Shape Histogram
The Shape Histogram module is a type of histogram transform and can be used as part of an object
classifier. A binary image is used to generate
a histogram that represents the run-length values of the given image in each of 4 direction.
The advantage of creating a histogram based on a shape's pixel-length span in many directions is that it
reduces orientation dependency and produces a similar histogram regardless of the shape's orientation.
For example,
if a square shape is being processed (100 width by 200 height) the algorithm will encounter the top left
corner of the square first and
proceed to the right corner counting how many pixels the top run length of the square is. Once a black
pixel is encountered the span is considered terminated and the span length is added into the histogram
by incrementing
the bucket in the histogram that represents the span size (in this case 100) by 1. This process of counting
the span lengths is then
continued for each horizontal line. Depending on the selected checkboxes further orientations may
be processed in a similar manner.
Interface
Instructions
1. Specify which axis to analyze. Using fewer axes will result in histograms that are more
dependent on orientation.
2. To smooth the histogram to a more usable shape select the amount of smoothness to
apply to the histogram.
3. Normalizing the histogram will scale the values to occupy the maximum range of
the histogram and can be used to view more detail of the histogram.
Example
| Source | Shape Histogram |
 |  |
| Roatated 45 degrees | Rotated 90 degrees |
 |  |
See Also
Histogram
|
