Abstract: The lithography processes, especially the defects introduced, have a direct impact on the yield of the integrated circuit (IC). Diagnosing and classifying lithography defects, which exhibit a great variety of shapes, are very important to improve the yield. As a result, the skelecton of defects is a key feature that can be used to describe the shape of defects and to recognize the defects. In this paper, a method for extracting the skeleton of IC real defects based on mathematical morphology is presented. Defined by means of the luminance (L) and saturation (S) space, the LS space eigenvalue is used in clustering segmentation for the IC real colour defect image and the wavelet decomposing is used in segmentation for the IC real gray image. Then, the proposed method removes noisy spurs and extracts a skeleton set from noisy images. This approach is very useful in the IC extra material defects recognition. Experimental results show that the proposed method gives a simple description of IC defect shapes, which provides the foundation for detection and classification of defects accurately.

Key words: IC defect characterization, LS color space clustering, morphological filtering, skelecton extraction