An Adaptive Down-Sampling Video Coding With Hybrid Super-Resolution Method

With the rapid development of computer network, people’s demand of video application is growing fast. The size of video grows from QVGA(320x240),VGA(640x480), 720P( 1280x720), 1080P( 1920×1080) to 2K(2560×1600),4K(3840x2160). In order to adapt to the grows of video size and the network bandwidth, ITU-T and ISO/IEC which are the two famous International Standardization Organization have established a series of video coding standard including H.26x and MPEG-x. Recently, the two famous international standardization organization worked together and published the new generation video coding standard named H.265|HEVC.Despite the significant increase of video coding due to the development of video coding standard, it cannot match the growing of video size and network bandwidth,especially under limited bandwidth. There are also many applications like video surveillance and video conferencing which only take limited bandwidth, therefore we emphasize on how to keep a good video quality under low bitrate.The main contents and novelties are as follows:1. Present an adaptive down-sampling video coding method which chooses between regular coding and down-sampling adaptively. This method avoids the disadvantage of traditional down-sampling method which fixes video coding structure,thus it better fits the characteristic of video sequence itself, providing more coding efficient.2. Present a space super-resolution method based on wiener filter. During down-sampling process the high frequency part of video frame will loss which is the key of restoring down-sampled video frame. If we only restore the down-sampled frame through up-sampling, the lost high frequency cannot be recovered which results in poor restoring quality. The object of wiener filtering is original frame, therefore the coefficients of wiener filter keep some information of the high frequency part.3. Present a coarse-to-fine de-interlace method.Experimental results show that our method is better than H.265 and existing down-sampling method both in rate-distortion performance and perceived visual quality at low bitrate.