Pattern image enhancement by automatic focus correction

Abstract Optical analysis techniques are key tools for the failure analysis and defect localization in integrated circuits. Using a confocal laser microscope, it is possible to extract different pieces of information such as spatial distribution of signals or voltage waveforms. Blur is getting a more and critical issue as technology pitch is getting smaller, very close to resolution limits. Find the correct focus, is a recurrent problem to solve in optical microscopy. The expert has to correct the blur disrupting the image, by manually searching the good focus. By consequences, this step may take a long time to identify regions of interest in the circuit. With this purpose in mind, the aim of this paper, is to propose an automatic process estimating the out-of-focus blur parameters characterized by specific attributes. Proposed technique takes advantage of extracted features in the Discrete Cosine Transform (DCT) of blurred images. The blur information is identified and allows to recover the blur's kernel. Finally, the accuracy and the robustness of the suggested process is demonstrated on real blurred images.