It has been recognized for a long time that a conventional sequential processor is inefficient for operations on pictorial data where relatively simple operations need to be performed on a large number of data elements (pixels). Though many parallel processing architectures for picture processing have been proposed in the past, very few have actually been implemented due to the costs involved. With LSI technology, it is becoming possible to realize parallel architectures at a modest cost. In the following the authors review some of the proposed architectures for pattern recognition and image processing.