Skip to Main Content
Current methods of processing television images at the receiver are either performed globally, or locally with globally constant parameters. By contrast, greater levels of enhancement can be achieved if television images are segmented into regions of similar features. Then, each region is enhanced using algorithms optimized for the local properties of the segment. New methods of segmentation have been developed to allow for real-time processing at video rates. Some examples of this include grass and sky detection. We present models for color, luminance, position and texture that have been successfully used to identify grass and sky areas in TV images.