To determine gesture for a specific image, we need to make the “hand” in the image easily be recognized. Therefore, we need to get rid of all the unnecessary parts which appear in the original image, and what we do is reference point detecting. During this procedure, we determined a reference point on the hand and cut out a new figure which cancels noises.

To determine the point, we first put the figure into a Cartesian coordinate as we shown in the figure (Figure 3). The Y value of the point can be determined by looking for which horizontal lines covers most gray spot. The X value can be determined by determining the edge point between the hand and the background, and then subtract it by a threshold value. The subtraction is done for the purpose of making the reference point shift a little bit to the center of the hand, and it is helpful when we do further calculation.

Furthermore, the way we count gray spot is shown below in mathematical way if you are interested in. If the difference of the pixel and its adjacent is smaller than 0.001 and its own value is smaller than 0.8, it is counted. The first formula restricts that it should not be the edge, and the second restricts that its color should be gray which corresponds to the hand (white pixel -> 1 and black pixel -> 0).