C1 False color and pseudocolor images

Observations with Charge-Coupled Devices (CCDs) do not really generate images, they produce a 3-D set of numbers: x, y, and intensity (z). We have just seen how these numbers are displayed as an image and how the use of a look-up table allows the translation from pixel value (z) to displayed grey scale. Sometimes, we wish to present our results in color either for dramatic effect or for science productivity, or at times, a little of both. Additionally, color images often truly help us to see the details better.

A black-and-white image may make the digital data more understandable, but the number of different grey tones that the human eye can separate is very limited. Our eyes can only see or interpret 20-30 grey steps yet an 8-bit display shows us 200 or more steps on a contrast scale. On the other hand, our eyes can separate 20 000 or more different color tints or shades making pseudo and false color images much better at showing the real variation in the original data.

Pseudo-color images are single displayed images (usually 8-bits) in which each grey level is assigned a color. A simply approach may be to assign a LUT that is a ramp of red values from a light red to a dark red, with an ever increasing redness as the LUT goes form 0 to 255. A more complex pseudo-color representation may be to take three colors, red, green, and blue, and assign LUT values as follows: 0-85 is a red ramp, 86-171 is a green ramp, and 172-255 is a blue ramp. Another scheme might use these same three LUT ranges but only one color. In this mode, the real intensity values are represented by color that wraps around at each numeric boundary. The pixel values of 9, 95, and 181 will be displayed as the same color. In the end, we can use the 8-bit display to show a color image in which the grey levels of 0-255 are assigned a single color per datum value. There may be logic to the color assignment such as showing an image of the planet Neptune in shades of blue to express coldness or showing the crab nebula in red to infer hotness. This type of color scheme tries to help us interpret the scientific meaning of the image through casual visual inspection. Another choice of pseudo-color may be to try to separate image features close in value through use of color. Let's say we have a planetary nebula image and we wish to separate the central star from the gaseous nebula for contrast in a presentation. However, the central star is faint and its pixel values are 100-200 ADU while the nebula itself is 400-600 ADU. Using the above scheme of color ramps would assign shades of the same color to both components, reducing their contrast in the displayed image. However, use of a different color for each block of say 100 ADU levels will allow the user to stretch the 8-bit displayed image such that the star and the nebula are separated by a color boundary, thereby making them distinct in the final display.

A "false color" image does not mean that the data are wrong or that the picture is deceiving you. The term simply means that the image is not a real color photograph. False color images are composed of three separate 8-bit images displayed simultaneously (i.e., added together) allowing 24 bits or roughly 16 million possible color values to be represented. Each 8-bit color plane has a unique LUT of color values for its 256 possible numeric levels. An example of a display scheme may be to assign each 8-bit plane to a ramp of red, green, and blue. Colors are produced when the three LUTs overlap (the three images are displayed simultaneously) for any given pixel. If the red, green, blue (RGB) values are set to (255,255,255) the color appears white, (0,0,0) is black, (255,0,0) is pure red, (255,0,255) is pure purple, etc. The primary colors (RGB), when mixed in groups of two, produce the secondary colors of cyan, magenta, and yellow. The production of color and shades of color relies on the combination of the three LUTs as laid out in a chromaticity diagram or color wheel (Rector et al., 2005). This method of producing various colors works in much the same way as a color television. If you look closely at a color TV screen, you'll see that it is made up of many sets of three circular or hexagonal color regions (red, green, blue) and by varying the intensity in each, the color is changed. Your eye can not resolve (from far away) the three segments and thus blends them together to form the desired color. Color prints follow this scheme as well, usually printing images with three passes, one for each of three printed colors.

Let us look at an example. We wish to present a color image of a star-forming region in Orion using three digital images. One of the images was obtained with a CCD camera through an [O III] filter centered on the 4959 A and 5007 A emission lines. The second image was obtained with the same camera but using an Ha filter centered at 6563 A. The third image, produced by the same camera, used a broad-band U band filter centered at 3500 A. One choice of color presentation is to make each image a pseudo-color image assigning each a color LUT in red, green, and blue. The three images can then be displayed beside each other or printed out and viewed next to each other.

Another method is to produce a false color image by combining the three 8-bit images into a single 24-bit image. We might assign each image a color scheme in a similar way as just described (or not) and then combine the three into a single image. Tweaking of the color LUTs can then be performed to produce the final scientifically useful, as well as eye-pleasing, image. Other wavelengths, such as X-ray or infrared light images, can also be combined with or without an optical component to produce false color images. In these images, the color choice is often a matter of personal taste, and is used in a manner generally associated with the intensity or brightness of the radiation from different regions of the image.

For example, in a greyscale or black and white Chandra X-ray image of a supernova remnant, the darker shades of grey might represent the most intense X-ray emission, the lighter shades of gray could represent the areas of less intense emission, and the white areas could be used to show areas of little to no emission. In a color version, lighter colors such as yellow or orange could represent areas of high X-ray intensity, orange to red areas of lower intensity, and black representing little or no emission. This false color image representation provides the eye with a seemingly hot region (bright colors) for the highest X-ray emission and ends with darker, low emission regions in black. The color assignments in each image LUT in this case follow variations in intensity of the X-ray counts, which in reality are associated with variations in the density, or concentration, of hot gas. A typical method used to produce false color images from Chandra data is to use three images obtained in the energy bands of 0.3 to 1.55 keV, 1.55 to 3.34 keV, and 3.34 to 10keV, respectively. The intensity value within each image is mapped to a color LUT value of 0-255 and the three images are then combined to produce a false color X-ray image.

If we combine three optical images in a manner akin to how our eyes would see it, e.g., with similar intensity ratios for red, green, and blue light, the image is called a "true color" image. As usual, the best way to learn all about color imagery is to experiment yourself. Programs such as "The Gimp," XV, Photoshop, and others can be used for this purpose. An excellent account of how to produce color astronomical images including all the factors and intricacies involved, as well as examples, is provided in Rector et al. (2005).

http://www.photo.net/photo/edscott/ms000050.html

http://www.allthesky.com/articles/imagecolor.html

http://chandra.harvard.edu/photo/false_color.html