Image Processing in 2-D Array Space with a Minimized Amount of Computation

摘要

In this paper, a computation scheme for parallel image processing in a 2-D pixel array structure is proposed. To make the structure simple, each pixel has access to only four of its eight nearest neighbors. However, despite this limit, it is possible, in this scheme, to implement spatial domain computations with a largesize neighborhood. The expansion of the neighborhood can easily be done by means of a simple multi-phase procedure. To this end, the computation in a pixel is decomposed into very simple calculations that are performed by the multi-phase procedure. The proposed scheme makes good use of the feature of partially superposed neighborhoods of adjacent pixels. The result of every simple calculation in each step of the multi-phase procedure is used by a maximum number of pixels in the neighborhood, which makes the computations in the pixel array redundancy-free. The proposed scheme can lead to an easy implementation of a simple pixel circuit for fast image processing.