This study presents significant advancements in Four-Dimensional Scanning Transmission Electron Microscopy (4D-STEM) for analyzing strain and crystal orientation in thin films by introducing three novel methods. First, we developed an area-selective filtering technique that leverages unsupervised learning to reduce noise in 4D-STEM datasets. This approach achieved up to a 70% noise reduction for WS2-WSe2 superlattice data. Second, we introduce a strain correction method tailored for buckled two-dimensional materials. Guided by kinematical diffraction simulations, this method produces surface morphology maps that enable surface tilt and strain decoupling. Its application to MoSe2-MoS2 heterojunction data successfully reduced compressive strain measurements from an overestimated 6.5% to a more accurate ~1.5%. Lastly, we present a technique for precisely mapping crystal orientation in thin films. This technique was effectively applied to a gold nanoplate using a combination of 4D-STEM data, abTEM multislice simulations, and electron tomography validation. These advancements significantly improve the accuracy of strain measurements and crystallographic analysis, thereby enhancing our understanding of deformed nanofilms and expanding the capabilities of 4D-STEM for future materials science research.