Efficient computation of objects' spatial relations in digital images

Both quantitative and qualitative measures of directional spatial relationships (e.g., left, right, above) between two raster objects in a digital image are important for high-level computer vision tasks such as scene analysis and robot navigation. The histogram of forces can provide such measures, but cannot be computed in real-time. A new approach for real-time computation, based on a vector representation of raster objects, is presented. The performance of the proposed approach is examined in an extensive experiment. Considering processing time and accuracy, optimal assessments of directional spatial relationships for use in real-time applications can be obtained.