All flags are of the form key:value. For instance, if you would like to set the value of n to 5, add the argument n:5. Execute it from the command-line as follows:

java -jar -Xmx1000m Targeter.jar

Followed by your list of arguments. All arguments except input and target are optional.

An example command-line:

java -jar -Xmx1000m Targeter.jar n:15 m:5 depth:4 a:0.5 sample:2000 input:sourceimage.jpg,sourceimage_normalmap.jpg target:sketch.jpg

1. n: Integer size of the square window to use for targeting.
2. m: Integer size of the square of pixels to drop in place on each search
3. a or accuracy: Lower values make the distance function care less about the target and more about the coherency with the pixels it has already placed. Set to 0 to synthesize an image from scratch. (Currently you must still set the target.)
4. d or depth: Integer term to control the neighborhood of each pixel. When we examine the n by n block of pixels around the target pixel, we also have the option of viewing the corresponding n by n pixels in a lower resolution version of the image. This allows large regions of the image to diffusely effect the pixel choice without a lot of overhead. This parameter determines how many levels of resolution will be used. The region that will finally effect the choice of pixel has size n × n × 2^d.
5. s or sample: This parameter controls how many neighborhoods are searched for each matching. Whenever a new best matching is found, this many more neighborhoods will be searched. Half of the searching will be done near the current best point, and half will be random. This should be well over 1000 for good results, particularly if the source image is large and diverse.
6. i or input: The image to draw pixels from. This can have multiple values separated by commas to load the channels from multiple images into one fat image. (Useful for appending normal maps and specularity to an image.)
7. t or target: The image to try to match with pixels. This can also have multiple values.
8. tm or targetmask: This allows you to control a on a per-channel per-pixel basis. The set of files specified here will be used to set the importance of each pixel-channel in the target image. Note that even if you want to weight all colors the same, this image should still be in RGB format so that the channels match up properly. A greyscale image will only control the weights given to the red channel.
9. o or outputfile: You are probably better off not setting this and letting the software make a good name for you. If you need to set this, make sure you include enough filenames for the number of channels you plan on outputting. (one file per 3 channels currently.)
10. display: Setting this to false will turn off the display. You can also just close the window.

The amount of time it takes to run will be proportional to n²×d×target_size×f(s)/m² where f(s) is some nearly linear function of s.