Binary model shapes were loaded in an ipython notebook with
skimage which was capable of handling
.pgm format :
Skeletons were generated by
morphological thinning implemented in
mahotas:

endpoints (blue), branched points (green) 
Endpoints Spectrum :
The skeleton of the hand has two endpoints close to top of each fingers, as it would be fine to have one endpoint per finger, let's prune the skeleton to meet this condition. This can be achieved after some trial, however if robust features could be found from the skeleton a systematic method is needed. So, let's prune the skeleton until it has no endpoint and count the number of pruning step necessary:
From the curve, the hand skeleton has 13 endpoints before any pruning, after 40 pruning steps, it has 6 endpoints. Around 90 pruning steps are necessary to remove all endpoints. By only looking at their curve, the hand shaped silhouette and the fish shaped silhouette are different.
Endpoints spectrum resists to scale variation after normalization:
As scale invariance is a nice feature, let's see what is happening when the scale of the shape is successively reduced by a factor of 0.9², 0.7², 0.5², 0.3²:
From the images above, it is clear that the skeleton resists to scale variations but as the skeleton becomes smaller less pruning steps are necessary to "burn" it to the last pruned pixels set. So let's norme the pruning steps by the initial skeleton size such for 0 pruning step, the normed pruning step is 100. As the size of the hand shaped silhouette decreases (from 81%, 49%, 25%, 9% of the original size), the endpoints spectrums remain comparable:
Endpoints spectrum of similar shapes are similar:
How does the curve behave with similar shapes? Let's compare the endpoints spectrum of the serie of six hands:
The spectrums seem to resist to shape rotation (hand, hand10°, hand90°) and to moderate shape deformation (handbent1,2). The sketchhand silhouette is a right hand and its spectrum is comparable to the other spectrums which corresponds to left hands, so the endpoints spectrum is symetry invariant too.
Deriving a features vector from endpoints spectrum:
If shapes should be classified according to their topology, the whole endpoints spectrum is not n
ecessary and only the transitions may be retained. Taking a hypothetical skeleton whose endpoints spectrum before normalization is:
 For 7 prunings the skeleton has 0 endpoints
So a possible features vector would be:
(7,6,1,2,0)
The value 1 indicates that the value 2 endpoints is missing. In the python script, the endpoints spectrum is coded as two lists:
end_points_test = [4,3,3,1,1,1,1,0]
pruning_steps_test = [0,1,2,3,4,5,6,7]
A python funct
ion taking the steps list and the endpoints (spectrum) list yields a features vector :
def EPSpectrumToFeatures(steps, spectrum):
'''
eptest = [4,3,3,1,1,1,1,0]
pruntest = [0,1,2,3,4,5,6,7]
the returned vec:[7,6,1,2,0] which means:
0 enpoints > 7 prunings necessary
1 endpoint > 6 prunings
4 endpoints> 0 pruning (initial skeleton size)
'''
values = set(spectrum)
feat_vect = [1]*(len(values)+1)
spec = np.asarray(spectrum)
for epn in values:
indexmin = np.where(spec == epn)[0][1]
feat_vect[epn] = steps[indexmin]
return feat_vect
Supposing now the the skeleton size is 25, let's normalise the pruning steps as follow:
normed_pruning_steps_test = [100*(1i/25.0) for i in pruntest]
The endpoints spectrum is:
Let's take a normalised endpoints spectrum:
normedsteps = [100.0, 96.0, 92.0, 88.0, 84.0, 80.0, 76.0, 72.0]
endpoints = [4, 3, 3, 1, 1, 1, 1, 0]
And let's submit it the previous function :
EPSpectrumToFeatures(normedsteps, endpoints)
The function yields the vector:
[72.0, 76.0, 1, 92.0, 100.0]
which means that the skeleton has:
 0 endpoint for a normed number of prunings of 72%
 1 76%
 no 2 endpoints, set normed number of prunings to 1
 3  92%
 4  100%
shape= Shape(image)
The feature vector can be computed by:
vector = shape.End_Points_Features(normed = True)
or by:
vector = shape.End_Points_Features(normed = False)
Feature vectors from 216 images :
216 images of small 2D binary images (silhouettes) [
Sharvit et al.] were used to derive feature vectors. This set of images consists in 18 type of silhouettes with 12 instance of each. For example, there are 12 birds si
lhouettes:
Prior any kind of classification the 216 normed feature vectors were aligned for visual inspection:
Some silhouettes, as the
Bone silhouettes seems to show some intraclass structure, while
Brick and
Camel silhouettes seems different.
ipython notebook (updated 24/05/2013):
All the tasks were performed in a
ipython notebook available here.