Partial volume correction (PVC) is needed for an accurate quantification in clinical PET. Novel PVC methods can be divided into 4 categories: Post-reconstruction region-based methods, returning PVC regional values, making them suitable for quantification but not visual analysis ; post-reconstruction image enhancement techniques, either based on deconvolution or on the combination in the wavelet domain of PET functional information with high resolution information ; in-reconstruction resolution modelling, by incorporating a resolution model in an iterative reconstruction algorithm combined with regularization ; and deep learning-based approaches, which are based on networks that can be trained either with simulated data, real data, or data corrected with another PVC method, or can be self-supervised.