Adaptive thresholding ASIC for image processing application

The ASIC implementation of a simple and effective image thresholding method is presented. The method uses a novel weight based clustering technique using a neural-network-like structure to compute a global threshold. The uniqueness of the method is its simplicity and ease of implementation. The method has been found to perform well on mailpiece images. It is a single pass algorithm that requires minimal post processing. The network structure and implementation issues are discussed.<<ETX>>