7: Image Plots

Note: if your diagram is to be reproduced by a journal, it is unlikely that the reproduction will be able to distinguish between any two graylevels which differ by less than 0.2. Also, graylevels less than 0.2 may appear as pure black, while those of 0.8 or more may appear as pure white. These guidelines are as specified by American Geophysical Union (publishers of J. Geophysical Res.), as of 1998.

• Reading and Creating Image Data: Reading image, and creating from grid data
• Image PostScript Output: How the image is embedded in the PostScript
• Example Image: How to plot a satellite image

7.1: Reading and Creating Image Data

• Create images from gridded data using `convert grid to image'. For examples see see Grayscale Images), see Combination, and see Contouring.
• Read raw ascii image data files. Use `read grid'.
• Read PGM (portable graymap) ascii files. (That is, a file with magic characters `P1' or `P3' at the start.) Use the `read image pgm' command, for a file opened in ascii mode with `open filename'.
• Read raw binary data, with or without headers. Use `read image', after skipping any header bytes using the `skip' command, for a file opened in binary mode with `open filename binary'.
• Read a Sun ``rasterfile'' file (but only in uncompressed form). Use `read image rasterfile' for a file opened in binary mode with `open filename binary'.
• Read a PGM (portable graymap) binary file. (A file with magic characters `P2' or `P4' at the start.) Use the `read image pgm' for a file opened in binary mode with `open filename binary'.
• Aside: Images can be converted to grids (for contouring) using `convert image to grid' (see Convert).

Once the image is created, its grayscale/colorscale may be manipulated with the commands `set image grayscale' and `set image colorscale', which permit linear and histogram-equalized blendings over the grayscale or color range, or with `read image grayscale' and `read image colorscale', which permit reading in the grayscale or color values individually, one for each of the 256 pixel values.

It is important to understand the structure of image data. Gri works only with 8-bit image data. This means that a given pixel of the image may have only one of 256 possible values. The example below uses a satellite image of surface temperature. The suppliers of the data dictate that pixel value 0 corresponds to a temperature of 5C, and a pixel value of 255 corresponds to 30.5C, so the resolution is 0.1C per pixel value. This resolution will be apparent if the output of the example below is previewed on a grayscale/color monitor --- notice the quantization in the palette. This resolution issue is not very important with satellite images, since you have to use what you are given by the suppliers of the data. However, the issue is very important when you are converting grid data to images. When Gri converts grid data to image data, it neccessarily discards information, because the grid data have resolution to about 6 digits, whereas the image data have only 8-bit (2-3 digit) resolution. The `set image range' commands determines the range of this 8-bit resolution in terms of user units. All other things being equal, it would be preferable to use the smallest range consistent with the range of your data. If your grid data ranged from 0 to 1, say, you might `set image range 0 1'. This would give a resolution in the image of 1/255 in the user units. But, when Gri converts the grid into an image, it will clip all data outside the indicated range. In this case, any data greater than 1 in the grid would translate to exactly 1 in the image. Naturally there is a tradeoff between having a range large enough to encompass any data in the grid, and a range small enough to yield adequate resolution. In most cases, 8-bit resolution will be adequate, but it is good to be aware of the limitations. One should always `draw image palette', and check it on a color monitor for bandedness, which is a sign of resolution problems.

7.2: About The PostScript Output

%BEGIN_IMAGE 170.70 170.70 534.86 534.86 128 128 im

The first four numbers are the (x,y) locations of the lower-left and upper-right corners of the image, in units of points on the page (72 points = 1 inch). The fifth and sixth numbers are the width of the image and the height of the image. The keyword `im' is always present on this line. Gri inserts the following comment line at the end of the image data

%END_IMAGE

7.3: Example (Satellite image)

The file in this example covers 128 * 128 pixels over the Gulf of Maine. The numbers in `\filename' correspond to surface temperatures according to the equation

T = 5 + 0.1 * pixel_value

which explains the following lines in the command file:

\0val = "5" # 0 in image \255val = "30.5" # 255 in image

Depending on `\histo', the graymap will be linear or histogram-enhanced. The histogram method consists of dividing the cumulative histogram for the values in the image up into 256 levels, and assigning a graylevel to each. This has the effect of creating maximal contrast in all ranges of graylevel. It points up features really well, but it is a nonlinear mapping, so it is not good for telling you where gradients are strong or weak.

The command-file.