Mark Eaton

Homework 3b

Shape Analysis with basic Rubine Features

Shape 1

The bounding box for this shape is easy to picture, and is close to a square. So the bounding box angle should be close to 45 degrees (.785 radians), which it is (.82704). The fairly long stroke length, and small start to endpoint distance are clear to see.

There is a large relative and absolute rotation, which matches the spiral shape. One thing I don’t understand is why there is so much difference between the relative and absolute rotation. I would think that they would be closer, but I guess some of the jitter/jiggling of the line accounts for this.

Shape 2

This is very similar to Shape 1. Just as in the previous example, you can clearly picture the bounding box, and the relative and absolute rotations make sense for the shape. It is odd to me that the relative and absolute rotations are closer in this example. Just by looking at the picture, I can’t really see why these are closer in Shape 2 than in Shape 1.

Shape 3

F1 and F2 are very telling in this case. Since the cosine of the starting angle is 0, and the sin is 1, the gesture began straight up. In the picture, you can see that the left endpoint began in a vertical direction.

The high absolute, negative relative rotation fit the shape. The low sharpness value fits for the smooth, loopy shape. The max speed is very high, so I guess at some point this drawing was created with a high velocity (or the computer lagged).

Shape 4

Based on the starting/ending angle F1 and F2 values, I would also guess that this shape began with the left endpoint. The high absolute rotation makes sense. I think the relative rotation is so low because of the small dip at the end which brings it much closer to vertical (how the line started).

The shape seems fairly angular, so I would think the sharpness would be more. Strangely enough (to me), is that the sharpness is actually lower than Shape 1 (a spiral).

Shape 5

Compared to the other shapes, this shape should have a high stroke length. Looking at the numbers, I think it is the longest of all 8. The bounding box would be fairly squished (far from being square). You can see that the diagonal for the box would be fairly long, since there is large difference in the minimum and maximum x values. The relative rotation is very close to 0, and you can see in the image that both endpoints start/end at almost the same angle.

Shape 6

This shape should have a long stroke length, and it is very angular. The sharpness value should be high. According to the data, it is the most “sharp”, second only to Shape 8. Just by the image, I think the relative rotation should be a bit more, but the right most endpoint does go back to more of a horizontal path. This shape happens to have the longest total time, but there is not really a way to tell or evaluate this simply by looking at the pictures.

Shape 7

It’s a vertical line, so there should be no relative and absolute rotation. However, there is quite a bit of absolute rotation. I guess these accounts for very subtle (but constant) movement left and right (since humans aren’t good at drawing straight lines). The cosine / sine values of the angle between start and end point match up perfectly with what you would expect form looking at the drawing.

This line is supposedly the smoothest of all shapes. Looking at it, it’s not terribly smooth in my opinion. And if you compare it to the spirals, can you say a line is smoother? Granted, at this stage the code doesn’t know it is supposed to be a line, but since there really isn’t a major angle change the sharpness value seems moot to me.

Shape 8

Since the line is horizontal, it should have inverted values compared to Shape 7 for F6 and F7. This is indeed the case; it is almost 1 and 0. A really odd thing about this line is the ridiculously high absolute rotation. This makes a good argument for some kind of jitter control. Since movement is in the horizontal, the vertical jitter (could argue it as noise, depending on the user’s intent) could be creating many cases in which the angle changes by pi. This would account for a high absolute rotation. Based on this, it seems to me that vertical jitter can really mess up the features. This also explains why the shape is the 2nd “sharpest” image. It shouldn’t be any different from the vertical case (equally as smooth). However, I think the vertical jitter is causing this.