In geometric shape search, matching is performed by extracting contour shape from both model and target image. Compared with the normalized correlation search which calculates the normalized correlation between the region of the target image and the model, it has the advantage that it is resistant to disturbance such as irregular brightness change. Moreover, flexible search is possible for rotation and scale change.

For example, in the next image, the brightness changes to a gradation like compared to model registration, and the image changes like spots, chips, etc. If normalized correlation search is used for such images, the correlation coefficient decreases, so detection becomes unstable. In the case of geometric shape search, matching is performed based on the contour extracted from the image, so matching becomes high score and stable detection is possible unless there is no problem in contour extraction.

In geometric shape search, the performance of contour extraction greatly affects accuracy, speed, and robustness. Our X-Match® enables high-speed and high-precision contour extraction with accurate subpixel accuracy with original algorithm.