This program is part of Netpbm.
ppmshadow adds a simulated shadow to an image, giving the appearance that the contents of the image float above the page, casting a diffuse shadow on the background. Shadows can either be black, as cast by opaque objects, or translucent, where the shadow takes on the color of the object which casts it. You can specify the crispness of the shadow and its displacement from the image with command line options.
ppmshadow sees your image as a foreground on a background. The background color is whatever color the top left pixel of your image is. The background is all the pixels that are that color and the foreground is everything else. The shadow that ppmshadow generates is a shadow of the foreground, cast on the background.
The shadow is the same size as the foreground, plus some fringes as determined by the -b option. It is truncated to fit in your image. The output image is the same dimensions as the input image.
You can use pamcomp to place a foreground image over a background before running ppmshadow on it. You can use ppmmake to make the background image (just an image of a solid color).
The output has the same dimensions and maxval as the input.
The blurring to make the fringes of the shadow will not have a desirable effect if the color depth (maxval) of the image is too low -- you need a high maxval to get all the shades needed to create a smooth gradient. So if your input has low maxval (including most notably if the input is PBM, which means its maxval is 1), run it through pamdepth to raise its maxval. 255 is usually a good choice.
Input is a PPM file named by the ppmfile command line argument; if you don't specify ppmfile, the input is Standard Input.
The output is a PPM file, written to Standard Output.
The default is 11 pixels.
Note that this option controls only the fringing effect of moving the light source closer to the object. It does not make the shadow grow or shrink as would happpen in the real world if you moved a point light source closer to and further from an object.
The source image must contain sufficient space on the edges in the direction in which the shadow is cast to contain the shadow -- if it doesn't some of the internal steps may fail. You can usually expand the border of a too-tightly-cropped image with pnmmargin before processing it with ppmshadow.
Black pixels and pixels with the same color as the image background don't cast a shadow. If this causes unintentional "holes" in the shadow, fill the offending areas with a color which differs from black or the background by RGB values of 1, which will be imperceptible to the viewer. Since the comparison is exact, the modified areas will now cast shadows.
The background color of the source image (which is preserved in the output) is deemed to be the color of the pixel at the top left of the input image. If that pixel isn't part of the background, simply add a one-pixel border at the top of the image, generate the shadow image, then delete the border from it.
If something goes wrong along the way, the error messages from the various Netpbm programs ppmshadow calls will, in general, provide little or no clue as to where ppmshadow went astray. In this case, Specify the -k option and examine the intermediate results in the temporary files (which this option causes to be preserved). If you manually run the commands that ppmshadow runs on these files, you can figure out where the problem is. In problem cases where you want to manually tweak the image generation process along the way, you can keep the intermediate files with the -k option, modify them appropriately with an image editor, then recombine them with the steps used by the code in ppmshadow.
Shadows are by default black, as cast by opaque material in the image occluding white light. Use the -t option to simulate translucent material, where the shadow takes on the color of the object that casts it. If the contrast between the image and background is insufficient, the -t option may yield unattractive results which resemble simple blurring of the original image.
Because Netpbm used to have a maximum maxval of 255, which meant that the largest convolution kernel pnmconvol could use was 11 by 11, ppmshadow includes a horrid, CPU-time-burning kludge which, if a blur of greater than 11 is requested, performs an initial convolution with an 11 x 11 kernel, then calls pnmsmooth (which is itself a program that calls pnmconvol with a 3 x 3 kernel) as many times as the requested blur exceeds 11. It's ugly, but it gets the job done on those rare occasions where you need a blur greater than 11.
If you wish to generate an image at high resolution, then scale it to publication size with pamscale in order to eliminate jagged edges by resampling, it's best to generate the shadow in the original high resolution image, prior to scaling it down in size. If you scale first and then add the shadow, you'll get an unsightly jagged stripe between the edge of material and its shadow, due to resampled pixels intermediate between the image and background obscuring the shadow.
ppmshadow creates a number of temporary files as it executes. It creates a new directory for them in the directory named by the TMPDIR environment variable, defaulting to /tmp if it is not set.
In normal operation, ppmshadow finds a unique name for the temporary directory and deletes each temporary file as soon as it is done with it and leaves no debris around after it completes. To preserve the intermediate files for debugging, use the -k command line option. In that case, the directory name is ppmshadowpid, where pid is the process ID of the ppmshadow process, and the program fails if ppmshadow cannot create that directory because the name is already in use.
The temporary files are: