Advanced Computer Vision With Object-Based Image Analysis

Common pixel-based image analysis methods cannot match the remarkable pattern recognition capabilities of humans. For the human mind, an image represents a complex web of interrelated objects at different scales. Prior knowledge of the characteristic properties and mutual relations of theses image objects enables humans to understand the image content.

Fraunhofer MEVIS has developed a comprehensive toolkit for analyzing images on the basis of objects inspired by the way a human would understand an image.

• Transformation of an image into an object-based representation
• Rich set of spectral, textural, shape, and relational object properties
• Declarative language for explicit knowledge-based evaluation of image objects
• State-of-the-art machine learning algorithms
• Simple interface for easy integration into different software platforms
• Visual inspection tool