In current VVC, the picture width and height have to be multiple of 4 (the smallest CU size). However, when a picture with resolution (8M+4) x (8N+4) is input to the VTM-4.0, the encoder crashes because ALF only supports 4x4 SIMD filtering and (8M+4) x (8N+4) picture results in 2xN/Nx2 chroma region which triggers the encoder crash. It could be fixed easily by supporting 2xN/Nx2 ALF filtering. Another issue is that the merge sharing node of the 4x4 block located at the bottom-right corner of a picture may have region located outside the picture and it may result in undefined operations for MVP derivation. To solve the issue, it is proposed that the picture width and picture height are not necessarily equal to the minimum CU width/height. Instead, the picture width and height are both constrained to be multiples of 8 which is aligned with HEVC; alternatively, to provide more flexibility, the picture width is constrained to be multiples of 8 and picture height is constrained to be multiples of 4. Another solution is that when a boundary of a merge sharing node is located outside a picture, applying a truncation to the boundary to move the boundary of the merge sharing node to be aligned with a picture boundary. The proposed modifications do not bring impact to the coding performance under CTC.