A Content-Aware Quantisation Mechanism for Transform Domain Distributed Video Coding

The discrete cosine transform (DCT) is widely applied in modern codecs to remove spatial redundancies, with the resulting DCT coefficients being quantised to achieve compression as well as bit-rate control. In distributed video coding (DVC) architectures like DISCOVER, DCT coefficient quantisation is traditionally performed using predetermined quantisation matrices (QM), which means the compression is heavily dependent on the sequence being coded. This makes bit-rate control challenging, with the situation exacerbated in the coding of high resolution sequences due to QM scarcity and the non-uniform bit-rate gaps between them. This paper introduces a novel content-aware quantisation (CAQ) mechanism to overcome the limitations of existing quantisation methods in transform domain DVC. CAQ creates a framespecific QM to reduce quantisation errors by analysing the distribution of DCT coefficients. In contrast to the predetermined QM that is applicable to only 4x4 block sizes, CAQ produces QM for larger block sizes to enhance compression at higher resolutions. This provides superior bitrate control and better output quality by seeking to fully exploit the available bandwidth, which is especially beneficial in bandwidth constrained scenarios. In addition, CAQ generates superior perceptual results by innovatively applying different weightings to the DCT coefficients to reflect the human visual system. Experimental results corroborate that CAQ both quantitatively and qualitatively provides enhanced output quality in bandwidth limited scenarios, by consistently utilising over 90% of available bandwidth. Keywords—distributed video coding; Wyner-Ziv video coding; content-aware quantisation; adaptive quantisation; bit