changing data values with imshift and rotate
bwp wrote on Sep 27, 2006
I am attempting to analyze data from the Hubble archives. At this stage, my goal is to align and median combine 4 images.
I used the STSDAS task wfits to make a standard FITS file of the image from the planetary camera (I intend to deal with the wide field images later.) However, when I align the images using the standard IRAF tasks imshift and rotate, I get a sort of blurring and streaking. Stars get elongated and the become brighter and "blocky," so that an object (or cosmic ray) that was basicly 2x2 pixels might become 4x2 pixels. This is not just a matter of contrast: the data values are in fact different from what they were in the original image.
Has anyone encountered this problem before? Is there a tool in the STSDAS package that I have overlooked?
-Brad Peterson
Iowa State University
I used the STSDAS task wfits to make a standard FITS file of the image from the planetary camera (I intend to deal with the wide field images later.) However, when I align the images using the standard IRAF tasks imshift and rotate, I get a sort of blurring and streaking. Stars get elongated and the become brighter and "blocky," so that an object (or cosmic ray) that was basicly 2x2 pixels might become 4x2 pixels. This is not just a matter of contrast: the data values are in fact different from what they were in the original image.
Has anyone encountered this problem before? Is there a tool in the STSDAS package that I have overlooked?
-Brad Peterson
Iowa State University
Francisco Valdes wrote on Sep 27, 2006
Hello Brad,
Resampling an image, which is what IMSHIFT and ROTATE do, inevitably smears the data because one must estimate unknown data from the finite pixel size. There is no solution though using different interpolators can give poorer or better appearance. This subject is a major one in astronomical image processing. If you data or simply offset for each other in x and y then you might get something closer to what you want using the "offsets" option of IMCOMBINE. Note this does only integer registration so on the one hand you don't get interpolation smoothing but on the other hand you can register more than to the nearest pixel. The alternative to functional interpolation, which is what IMSHIFT, ROTATE, and GEOTRAN do, is a method called "drizzle". This also is fundamentally an interpolation method and can't beat the nature of trying to estimate data at a sampling lost due to finite pixels but it might be something you could look at. The drizzle style resampling is part of the STSDAS package.
Yours,
Frank Valdes
Resampling an image, which is what IMSHIFT and ROTATE do, inevitably smears the data because one must estimate unknown data from the finite pixel size. There is no solution though using different interpolators can give poorer or better appearance. This subject is a major one in astronomical image processing. If you data or simply offset for each other in x and y then you might get something closer to what you want using the "offsets" option of IMCOMBINE. Note this does only integer registration so on the one hand you don't get interpolation smoothing but on the other hand you can register more than to the nearest pixel. The alternative to functional interpolation, which is what IMSHIFT, ROTATE, and GEOTRAN do, is a method called "drizzle". This also is fundamentally an interpolation method and can't beat the nature of trying to estimate data at a sampling lost due to finite pixels but it might be something you could look at. The drizzle style resampling is part of the STSDAS package.
Yours,
Frank Valdes
Last post on Sep 27, 2006