Area thresholding and silhouette extraction of simple coloured objects using hue

The paper describes a method of extracting object silhouettes from a hue image. The stages involved are first to extract the hue from the red, green and blue images. Once the hue has been extracted it is processed by an area thresholding algorithm that extracts objects corresponding to a specified hue value. If there are a number of objects with similar hue then the algorithm stores the objects in separate records. Objects with small areas are filtered out. The resulting objects are then processed by simple convolution kernels to obtain a silhouette outline which can be analysed to detect features such as corners, curves etc