Hubble Space Telescope: Image Deblurring with a Parallel Comptuer

This visualization is demonstrating that data collected from cameras on satellites have flaws or problems and that those issues can be fixed using a mathematical technique called maximum entropy using a parallel computer system.

An accretion disk is formed around the hot star as it captures the mass being lost by the giant. The accretion disk beams matter away from the system in the form of a symmetrical jet.

This is a raw image data from the Hubble Space Telescope faint object camera. We see that hot nebula matter that surrounds the entire system and the extended jet feature.

The raw data suffers from two problems. First, the flaw in the Hubble primary mirror; causes point sources of light to be spread out across many picture elements or pixels.

Secondly, the faint object camera has recorded the light in a highly non-linear fashion; known as Detector Saturation, the saturation manifests itself in the dark central pixels.

This is a colorized three-dimensional plot.

The saturated region is a valley of missing data. Restoring the data with a mathematical technique called maximum entropy.

Using the surrounding unsaturated but highly blurred data. The method is able to reconstruct the emission peaks that occupy the saturated region.

This region contains the main emission from the gas surrounding the hot star, the accretion disk and the cool star.

This is the most intense emission peak (shown on the left). To the right are smaller emission peaks coming from the more extended jet.

These techniques were performed on a parallel computer system, comprised of 8192 processors.

Massively parallel computations like this are able to correct imagery from spacecraft.