AI-Based Image Processing for Photoresist Latent Image Enhancement : AM: Advanced Metrology

A new image processing and AI based methodology is described that accurately locates alignment marks in the latent images of an exposed wafer before post exposure bake (PEB) despite minimal contrast. The detected marks will be used for the alignment of low-contrast PEB images and allow multiple exposures without the need to develop the photoresist or PEB for each exposure. The critical steps in the algorithm are to remove high contrast elements from the image and perform adaptive histogram equalization followed by template matching to determine the location of the marks in units of pixels. In 98% of the test cases, the algorithm correctly located the alignment marks; processing an image requires 3.5 sec on a consumer laptop.