Temporal-spatial sparse coding for X-ray image analysis and interpretation

Digital X-ray cameras are experiencing an ``arms race'' towards ever higher resolutions and frame
rates. However, the cost of digital X-ray cameras continues to increase as their spatial and
temporal resolution grows higher. Here we illustrate that by exploiting prior knowledge about
the data, we can increase the effective frame rate of an X-ray video sequence, thereby increasing
the dynamic range of digital X-ray cameras whose frame rate would otherwise be too slow to
capture all of the relevant physical phenomena. Here, spatiotemporal upsampling is
accomplished by sparse coding the video at a high frame rate and learning spatial-temporal
dictionaries that encode multiple frames, allowing us to interpolate the missing frames that
would otherwise be lost when recording at a lower frame rate. Our approach motivates the
construction of less expensive X-ray sensors with a smaller sampling frequency that would by
required by the Nyquist-Shannon sampling theorem, while retaining the ability to capture
physical phenomena the required spatiotemporal resolution.