How PinPoint Works
Fundamentally, i2k PinPoint works by recognizing the same regions between two different images (nominally, these are the "base" image and the "target" image). When there are appearance differences between the base and target images - due to changes in illumination, modality, viewpoint, or scene content - the software only pays attention to what is similar. When enough image content is similar, the software decides that regions of the base and target images are recognized. When enough regions are recognized, the software decides the base and the target images are matched. Image regions that have not been recognized either do not contribute correspondences, or, in the case of pre-specified control point locations, contribute correspondences with very low weight.
When matching a target image against multiple base images, i2k PinPoint quickly evaluates each base image to see which are the most promising to match against the target image. It then focuses its effort on the best possibilities.
When handling multiple target images, each succeeding target image starts from the matching location (in the selected base image) of the previous target image. If this matching fails for a new target image, i2k PinPoint starts over, working as though the new target image were the first one.