The Neighboring Pixel Representation for Efficient Binary Image Processing Operations

Many image processing operations such as: segmentation, contour tracking, morphological operations, etc. needs an additional information about pixel neighborhood. In the other hand memory access operations are often (comparing to working time of other hardware devices) time-consuming, so checking values of neighboring pixels via memory access could lead to low performance of the algorithm. Our binary pixel representation contains information of it´s neighborhood on one byte only, which is stored at the pixel´s address. Moreover, because of information about zeros in neighboring pixels, computational time can be efficiently reduced - no need to checking memory cells where zero values are stored. Putting the pixels´ values in dual-memory bank two image processing operations can be done simultaneously. This paper presents the algorithm of converting binary image to our notation, it´s hardware realization and description of basics binary image operations based on the new notation. As an example of usage the notation the hardware realization of contour tracking algorithm was introduced as a preprocessor for contour shape descriptor of MPEG-7 standard