In 2005 July, the NASA Deep Impact mission successfully completed a flyby of comet 9P/Tempel 1. A flaw in the prelaunch calibration system resulted in an inability to accurately focus the telescope's High Resolution Instrument, resulting in a significant loss of resolution. However, because of the nature of the blurring process, much of the high-frequency information is retained and can be recovered using deconvolution. We discuss the data-processing requirements and noise modeling needed prior to deconvolution and compare a few popular deconvolution algorithms. All of the algorithms give very similar results. More importantly, we find that none of the algorithms correctly handle the errors resulting from the data compression used on board the spacecraft to compress the 16 bit CCD data to 8 bits prior to transmission of the image to the ground. We present a modification to the Richardson-Lucy deconvolution algorithm that we have successfully used to account for these compression errors.
[1]
John W. Fowler,et al.
A maximum correlation method for image construction of IRAS survey data
,
1990
.
[2]
A. S. Fruchter,et al.
Drizzle: A Method for the Linear Reconstruction of Undersampled Images
,
1998
.
[3]
R. White,et al.
Image recovery from data acquired with a charge-coupled-device camera.
,
1993,
Journal of the Optical Society of America. A, Optics and image science.
[4]
William H. Richardson,et al.
Bayesian-Based Iterative Method of Image Restoration
,
1972
.
[5]
H. Melosh,et al.
Deep Impact: Excavating Comet Tempel 1
,
2005,
Science.
[6]
L. Lucy.
An iterative technique for the rectification of observed distributions
,
1974
.
[7]
Fionn Murtagh,et al.
Deconvolution in Astronomy: A Review
,
2002
.