To determine the hand, in the moving image search for the hand and determines their position in screen coordinates. Whose position is described by an image mask.
A pixel is marked if it assigned position of the hand can be. In addition, the location of the center point calculated by hand.
I would first like to narrow down the area of the image, the closest hand.
it is given a random search. The color values randomly selected pixels with the color values of the hand are compared.
What is the color values of the hand is determined in a histogram. As a guest
area now the area is delimited, in which the density of the pixels is highest.
in this area was limited, it is also the hand.
The area is defined by a starting point (xS, yS) and
an endpoint (xE, yE). Start and end point in image coordinates described.
picture: view hand detection and identification
hand
be seen in the image were identified a few pixels of the hand. Through a Segmentation in the restricted space BS the entire palm is exposed. For this, a region growing algorithm is used. As a starting point for the algorithm
the center of the area BS is chosen. From this starting point
the color values \u200b\u200bof all neighbors are compared with the color values \u200b\u200bthat are defined as the color of hand
. It defines no fixed color value but a range of colors.
If a color value in the desired range of colors, the pixel marked in an image mask.
Each field in the mask corresponds to one pixel in the video image. Then
the other neighbors are examined. The process stops when none can be found further
more pixels that correspond to the values \u200b\u200bdefined area of \u200b\u200bthe hand
or if the edge of the area BS is achieved. The result of region growing-
algorithm is the segmented surface of the hand.
side of the area is the edge pixel is determined and marked in the image mask.
by the contour of the hand is determined.
Next, check whether the segmented area and the contour of the hand is
. But two features are defined, being tested set of rules: first
The number of segmented pixels in relation to the surface of the fenced area BS.
second The number of edge pixels in proportion to the area BS.
Are the characteristics of the segmented areas the desired characteristics such
defines the area as a hand. Otherwise, the process is interrupted and continued until the next
recorded video image.
Source:
picture and procedures from the review:
process for the preparation of hands in an interactive Augmented Reality applications.