Using Edge SIFT Points for Simple Shape Object Matching

SIFT the well-known image recognition algorithm performs reliable matching between different views of an object even it is rotated, in different scales, or has different brightness than of the original image. But in our button match works, the SIFT points detected in simple object may be few, and the wrong match between objects that has same textures but different shapes may lead wrong matching results. To address this problem, we only retain the detected SIFT points that on the edges and also compared the edge SIFT points with the classic SIFT points, then combined them to perform a better result. Experiments demonstrate that morphological operation tried to decrease the light problems and get a substantial improvement.