Hi,
I’ve built a prototype of a multitouch-panel that provides a grayscale depth-map.
The blobs aren’t white. The brightness of the blob is dependent on the depth (finger pressure) and
the blobs aren’t sharp. You get blobs which look like mountains.
It’s like looking for peaks on a topographic map.
What would you do to find local maxima? Or do you know any well-known algorithms?
And which algorithm is the quickest?

Some possible solutions:
1. Convert grayscale-image to bw-image then do blob-recognition. And then determine the brightness of the centroid pixel of the blob.
-Not good when the blob isn’t round and when you have mointains ontop of a plateau.
2. Apply hill-climbing algorithm line by line (vertical and horizontal) and combine the acquired local maxima.
3. Separate the image into a series of images with each image containing only one depth layer. Then apply optical-flow algorithm.
-Very many steps
4. Separate the image like solution no. 3 but then apply a classification algorithm which finds circles without holes inside circles with holes. (onion-ring)
-Very many steps; Not good when sensor-image has too much noise or isn’t crisp.

This is not the actual sensor image. Just a sample how it could look like.

David