Making HST Pictures: Examples and Illustrations
Zolt Levay (STScI)

Summaries based on poster papers presented at AAS meetings:
  1. January 2002 (487kB)
  2. January 2003 (655kB)
  3. January 2004 (341kB)

These illustrations represent some of the steps and techniques used to produce “pretty pictures” from HST images. Each thumbnail image links to a web-sized JPEG and the PDF icon links to an Acrobat PDF document.

Remove Cosmic Rays


TIFFs (ftp)

Bright, sharp spots and streaks from cosmic rays appear on every HST exposure. Multiple exposures at the same pointing allow the randomly-appearing noise to be removed relatively easily, as long as nothing else changed in the meantime (easier for non-variable deep space targets such as nebulae and galaxies rather than closer, fast-moving solar-system targets).

“Good” pixels from one exposure replace “bad” (cosmic ray) pixels from another exposure in a combined image. Otherwise, the exposures add together exactly as if it were one longer exposure.
     
Getting the most out of the data
TIFFs (ftp)

Automatically setting the smallest image value to be black and the highest value to white can force important areas of the image to be rendered invisible. By selecting a minimum and maximum data value to render as black and white, repsectively, (clipping) the significant regions of the image can be made visible.

Clipping the bright end to a lower pixel value shows some detail, but fainter details still may not visible. Clipping to still lower levels brings out fainter details but “saturates” (forces larger areas to be white) the bright regions.

Applying a non-linear (log, square root, etc.) transformation can compress the dynamic range so that more detail becomes visible. Fainter details can become visible without saturating the brightest regions.

Histograms are plots of the numbers of pixels at each brightness value, showing the relative distribution of intensities. They can be useful to guide the selection of clipping and intensity scaling. A flatter histogram reflects more detail visible in all intensity levels. Intense peaks indicate a concentration of values at a particular brightness.

     
Combine images in different filters
TIFFs (ftp)

Separate images are exposed through different color filters. The resulting black and white or “grayscale” images are assigned a color that can be viewed or reproduced (that may or may not be the color of the filter used for the exposure). The separate images are combined in color image “channels” that together produce a full range of hues.

Using filters that match the eye’s response can result in a natural or “true” color image. Otherwise, the colors are enhanced or shifted from what we would be able to see. In most cases, though, the colors in the picture represent actual colors in the observed object. That is, what looks red is redder in the object, but perhaps not as red as it appears.

In this case, four filters were used, with three assigned the additive primaries: red, green and blue and the fourth assigned violet. In principle, any hue may be assigned to any filter layer, but the maximum color range and contrast usually is available by assigning the primaries. Other hues are combination of the primaries so will generally desaturate the colors unless the structure that appears in that filter is different from the structure appearing in the other filters (as different hues in other layers).

     
An alternate way to make color images
TIFFs (ftp)

Eight bit (one byte) images contain 256 different values that may be represented by 256 shades of gray or 256 arbitrary colors from a color map (or palette) of all possible colors.

In this case, the colors do not necessarily represent actual color of the object or any physical feature other than brightness.
     
Why is part of the picture missing?
TIFFs (ftp)

HST’s WFPC2 instrument contains four CCD detectors (“chips”), each producing an image 800x800 pixels. Three “wide-field” (WF) chips have a field of view of 2.5 arcminutes. The fourth “planetary camera” (PC) chip sees an adjacent, narrower field of 35 arcseconds.

Because of the instrument’s optical path, the resulting images are rotated with respect to each other. The separate images must be rotated and scaled to the same pixel scale before being combined into a mosaic.

The fields overlap somewhat but “seams” remain in the resulting mosaic images.

Hubble’s new Advanced Camera for Surveys (ACS) does not pose this particular challenge. The camera does consist of two separate CCD detectors (each 2048x4096 pixels in size), separated by a narrow gap, but they are the same pixel size and scale. Most observations consist of overlapping, offset exposures which can be combined to fill in the gap.

     
Combine images for a wider field
TIFFs (ftp)

The relatively narrow field of view of the WFPC2 camera means that many targets overfill a single image. Several carefully aimed exposures (“pointings”) may be used to cover the field but must be combined afterward.

Overlapping areas can be cut around (masked) or combined. The resulting images retain the resolution of the individual WFPC2 fields.

All of the above illustrations together

These and some other Illustrations showing specific Adobe Photoshop tools.

Additional Photoshop techniques.
Combine images processed differently (TIFFs ftp: M51 or IC 4406)
Composite images of different resolution (TIFFs ftp)

Blue-Red Bar

January 15, 2004