pattern segmentation request

[Neural Network Group]

Using combinations of classical transforms (fourier-transform, 
KLT and modifications) we constructed some modules for feature 
extraction and made investigations with recurrent NN-topolgies based on
statisticel methods like for example fractal NN for document-analysis
problems. 

In our studies we've found some help in publications from National Institute
of Standards and Technology in the following papers:
ir_4766.ps      Optimization of Neural Network Topology and Information
                Content Using Boltzman Methods
ir_4776.ps      Training Feed Forward Neural Networks Using Conjugate
                Gradients
ir_4824.ps      Karhunen Loeve Feature Extraction for Neural Handwritten
                Character Recognition
ir_4893.ps      Topological Seperation Versus Weight Sharing in Neural
                Network Optimization

Here we got interesting results and a good performance in pattern
recognition tasks. Recognition of pattern is an interesting field at all, 
but there is also the problem of finding any pattern or a special pattern
in an image, the problem of pattern segmentation. 
Or in a more general case: How to find a special object at any places and 
different sizes in an image?

Referring to that problem I've read an article about pattern segmentation
using gabor functions and genetic algorithms from NIST (ir_xxxx.ps). 
But we could'nt get any good results in algorithms for object search using 
Neural Networks.

During last weeks I've read a lot of publications from connectionists
archive, but could'nt find much help in solving that problem.
Do you have any hints for me how to get some information about algorithms 
and methods in pattern segmentation or/and in a more general sense object
search-algorithms in 2-dimensional images?

Many thanks

                                Gratefully yours
				Welf Wustlich
                                from Neural Network Group Rostock