Directional Selection Property of 2-Dimensional Complex Discrete Wavelet Transform and Its Application on Defect Inspection of Semiconductor Wafer Circuits

Abstract

Previously, Discrete Wavelet Transform(DWT) is widely used as the image processing method. Some of the DWT, Complex Discrete Wavelet Transform(CDWT) and Complex Wavelet Packet Transform(CWPT) have directional selection that can detect the directional edges from the high frequency components. This property is expected to be used for shape and texture analysis such like a circuit pattern images. However, the principle of directional selection is still unclear. Therefore, fundamental properties such as the detect direction and detection angular range(resolution) are also still unclear. Thus, it is difficult to apply the defect detection that has the directional shape and textures. In this research, we firstly consider the principle of directional selection and we proposed the evaluation method of directional selection based on the consideration result. Proposed evaluation method can make clear the detect direction and resolution quantitatively. Finally we apply the directional selection to defect recognition of semiconductor wafer circuit in combination with Kmeans method. As the result, directional selection of 2D-CDWT and 2D-CWPT achieved highest recognition rate and lowest miss recognition rate compared with previous DWT and FFT.