Browse Prior Art Database

Indirect, N Dimensional Displacement Addressing Mechanism for an Image Processor

IP.com Disclosure Number: IPCOM000045315D
Original Publication Date: 1983-Mar-01
Included in the Prior Art Database: 2005-Feb-06
Document File: 2 page(s) / 13K

Publishing Venue

IBM

Related People

Gonzalez, J: AUTHOR [+2]

Abstract

For addressing picture elements (pixels) in the vicinity of a current operating point during image processing operations, the following technique is suggested: The relative displacement between a pixel and the operating point is inserted as the immediate operand into the instruction for that pixel. A table comprising the absolute displacement values (based on selected dimension, etc.) is provided which can be looked up by using relative displacements as selectors or addresses. When a pixel is being evaluated, i.e., the respective instruction is executed, it is only necessary to access the table using the relative displacement to obtain the absolute displacement. This saves arithmetic operations which would otherwise be necessary during execution.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 54% of the total text.

Page 1 of 2

Indirect, N Dimensional Displacement Addressing Mechanism for an Image Processor

For addressing picture elements (pixels) in the vicinity of a current operating point during image processing operations, the following technique is suggested: The relative displacement between a pixel and the operating point is inserted as the immediate operand into the instruction for that pixel. A table comprising the absolute displacement values (based on selected dimension, etc.) is provided which can be looked up by using relative displacements as selectors or addresses. When a pixel is being evaluated, i.e., the respective instruction is executed, it is only necessary to access the table using the relative displacement to obtain the absolute displacement. This saves arithmetic operations which would otherwise be necessary during execution. If the image size is changed, a new table has to be loaded, but the instructions of the compiled program containing immediate operands specifying relative displacements need not be changed. A detailed description and an example are given below.

Kernel operations play an important role in digital image processing. The easiest example is filtering in which each pixel of an image to be computed is obtained as a weighted average of some surrounding pixels of the input image. Thinking in a two-dimensional case (the idea presented here can be easily extended to n dimensions), the function to be computed is: O(i,j)=SUM(H(n,m) I(i-n,j-m)) (1)

n,m where I() is the input image, O() is the resultant image, H() is the filter kernel, and i, j, n and m are indices.

When the processor memory is organized linearly, i.e., a one-dimensional array of successive memory locations, and...