We propose a fast DCT domain downsampling technique for video transcoding. Experimental results show that the proposed method is not only computationally efficient, but also gives significant improvements in PSNR.