Databases Reference
In-Depth Information
FIGURE 8.8: Store/recall times for each subpattern in each DHGN
subnet during the edge recognition process.(With kind permission from
Springer Science+Business Media: Neural Process. Lett., “Distributed Multi-
Feature Recognition Scheme for Greyscale Images,” vol. 33, issue 1, pp.
45-59, (February 2011), Muhamad Amin, A.H., and Khan, A.I., Fig.12,
http://dx.doi.org/10.1007/s11063-010-9163-8.)
of the images in less than 30 seconds. The processing times will substantially
be less for a real computational network with parallel processing resources.
These speeds make it possible to process live image data streams and large
data sets in real time.
The DHGN multi-feature scheme provides a highly e
cient and scalable
mechanism for multi-feature pattern recognition on coarse-grained computa-
tional networks. This multi-feature recognition approach represents a holistic
process where more features can be taken into consideration without any
changes to the approach. The scheme has shown to be highly scalable and
the processing time and recognition accuracy are not adversely affected with
the increase in number of processed patterns. The approach discussed in this
section works well on gray-scale images and it can be applied to a number of
fields that require gray-scale image analysis. The flexibility to include any im-
age feature at any point creates a“plug-and-play”capability for dynamic image
analysis. This scheme opens up the possibility for real-time image recognition
on Internet-scale data sets in biomedical imaging and video streaming. Fur-
thermore, through distribution of features, the DHGN is capable of performing
the recognition process on patterns with increasing size and dimension. Note
