A method and apparatus for image processing provides a memory having a plurality of individual parallel buffers constructed from random access memories (RAMs) for storing data related to a group of image pixels. The buffers each store a parallel, identical version of the image data so that an image processor can access data related to a given pixel in the overall data from each buffer simultaneously. An address expander for the buffer rows and buffer columns is used to convert a row and column address of a selected “central” pixel into a plurality of related pixel data addresses offset at predetermined distances from the selected pixel data's address. In this manner, the address expanders enable a group of related pixels, each in a different parallel buffer, to be accessed simultaneously, without requiring the processor to be interconnected with all of the buffers. This substantially reduces the complexity of processor interconnection design, while substantially enhancing processor speed.