Ultrapointe Corporation

Kla Tencor Corporation

Ke Han

Ken K Lee

Bruce W Worster

Lakshman Srinivasan

Dilation and erosion operations are performed on imaged objects that are represented using a number of object pixels; the object pixels are a subset of a larger number of image pixels. A ring-shaped structuring element is specified by a radius and an origin. In the ring dilation operation, the boundary of the structuring element includes two concentric circles, one having a radius that is one pixel greater than the other so that the boundary is two-pixels wide. The origin of a copy of the structuring element is then overlaid onto each object pixel such that the boundaries of the copies intersect to form a dilated image feature. The image feature includes a feature boundary defined by the outermost portions of the boundaries of the overlapping copies of the structuring element. This feature boundary is identified and described using conventional chain-coding techniques. The dilation is then completed by filling an area defined within the feature boundary. To erode objects, a second ring-shaped structuring element is specified by a second origin and by a second radius. The origin of a copy of the second structuring element is overlaid onto each pixel within the object. A pixel within the object is selected only if the boundary of the copy of the second structuring element lies completely within the object when the origin of the copy is overlaid onto that pixel. The selected pixels represent an eroded version of the object.

Ring dilation and erosion techniques for digital image processing

A method for automatically generating a knowledge database in an object classification system having a digital image data source, and a computer, includes the steps of inputting digital image data corresponding to a plurality of training images, and characterizing the digital image data according to pre-defined variables, or descriptors, to thereby provide a plurality of descriptor vectors corresponding to the training images. Predetermined classification codes are inputted for the plurality of training images, to thereby define object class clusters comprising descriptor vector points having the same classification codes in N-dimensional Euclidean space. The descriptor vectors, or points, are reduced using a similarity matrix indicating proximity in N-dimensional Euclidean space, to select those descriptors vectors, called extreme points, which lie on the boundary surface of their respective class cluster. The non-selected points interior to the class cluster are not included in the knowledge database. The extreme points are balanced by eliminating functionally redundant extreme points from each class cluster to provide a preliminary knowledge database. Fine tuning of the preliminary knowledge database is performed by either deleting extreme points that tend to reduce the accuracy of the database, or adding new rules which enhance the accuracy of the data base. Alternately, the fine tuning step may be skipped.

Automatic knowledge database generation for classifying objects and

10A method is described for detecting and characterizing defects on a test surface of a semiconductor wafer. A three-dimensional image of the test surface is aligned and compared with a three-dimensional image of a defect-free reference surface. Intensity differences between corresponding pixels in the two images that exceed a predefined threshold value are deemed defect pixels. According to the method, the pixels of the reference image are grouped according to their respective z values (elevation) to identify different physical layers of the reference surface. Because different surface layers can have different image properties, such as reflectance and image texture, the groups of pixels are analyzed separately to determine an optimal threshold value for each of the groups, and therefore for each layer of the reference surface.

Growing community of inventors

San Francisco, CA, United States of America

Ke Han