A new programmable VLSI architecture for histogram and statistics computation in different windows

The grey-scale histogram of an image has become important for image preprocessing methods. Histograms are employed in generating binary images (segmentation), data compression, image enhancement and elimination of inhomogeneous illumination. Even the co-occurrence-matrix (texture analysis) is obtained by performing simple increment operations, e.g. histogramming. A new concept and architecture has been developed to compute real-time histograms and statistics in up to four window areas. The size and location of these windows can be defined and the methods mentioned above can be verified. This paper introduces the recently developed VLSI architecture (histogrammer) and the required instances to realise window handling. The special configuration of the memory and arithmetic unit is explained. Furthermore, it describes other applications of the architecture for texture analysis and equalisation. Simulation results and implementation of the design are presented.