Importance assignment to regions in surveillance imagery to aid visual examination and interpretation of compressed images

A simple bottom-up, context-free, algorithm for the identification of regions-of-interest in surveillance imagery that correlates well with the visual attention processes in humans is presented. The aim is to develop new criteria related to human performance in visual examination and interpretation, and perhaps to be able to design a more biologically plausible perceptive JPEG2000 image compression system. The paper introduces a quadtree computational platform to assign the importance of regions in an importance map, using several bottom-up, low-level factors, which have been found to influence visual attention. This technique is useful for images with objects or structures with importance dependent on resolution and field of view. Further work incorporating higher-level visual factors and context dependent criteria will allow regions of importance to be determined with greater accuracy