The CCLM (cross-component linear model) method in the current VTM3 requires complex process to derive the linear model parameter and based on the straight line equation with reference pixel comparison step. Considering that the derivation process of the linear model parameters is also required in decoder side, it is necessary to reduce the required complexity. In this proposal, the number of reference sample pairs, which are used in the CCLM parameter determination is reduced to half based on pre-defined upper limit so that the worst pixel comparison complexity is reduced by half (i.e., from 4 samples to 2 samples).