My second needlepoint project is almost done; I’m stitching in the blue-gray between the orange checkerboard squares. Apparently, I like to photograph and post this stuff when it reaches 70% completion. Here it is:

I found the tree design while searching for “Pixel Art Tutorial.” I came across this great page full of swords, potions, and white-haired warlords scowling into the wind (all traditional eight-bit subjects). Here, pixel-artist Mithlomion shows you how to add greenery into your three-quarter-view adventure game:

Marianne from local needlepoint joint Fireside Stitchery showed me how to stretch a canvas onto a lightweight frame (wooden bars, rabbet joints, lots and lots of brass thumbtacks), and I bought a bunch of short hanks of tapestry wool (three-ply yarn; you separate the strands and stitch with just one at a time.

The challenge was to get the design on to the canvas. Since the design has so many different scattered pixels all over the place — and I didn’t want to freehand it yet — I tried making color separations.

I opened the tutorial image in Photoshop cropped to just the tree, and zoomed way in to the pixel level. Then did a “select color range” on each of the seven colors, floating each color as a separate layer. I then added register marks around the border, and a number to each separation to keep the colors straight. Here’s what the color separations look like:

Next, I monkeyed with the pixel resolution and device resolution of the file to match the canvas’ 18 threads-per-inch resolution. I tried sending a 53-pixel-wide file to the printer at a device resolution of 18 pixels per inch, but the printer’s own scaling algorithms softened the hard edges of the pixels, and everything looked mushy.

> Image Size dialog” border=”0″ align=”right” style=”padding-top: 10px; padding-left: 10px; padding-bottom: 10px;” />
So, after a few false starts, I rezzed up the image to 10 times its original size, using Photoshop’s Image >> Image Size dialog. I used “Nearest Neighbor” for the resampling, to preserve the hard edges of each pixel. Then I set the device resolution to 180pixels per inch, and sent to the printer. This resulted in a nice, clean print of each color separation at the right size, theoretically suitable for tracing directly to the canvas.

I cut out each separation, put it on a light box, and laid the canvas over that, marking each thread intersection to be stitched with a Pigma marker (in needlepoint, a “pixel” corresponds to the cross where two threads of the canvas join; the yarn gets stiched around that.) Marking in four shades of green would be impossible to tell apart, so I mapped the colors: black for the darkest green, green for the next lightest, red for the next lightest, et cetera. The canvas ended up looking like a thermal-heat map, or a particularly lurid paint-by-numbers design.

Frankly, this didn’t work so well. Needlepoint canvas isn’t a strictly regular 18 threads per inch, and unless you possess superhuman canvas-stretching skills, it’s not really square, either. So I think I was working against the medium, here, and the registration ended up being pretty significantly off. For example, my canvas has a _lot_ more of the lightest green (separation 7) than the original art does.

Of course, this does not matter — I think the design looks cool, and each step of the process was (mostly) a Fun Learning Adventure. One solution would be “start freehanding it, you obsessive nerd!” However, I enjoy having pixel-level control over the design before it’s stitched, so I think I’m going to continue to find a way to assign each thread-intersection a color explicitly. Like counted cross-stitch, except I much prefer the wooly texture and thick dimension of needlepoint. Maybe if I get really good at stretching a canvas square, I’ll have better luck maintaining precise resolutions.

Here’s a closeup of the mostly-finished design. What do you think?