False-Positive Detection in Graph Specifying Additional Contextual Clues for Recognizing of Image Classes

The core idea is that a context call be described by a recursive method. This template describes the configuration and properties of sub-regions in the image. An image with sub-regions that match the template is an example of that context. It is generated the set of models which it applies to the database. One of main examples is of a "snowy mountain" template. It is used qualitative models to represent and recognize image classes. The qualitative models are formed by generating pair-wise relationships between groups of pixels. An image class is represented as a graph specifying necessary relationships between pixel groups. If the graph for a given image class can be matched against an image, that image is part of that class. It is defined four features that can be used in such a model: size, position, luminance, and color.