NWE Help: Graphics: XV: Image Algorithms

NWE Home :: Help :: Graphics :: XV

xv now has the ability to run a number of standard image-processing algorithms on the current image, or any rectangular portion of the current image. If there is a current selection rectangle (See "Selection Rectangle" ), the algorithm will only affect the selected area. Otherwise, the algorithms operate on the entire image.

Algorithms are chosen via the Algorithms menu, and are executed immediately. Algorithms are cumulative, in that if you run an algorithm on an image, and then run a second algorithm, the second algorithm operates on the modified image produced by the first algorithm. And so on.

See "Adding Algorithms to xv" for information on how you can add additional algorithms to this menu.

Also, it should be noted that the algorithms operate on 24-bit images. If you are currently operating on an 8-bit image, and you run an algorithm, the image will be converted up to 24-bits, the algorithm will be run, and the result will be converted back to 8-bits using the currently selected 24->8 algorithm. As such, if you're going to be doing a lot of algorithm-ing, you may find it faster to temporarily switch to 24-bit mode . Likewise, if you intend to run multiple algorithms on the same image (say, a Blur followed by an Emboss ), you should definitely switch to 24-bit mode , to maintain full 24-bit precision on the intermediate images (i.e., to prevent noise from being added to the image by any intermediate 24->8 conversions).

The (normally dimmed-out) Undo All selection undoes any and all algorithms that have been run on the current image. It restores the image to the state it was in when the first algorithm was executed, and it also puts xv back into the 8/24-bit mode it was in.
Keyboard Equivalent: <Meta> u

Runs a convolution over each plane (red, green, blue) of the image, using a n*n convolution mask consisting of all 1's. It has the effect of, for each pixel in the image, replacing it with the average value of all the pixels in the n*n rectangle centered around the pixel in question. When you select this command, a dialog box is opened up to request the value for ' n'. Values must be greater than or equal to '3', and they must also be odd (e.g. 3,7,5,9,11). Larger values blur the image to a greater extent.

Warning! This algorithm, due to its n^2 nature, can get very slow as n is increased. Values above '15' aren't particularly recommended.

Note: This command is a reasonable way to turn dithered black-and-white images back into some semblance of the greyscale that they probably originally started as.
Keyboard Equivalent: <Meta> b

Runs an edge-sharpening algorithm on the image, which attempts to maximize contrast between adjacent pixels. Prompts for a 'strength' value of 0% through 99%.
Keyboard Equivalent: <Meta> s

Runs a convolution using a pair of convolutions, one which detects horizontal edges, and one which detects vertical edges. The convolution is done separately for each plane (red, green, blue) of the image. It is only done for pixels that have the 3x3 mask fully contained within the image, to avoid pesky edge conditions. One note: since it runs an edge detection separately for each plane of the image, the results are colorful. You'll get red edges when there are edges in the red plane, yellow edges when there are edges in the red and green planes, and so on. If you'd like a traditional grey edge detection (based on the overall intensity of each pixel), you should use the Saturation dial in the xv color editor (See "The Saturation Control" ) to completely desaturate all the colors in the image (turning it grey) first. Then, the results will also be grey.
Keyboard Equivalent: <Meta> e

Runs an algorithm that produces nifty 'embossed' images by using a variation of the edge detection algorithm. It produces greyscale (8-bit) images where most of the image is shown as a medium gray. 'Leading' edges (edges found on the top and left sides) are shown as a lighter gray, and 'trailing' edges (bottom and right edges) are shown as a darker gray. The image wind up looking like pseudo-3-d, sort of like the Motif toolkit. You can then use the White dial in the xv color editor (See "The White Remapping Control" ) to colorize the image as you see fit
Keyboard Equivalent: <Meta> m

Does an 'oil transfer', as described in the book "Beyond Photography", by Holzman. It is a sort of localized smearing. The basic algorithm is to take a rectangle (7x7 in this case) centered around the current pixel, compute a histogram of these (49) pixels, and replace the current pixel with the 'most popular' pixel as determined by the histogram.
Keyboard Equivalent: <Meta> o

Blends a rectangular selection out of existence by replacing the selected area with a radial gradient fill. Each pixel on the edge of the selection retains its original color, the pixel in the center is replaced by the average of all the edge pixels, and all remaining pixels are replaced by the weighted average of the center pixel and the edge pixel (along the line that runs through the center pixel and the pixel being computed).

This is much easier to see than it is to explain.
Keyboard Equivalent: <Meta> B

Does arbitrary (smooth) rotation of the entire image (or the selected region). You will be prompted to enter a rotation amount, in degrees. Positive values rotate the image counter-clockwise, negative values rotate the image clockwise. The Copy Rotate function makes a copy of the selection, rotates it, and pastes it back on top of the original.
Keyboard Equivalent: <Meta> t

Also does arbitrary smooth rotation, as with the Copy Rotate command, but this version clears the original image (using the 'current color', see "Displaying Pixel Values" ) before pasting the rotated version.
Keyboard Equivalent: <Meta> T

Runs a 'pixelization' algorithm, suitable for obscuring sensitive and/or naughty parts of an image. Pops open a dialog box which prompts for either a single value ' m', or a pair of values ' mx n'. Divides the image (or the selected region) up into m-by- m squares (if a single value was entered) or m-by- n rectangles (if a pair of values were entered). Each area is filled with the average color of all pixels contained within the area.
Keyboard Equivalent: <Meta> p

Runs a 'random spread' algorithm on the selected area of the image (or the entire image). Pops up a dialog box which prompts for either a single value ' m', or a pair of values ' mx n'. In the case of the single value, each pixel in the image is swapped with some other random pixel located within a distance of m pixels from the first pixel. If two values are entered, each pixel is swapped with another pixel located within m, n of the original pixel. This can produce an interesting 'de-rez' effect. Whether it's useful or not is for you to decide.
Keyboard Equivalent: <Meta> S

Runs a despeckling algorithm, also known as a median filter among image-processing circles. This algorithm is good at removing minor noise from an image, such as you'll normally find on scanned-in faxes and the like. It may also prove useful for de-dithering images, turning black-and-white dithered images into greyscale images, and dithered color images into undithered color images. Note that the Blur algorithm can also be used in the same way, and may do a better job of un-dithering. However, the DeSpeckle algorithm generally leaves edges alone, unlike the Blur algorithm, which blurs everything uniformly.

This algorithm runs through the selected area of the image, and for each pixel in the image, looks at the m-by- m square centered around the pixel. It replaces the pixel with the median value of the pixels found in the square. As with the Blur algorithm, you will be prompted for a value of ' m'. The value must not be less than '3', and it must be odd. Larger values have a greater effect, though '3' should be sufficient for most purposes.
Keyboard Equivalent: <Meta> k