The contribution proposes a method for implicitly partitioning CTUs or CUs spanning across a picture boundary. Specifically, for picture dimensions that are not an integer multiple of the CTU size, CTUs at the right and/or bottom of the picture will contain pixel regions that do not correspond to the intended, reconstructed image. In the proposed method, these CTUs undergo implicit partitioning that makes use of the available square and rectangular partitioning types until all the samples at the leaf of the partitioning tree created using implicit splits do not span across the picture boundary. The implicit splits can include both quad tree and binary tree splits. Moreover, the quad-tree splitting can be further controlled by receiving a minimum number of required quad-tree splits of the CTU within a parameter set. Experimental results show that the proposed method achieves a coding efficiency improvement of 0.20% and 0.25% relative to the VTM and BMS, respectively, using CTC random access test conditions. This includes a gain of 0.42% and 0.54%, respectively, for the Class B test sequences. In all cases, the impact on encoder run-time is measured to be in the range of 102%-108%.