Introduction - Hierarchical Neural Networks for Image Interpretation - page 10

bottom−up: top−down:

− data driven − hypothesis driven

− analysis − synthesis

− feature extraction − feature expansion

lateral: − grouping − competition − associative memory
Fig. 1.12. Integration of bottom-up, lateral, and top-down processing in the proposed hier-
archical architecture. Images are represented at different levels of abstraction. As the spatial
resolution decreases, feature diversity and invariance to transformations increase. Local re-
current connections mediate interactions of simple processing elements.

man visual system when designing computer vision systems. Although the human
visual system is far from being understood, some design patterns that may account
for parts of its performance have been revealed by researchers from neurobiology
and psychophysics.
The thesis tries to overcome some limitations of current computer vision systems
by focussing on three points:

- hierarchical architecture with increasingly abstract analog representations,
- iterative refinement of interpretation through integration of bottom-up, top-down,
and lateral processing, and
- adaptability and learning to make the generic architecture task-specific.

Hierarchical Architecture. While most computer vision systems maintain multi-
ple representations of an image with different degrees of abstraction, these repre-
sentations usually differ in the data structures and the algorithms employed. While
low-level image processing operators, like convolutions, are applied to matrices rep-
resenting discretized signals, high-level computer vision usually manipulates sym-
bols in data structures like graphs and collections. This leads to the difficulty of
establishing a correspondence between the symbols and the signals. Furthermore,
although the problems in high-level vision and low-level vision are similar, tech-
niques developed for the one cannot be applied for the other. What is needed is a
unifying framework that treats low-level vision and high-level vision in the same
way.
In the thesis, I propose to use a hierarchical architecture with local recurrent con-
nectivity to solve computer vision tasks. The architecture is sketched in Figure 1.12.
Images are transformed into a sequence of analog representations with an increas-
ing degree of abstraction. As one ascends the hierarchy, the spatial resolution of