A comparison of synthetic and human observer approaches to multispectral sensor resolution assessment

Resolution is one of the key parameters addressing the quality capability of a sensor. One approach to determining the resolution of a sensor/display system is to use a resolution target pattern to find the smallest "resolved" element using the system, which typically requires a human in the loop to make the assessment. This paper compares the results of a software approach to generate an effective resolution value for a sensor with human vision results using the same images. Landolt Cs were selected as the resolution target, which were imaged at multiple distances from multiple sensors. The images were analyzed using the software to determine the orientation of the C at each distance, resulting in a probability of correct orientation detection curve as a function of distance. Probability of correct orientation detection as a function of distance was also obtained directly from subjects that viewed the imagery. These curves were then used to generate "resolution" values for the sensor using the software results and the subject results. Resolution results for both the software and the participants were obtained for four different spectral band sensors as well as for fused images from two pairs of sensors.