ABSTRACT PRINCIPLE see autopoeisis
(for an example of autopoeitic algorithms http://www.eeng.dcu.ie/~alife/bmcm9702/bmcm9702.html see)
ORGANIZING PRINCIPLE fmo1 -→ DCT
[[DCT) and what mpeg4 ennables - see: http://www.eetimes.com/design_library/ad/m/OEG20030108S0043 MPEG-4 toolkit and see http://www.xiph.org/archives/tarkin-dev/200201/0008.html current state of video coding research
fm01 may incorporate byproducts effected by the contingent extraction of features of interest from heterogeneous large data sets, by the ad hoc mapping and convergence of data within those sets -→ emergent/mutating training thresholdsthese byproducts may at times overlap with formal image analysis tools, but their classification is not an organizational objective of fm01
NOTES
1) fm01 may generate (quasi) functional tool attributes eg. http://dynamo.ecn.purdue.edu/~ace/msel/msel.html edge recognition (hard or soft versions), iconic change,eg http://citeseer.ist.psu.edu/black98framework.html a framework for modeling appearance change in image sequences, + http://www.cs.brown.edu/people/black/ see etc as contingent by-products, softly perturbing the structure of fm01 at any given time, but are not treated in fm01 as implementable organizing principles
2) the http://www.genie.lanl.gov/ GENIE model is an anexact approach to imagery analysis in which http://www.1010.co.uk ap have an interest has been developed using genetic programming design principles in creating genetic imagery exploitation software http://www.genie.lanl.gov/publications.shtml see also Brumby, Mitchell et al
- its antecedents are from: ‡ evolutionary computation (John Holland. John Koza, etc http://www.cscs.umich.edu/~crshalizi/research/ see) ‡ and genetic programming as applied to: - edge detection – http://www.rfai.li.univ-tours.fr/rousselle/docum/ALGEN004.htm see - face recognition – http://www-2.cs.cmu.edu/~coral/publications/epia95.html see - image segmentation – http://citeseer.nj.nec.com/poli96genetic.html see - multispectral imagery - http://citeseer.nj.nec.com/brumby00genetic.html see - feature extraction – http://ltpwww.gsfc.nasa.gov/MAS/doc_new.html see King et al (.pdf)
genetic programming – brief design overview ‡ a population of candidate solutions (for GENIE image processing tools) is evaluated according to “fitness” with regard to its given environment of training data
‡ selection proceeds from using a fitness function, reproduction is attended by crossover and mutation with elitism as an option
environment: training data ‡ good quality training data as some “ground truth” and marked up ie. by known algorithms (eg.see http://citeseer.nj.nec.com/cachedpage/377748/2 genie 1 and http://216.239.39.104/search?q=cache:1PDAwDx1cCsJ:www.genie.lanl.gov/green/publications/hirschSPIE4132.pdf+ndvi+genetic+algorithm&hl=en&ie=UTF-8 genie 2 user training data, provided via a GUI, according to application) ie the specification of a truth plane presupposes some known feature of interest and also assumes, firstly, that it can be re-identified (where pixel feature = 1, non – feature pixel = 0), secondly, that several planes can be embedded with different weightings of pixels (extended as byte or float types)