Abstract: Spatial scalable video coding technique becomes a hot problem in recent years because this kind of multi-resolution coding technique can satisfy with the requirement of online video transmitting in low bit-rate and various terminal devices displaying.In this paper, we propose a scalable video coding of arbitrary resolution in DCT domain based on convolution.Firstly, we estimate each macro-block′s coefficients based on its sub-block coefficients referring to the relationship between the sub-block coefficients and the macro-block coefficients.Secondly, in order to reduce the original video resolution, a filter is designed to get the low frequency components of the estimated macro-block.Finally, these low frequency components can be saved or transmitted as one part of the current video coding stream, and a low computing complexity and an arbitrary on-demanded resolution video stream can be obtained, it is own to different size of low-pass filter and a scheme which can get macro-block DCT coefficients directly.Experimental results demonstrate that our scheme can provide flexible spatial scalable ability, and it can be used to process all kinds of the DCT-based video encoders including AVS and H.264, etc.