LEGO Minifigure head - AWFeldick/Rotostitch GitHub Wiki

In this tutorial, we'll go through the process of using Rotostitch to 'scan' a LEGO Minifigure head.

---

Prerequisites:

• An image sequence of an object being rotated 360 degrees
• 2 mins of time

The image sequence I will be using for this tutorial looks like this:

This is standard for most image sequences but bear repeating:

1. Each frame must be the same size
2. Each frame can have any name as long as the same number of characters at the end of the name are digits

The best results will be obtained if you keep the object (in this case the minifigure head) roughly in the center of the frame. Additionally, the axis of rotation should be as close to vertical as possible (in relation to the image sequence). How you obtain your image sequence is up to you. In general, you will probably take a video and then extract the frames from it.

Lets start!

Step 1

Once we have our image sequence, lets start Rotostitch, and load the sequence:

To open your sequence, simply navigate to the containing directory and select any image from the sequence.

Step 2

The sequences should open in just a moment (I zoomed out and resized the windows to get a better screenshot):

When the sequence is finished loading, the GUI will update left preview area with the first frame of the sequence, and the right with the last frame of the sequence. Everything else is left at their default settings. Lets take a moment to review the controls for navigating the image previews:

• Middle mouse button drag pans the previews (they are locked together, and you can drag in either preview)
• Mouse wheel roll zooms in and out on the previews (again, the mouse can be over either preview)

With our sequence loaded, the first thing that needs done is to tell the program where in the frame our object is by drawing the width cross-section line. We can see from the green 'light' on the width's 'Set' button that we are already in the right mode to do this. Zooming in and out as convenient, we can left click to place one end of the cross-section line, and right click to place the other. The main thing is to place the two green endpoints on the silhouette edges of the cylindrical object we want to 'scan':

The blue vertical line that updates along with our cross-section line shows the 'rotation' axis of object and the approximate area around which the vertical frame strip will be taken from each frame in the sequence.

Step 3

With the width cross-section line set, we now have to set the two height lines. Because of the perspective on the object, the center of the object appears larger in the frame than it should. This would slightly mess up the aspect ratio on the final output unless we do something about it. Thus we need to set two height lines that allow the program to calculate a simple ratio multiplier to correct the final image.

So, switch the mode from width to height and using the same method as in setting the width line set a line along one silhouette edge (preferably one of the silhouette edges that the width line ends at). We place the second line using the same mouse buttons except we also hold the 'Shift' key while doing so. When placing the second line, the idea is to place it on the image so that if both lines where actually there in reality, they would be the same length. It should also be place approximately around the rotation axis line of the object. All this may sound confusing, so just take a look at the picture, it really isn't to hard:

We can see from the picture that the lines would be approximately the same length in real life. I placed them by 'eye-balling' the pink, left line and then using the face printing to set the cyan line's bottom point, and the faint molding line to set the top. It doesn't really matter which sides the pink and cyan lines are on. In general, they will probably be placed in about the same positions (near one edge, and near the center).

Step 4

We can now choose our first and last frame. We can activate 'Difference Mode' on the left preview with the small button in-between and at the bottom of the previews. This allows us to see how similar the first and last frame are to each other. The goal is to set the first and last frame to be as similar as possible and between which the object rotated 360 degrees. If our images are really close, the left preview will be nearly black with Difference Mode active:

Step 5

The hard parts are finished! All we have to do now is select which direction our object is turning and set a save path:

Make sure you put a full path including the output filename and extension. With that finished, simply hit 'Create'! You should see the updating status of the stitching and then the successful output message. If everything went well, you can then go find the result and marvel at its awesomeness:

Step 6

Step 6 is up to you! Hope this tutorial was helpful.