Encoder-Aware Video Downscaling Using Encoding Parameters

To simultaneously produce 4K broadcast video with 8K broadcasts, 4K resolution videos can be captured using an 8K resolution video camera with an image downscaling process. A blocking artifact for the video coding of this 4K broadcast is likely to form in a pixel region with high coding difficulty. We developed the encoder-side preprocessing method of video coding for image downscaling to prevent the formation of blocking artifacts. Our proposed method adaptively controls the band limitation of the spatial frequency in an iteration using the spectrum power information of the original image and the quantization parameter (QP) and the motion vector (MV) in the video coding of its downscaled image. In addition, when iteration is not allowed, for example, for live broadcasts and low latency transmission, the spectrum power information of the original image and the estimated values of the QP and MV in the video coding of its downscaled image are used. These estimated QP and MV values were calculated previously using the video codec. We experimentally showed that coding in the proposed method is simpler than the conventional bicubic, Lanczos3, and wavelet methods to prevent the blocking of artifact formation. These experimental results using both methods are also discussed.