The invention relates to video image compression technologies, and discloses a method and system for upsampling a spatial scalable coded video image so that during upsampling computation complexity may be reduced while coding performance is substantially unchanged. In the invention, the principle that human eyes are far less sensitive to a chrominance components than to a luminance components is utilized, and a simpler filter is adopted for the chrominance components than that for the luminance components during upsampling in I_BLINTRA_Base inter-layer prediction or residual samples image inter-layer prediction, thereby reducing effectively calculation complexity while coding performance is substantially unchanged.