Directional dilation for the connection of piece-wise objects: a semiconductor manufacturing case study

A technique for intelligent dilation of objects in an image was developed that, when compared to standard isotropic techniques, can create a better representative resultant image for subsequent morphological analysis. Individual masses in an image are forced to grow in preferential directions with varying degrees of strength according to a set of morphological parameters. These parameters are determined by applying a model based on gravitational force to determine the strength and direction of the “attraction force” between objects. The result of this analysis is a collection of normalized dilation vectors for each object in the image, where each vector represents a potential direction for growth. Results of the application of this algorithm are shown on binary images of defect distributions on semiconductor wafers where a single defect (e.g., a scratch) can be made up of numerous, disconnected blobs