A computer program enables assessment of the quality of tie points in the image-correlation processes of the software described in the immediately preceding article. Tie points are computed in mappings between corresponding pixels in the left and right images of a stereoscopic pair. The mappings are sometimes not perfect because image data can be noisy and parallax can cause some points to appear in one image but not the other. The present computer program relies on the availability of a left→right correlation map in addition to the usual right→left correlation map. The additional map must be generated, which doubles the processing time. Such increased time can now be afforded in the data-processing pipeline, since the time for map generation is now reduced from about 60 to 3 minutes by the parallelization discussed in the previous article. Parallel cluster processing time, therefore, enabled this better science result. The first mapping is typically from a point (denoted by coordinates x,y) in the left image to a point (x',y') in the right image. The second mapping is from (x',y' ) in the right image to some point (x",y") in the left image. If (x,y) and(x",y") are identical, then the mapping is considered perfect. The perfectmatch criterion can be relaxed by introducing an error window that admits of round-off error and a small amount of noise. The mapping procedure can be repeated until all points in each image not connected to points in the other image are eliminated, so that what remains are verified correlation data.
This program was written by Gerhard Klimeck and Gary Yagi of Caltech for NASA's Jet Propulsion Laboratory.
This software is available for commercial licensing. Please contact Don Hart of the California Institute of Technology at (818) 393-3425. Refer to NPO-30632.
This Brief includes a Technical Support Package (TSP).
Software for Verifying Image-Correlation Tie Points
(reference NPO-30632) is currently available for download from the TSP library.
Don't have an account?
Overview
The document outlines NASA's Technical Support Package for "Software for Verifying Image-Correlation Tie Points," identified as NPO-30632. This software is developed to enhance the quality assessment of tie points generated during the correlation of stereo images, which is crucial for various aerospace applications.
The primary challenge addressed by this software is the potential for inaccuracies in the correlation process. When correlating stereo images, points can be incorrectly identified as valid or may be missed entirely, depending on the direction of the correlation (left-to-right or right-to-left). The software employs a two-way correlation approach, which significantly mitigates these issues by comparing results from both directions. This method allows for the elimination of spurious tie points—those incorrectly deemed valid—and the restoration of valid tie points that might otherwise be overlooked.
The correlation process involves mapping every pixel in a reference image to a corresponding pixel in its stereo pair. This is executed through a parallel algorithm that divides the reference image across multiple CPUs, enhancing processing speed. However, the quality of this mapping can be compromised due to factors such as parallax (where the left and right images may capture different objects) and noise in the image data. To assess the quality of the correlation, the software checks the mapping from left to right against the mapping from right to left. A perfect correlation is defined as the original point and the point returned from the reverse mapping being identical. An error window can be introduced to account for minor discrepancies due to noise and interpolation errors.
The document emphasizes the importance of this software in ensuring the reliability of tie points used in further analysis, which is critical for applications in remote sensing, robotics, and other fields that rely on accurate image data. For further inquiries or assistance, contact information for the Jet Propulsion Laboratory (JPL) is provided, highlighting the collaborative nature of NASA's efforts in advancing technology for broader scientific and commercial applications.
In summary, this Technical Support Package presents a robust solution for improving the accuracy of image correlation in stereo imaging, thereby enhancing the quality of data used in various technological and scientific endeavors.
