A Supervised Learning Approach to Dynamic Sampling (SLADS)

SLADS is a greedy sparse dynamic image sampling algorithm. In SLADS, the objective at each step is to find the pixel location that maximizes the expected reduction in distortion (ERD) given previous measurements.

• The most detailed SLADS publication can be downloaded here.
• The SLADS Python code can be downloaded here.

The Python package includes a ReadMe.txt file as well as some simulated Electron Back Scatter Diffraction images generated using the Dream3D software for experimentation.

SLADS was developed by G. M. Dilshan P. Godaliyadda , Dong Hye Ye and Charles A. Bouman from the Integrated Imaging Laboratory at Purdue University, Gregery T. Buzzard from the Department of Mathematics at Purdue University, and Michael D. Uchic and Michael A. Groeber from the Air Force Research Laboratory (AFRL).

Here, we show results comparing SLADS with random sampling (RS) and Low-discrepancy sampling (LS) when ~6% of the image is sampled. In this experiment we use Electron Back Scatter Diffraction images generated using the Dream3D software. The details of the experiment are published in [2], which is also where the results were first published.

Figure 1: Images illustrating the sampling locations, reconstructions and distortion images after 6% of the image is sampled. (a) Original image, (b) random sample (RS) locations, (c) low discrepancy (LS) sample locations, (d) SLADS sample locations, (e) reconstruction using RS samples, (f) reconstruction using LS samples, (g) reconstruction using SLADS samples, (h) distortion using RS samples, (i) distortion using LS samples, (j) distortion using SLADS samples.

• Air Force Office of Scientific Research (MURI - Managing the Mosaic of Microstructure, grant # FA9550-12-1-0458)
• Air Force Research Laboratory Materials and Manufacturing directorate (Contract # FA8650-10-D-5201-0038)

1. Godaliyadda, G. M., Ye, D. H., Uchic, M. D., Groeber, M. A., Buzzard, G. T., and Bouman, C. A. A Framework for Dynamic Image Sampling Based on Supervised Learning. IEEE Transactions on Computational Imaging. doi: 10.1109/TCI.2017.2777482.
2. Godaliyadda, G. M., Ye, D. H., Uchic, M. D., Groeber, M. A., Buzzard, G. T., and Bouman, C. A. (2016). A supervised learning approach for dynamic sampling. Electronic Imaging, 2016(19), 1-8.
3. Scarborough, N. M., Godaliyadda, G. D. P., Ye, D. H., Kissick, D. J., Zhang, S., Newman, J. A., ... and Buzzard, G. T. (2017). Dynamic X-ray diffraction sampling for protein crystal positioning. Journal of synchrotron radiation, 24(1), 188-195.
4. Zhang, Y., Godaliyadda, G. D., Ferrier, N., Gulsoy, E. B., Bouman, C. A., and Phatak, C. (2017). Reduced Electron Exposure for Energy-Dispersive Spectroscopy using Dynamic Sampling. Ultramicroscopy.