Browse Prior Art Database

Variable Core Size Linear Transformation Program

IP.com Disclosure Number: IPCOM000076514D
Original Publication Date: 1972-Mar-01
Included in the Prior Art Database: 2005-Feb-24
Document File: 3 page(s) / 43K

Publishing Venue

IBM

Related People

Herbst, N: AUTHOR [+3]

Abstract

Object of Program. This is a program for performing general linear transformations of the coordinates, comprising translations, rotations, scale changes and any combination of these operations, on two-dimensional arrays and subarrays. Such transformations can be specified by the coordinate values of three pairs of corresponding, noncollinear points. Images of interest are typically of the order of 4000 x 4000 pixels (2/24/ bytes), hence efficient means of carrying out the transformation required.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

Page 1 of 3

Variable Core Size Linear Transformation Program

Object of Program.

This is a program for performing general linear transformations of the coordinates, comprising translations, rotations, scale changes and any combination of these operations, on two-dimensional arrays and subarrays. Such transformations can be specified by the coordinate values of three pairs of corresponding, noncollinear points. Images of interest are typically of the order of 4000 x 4000 pixels (2/24/ bytes), hence efficient means of carrying out the transformation required.

Images appear in as many formats, as a rule, as there are data sources and must be converted into suitable input format for each processing step and for each output device, as required. Both the input and output array size and data formats are usually beyond the control of the problem investigator and present a time-consuming obstacle before useful work can begin. The transformation as described here is useful in editing or forming mosaic images to meet various I/O requirements, as well as performing translations, scale changes, or combinations of affine shifts for image registration. Data from remote sensing sources often show discrepancies over the same scene for different sensors, and comparisons of different runs over the same area generally requires registration for automatic computer processing.

The linear nearest neighbor transform, because it makes efficient use of computer resources, is also envisioned as a subprocessor for more general projective or Mercator mappings where the error over restricted mosaic portions of the image could be held within a given bound. General Scheme.

Any processing of an image of a size on the order of 2/24/ bytes would, of course, require some means of dividing the process into operations on subfields of the image. Because of the linearity of the transform, similarly shaped rectangular segments of the original image will transform into similarly shaped segments of the resultant. For brevity's sake, A and a will be taken to refer to the original image and one of its segments, respectively, and likewise B and b for the resultant. Fig. 1 shows given transform "T" transforming A and B.

Fig. 2 is a diagram of the segmentation and computation scheme. Note that working core holds only segments a', b and a transformation table, where a' is a rectangle circumscribing a as shown in Fig. 3. In general, it is necessary to bring all of a' rather than just a into core from the bulk store, because the images are stored in line-by-line or column-by-column form.

For a given size of...