Image Correlation on a Parallel Processor.

Digital photogrammetry requires that conjugate imagery be located by image correlation. Image correlation involves many computations and can be the most time consuming part of the digital photogrammetry process. This report, in line and area correlations, investigates the feasibility of performing the image correlation calculations on a parallel processor. Line correlation involves searching along epipolar lines using a one-dimensional window of gray shades. Area correlation pertains to using a two-dimensional window of gray shades to search for a match point either over an area or along a specific epipolar line. Functions include the linear correlation coefficient, the covariance, and the sums of absolute differences. The computer programs in this report were written for the GAC STARAN at ETL. The parallel processor may not reach the speeds of a special-purpose, hard-wired correlator, but it has the advantage of being readily reprogrammed. The parallel processor will be used in the interactive digital image processing facility at ETL. Author