted by Visual Computing Laboratory, School of Computing, KAIST.
|
|
|