Sample Size Comparison in Area-Based Image Matching: Cross-Correlation and Mean Relative Gray Value Difference

Abstract Geospatial information is data about a specific location with geographic coordinates that can be collected, modified, and displayed in real-time. Geospatial data can be collected through field surveys and aerial photogrammetry. Aerial photos and Digital Elevation Models are the raw results of aerial photography. An orthophoto is a photograph that shows an object in its true orthographic position. The production of orthophotos involves a process that produces stereo images through image matching using various methods. In this research, area-based and feature-based image matching are performed. For area-based image matching, sample sizes of 9×9, 15×15, and 33×33 pixels and two correlation equations were used: Cross-correlation and Mean Relative Grey Difference. Image matching results between samples from the right and left photographs were compared based on the resulting correlation values. From the data processing, area-based image matching with a sample size of 33×33 gives the best results that high Cross-Correlation values and low MRGVD values indicate strong image matching accuracy, especially for images with many details and variations. Thus, the combination of cross-correlation and MRGVD metrics provides an effective and comprehensive analysis for image matching.