The novel concept of Morphoscope enables reliable and reproducible
identification of visually relevant image structures. It considers the
context- and domain-dependent expert-knowledge of users and is
especially suited for the high demands of
biomedical image analysis. The
well-defined interface combines transparently advanced methods of
image understanding, such as nonlinear image processing and
user-controlled machine learning:
Morphoscope represents a novel modular approach to content-based image
analysis. It consists of two modules mimicing central aspects of
human perception:
Object extraction:
- Identification of all visually perceivable image structures
- Content-based object interpretation
Object evaluation:
- Observer provides expert knowledge by a "point-and-click" approach
- Knowledge is modelled, learned, and applied to the image objects
|