This contribution proposes two modifications to convolutional cross-component model (CCCM) based chroma prediction with an intention to simplify the implementation. The first modification removes an offset from reference luma and chroma samples. Reference sample values at the top-left corner of the PU are used as the offsets for simplicity. Using such offset is asserted to reduce the dynamic range of data and the bitdepth required for model derivation. The second modification replaces the division operation in CCCM by a piece-wise polynomial (power of 2) function.