Offset tricks -- Get pixel (X,Y)

Here's a nice little trick for ya.

It is sometimes a bit difficult to figure out how to offset something in the way you want.

What if you could just map the image with x and y coordinates?

With a fast Offset trick, you can.

The idea is: Black is 0, white is 100 in both x and y coordinates. Feed two images that contain the coordinate information into two inputs and, voila, you have distorted the original image exactly to those coordinates.

Take an easy example at first: You plug in a horizontal profile gradient black to white to x coordinates and a vertical profile gradient black to white to y coordinates, you get a copy of the original image.

Then you flip one of the gradients, you get a mirror image along one of the axis.

Or you switch the horizontal and vertical gradients and you get a rotated image.

Or you increase the Repeat amount of either and you get multiple copies.

Then you can get more fancy and plug in a curve to the Gradient Profile input and distort the image.

Or you plug in a radial gradient into one input and a circular into the other one, and you get polar coordinates transformation.

And it goes from there.

The whole idea is that it is easier to visualize the final outcome if you always have black = 0, white = 100 for the x and y coordinates.

And you can do distortions that you might have thought were impossible or very complex.

Sample snippet with sine curves below.

Get X, Y.ffxml

Here are the GetX and GetY maps for the above image:

Hi Corvus,

Yes, you've got the idea. It makes using Offset much more painless when you can use coordinates from the original image and know that black is always x=0 (or y=0) and white is always x=100 (or y=100), no matter where in the image that value appears.

While you don't need to understand how it works to make it work, it is always nice to understand it. So:

To get the idea of how the two offsets work, you could ask yourself "which X-offset will give me the x=0 point from the original image wherever my image is full black?" If you start with an image that is black everywhere, the answer is "a gradient that is mid-gray at left edge and white in the right edge." (You can try this)

Similarly, the X-offset that gives you the x=100 point from the original image is a gradient that is black at the left edge and mid-gray at the right edge.

If you follow so far, you can see that, simply enough, everything else is a blend between these two. Thus the offset images ("Get X" and "Get Y") are used as the opacity control to blend between these two gradients.

Same with the Y-offset, just vertically.

It is really easy to pre-visualize the final result of your offsets this way. Trying to do figure it out without this trick is a real pain in the butt.

I'll do some more simple visual examples to show how this works a bit later.

TD

Fun:

Here's a less intense one, using the same filter:

more craziness:

Even more fun

Here's the source filter

Get X, Y CC05.ffxml

So you're still planning to write that definitive wiki article on offsets, ThreeDee?