added a local extremum blob detector

[albenoit]

In

"Retina enhanced SURF descriptors for spatio-temporal concept detection"
Benoit A., Strat T., Lambert P., Caplier A. to appear in Multimedia Tools and Applications (2014) (http://hal.archives-ouvertes.fr/hal-00760192)

we proposed a method to automatically detect blobs within video frames that present high energy at the output of the retina model (magno output already available in the bioinspired module of OpenCV). Those blobs have a high energy generated from spatio-temporal events (object appear, move) and we can focus image description on them. The proposed blob detector applies spatio-temporal filtering that automatically adapts to the local context and is stable in space and time.
We showed that describing them within Bag of Visual Words (BOW) toolchain, helps to identify visual concepts. We experimented with the TRECVID Semantic Indexing challenge datasets.
In a further paper such as

Retina enhanced SIFT descriptors for video indexing Sabin Tiberius Strat, Alexandre Benoit, Patrick Lambert: . CBMI 2013: 201-206

We showed that it works fine with SIFT based BOW description and that we can design new spatio-temporal descriptors using different preprocessing with the retina model and this blob detector.

We now propose it to the community within OpenCV.

[albenoit]

If someone want's and has time to review (any recommendation will be appreciated to improve).
i will add doc in a few time