It has recently been demonstrated that spatial resolution adaptation can be integrated within video compression to improve overall coding performance by spatially down-sampling before encoding and superresolving at the decoder. Signiﬁcant improvements have been reported when convolutional neural networks (CNNs) were used to perform the resolution upsampling. However, this approach suﬀers from high complexity at the decoder due to the employment of CNN-based super-resolution. In this paper, a novel framework is proposed which supports the ﬂexible allocation of complexity between the encoder and decoder. This approach employs a CNN model for video downsampling at the encoder and uses a Lanczos3 ﬁlter to reconstruct full resolution at the decoder. The proposed method was integrated into the HEVC HM 16.20 software and evaluated on JVET UHD test sequences using the All Intra conﬁguration. The experimental results demonstrate the potential of the proposed approach, with signiﬁcant bitrate savings (more than 10%) over the original HEVC HM, coupled with reduced computational complexity at both encoder (29%) and decoder (10%).
|Title of host publication||Video compression with low complexity CNN-based spatial resolution adaptation|
|Publication status||Accepted/In press - 20 Apr 2020|
|Event||Applications of Digital Image Processing XLIII - |
Duration: 24 Aug 2020 → 28 Aug 2020
|Conference||Applications of Digital Image Processing XLIII|
|Period||24/08/20 → 28/08/20|