Hung, Wei-ChiWei-ChiHungYen, Ching-ChiaChing-ChiaYenTsai, Han-JuHan-JuTsaiChan, Feng-TseFeng-TseChanLIANG-CHIA CHEN2025-10-162025-10-16202597815106935790277786Xhttps://www.scopus.com/record/display.uri?eid=2-s2.0-105015148137&origin=resultslisthttps://scholars.lib.ntu.edu.tw/handle/123456789/732669This study presents a full-field chromatic confocal microscopy (CCM) system with galvanometer scanning and CNN-based deconvolution. A programmable liquid crystal on silicon (LCoS) replaces the spectrometer slit, enabling adaptive filtering. Finite Difference Time Domain (FDTD) and Zernike-based simulations are employed to generate training data. The deep learning model enhances lateral resolution, achieving a 32-fold reduction in bias compared to scanning probe microscopy (SPM) in microbump width.falseChromatic confocal microscopy (CCM)convolutional neural network (CNN)deep learningfull-field surface profilometryoptical metrologyconference paper10.1117/12.30735992-s2.0-105015148137