Non-separable mode dependent transforms for intra coding in HEVC

Abstract

Transform coding plays a crucial role in video coders. Recently, additional transforms based on the DST and the DCT have been included in the latest video coding standard, HEVC. Those transforms were introduced after a thoroughly analysis of the video signal properties. In this paper, we design additional transforms by using an alternative learning approach. The appropriateness of the design over the classical KLT learning is also shown. Subsequently, the additional designed transforms are applied to the latest HEVC scheme. Results show that coding performance is improved compared to the standard. Additional results show that the coding performance can be significantly further improved by using non-separable transforms. Bit-rate reductions in the range of 2% over HEVC are achieved with those proposed transforms.

Mode-dependent transform competition for HEVC

“…The results from previous work in intra coding using modedependent transforms [5] and transform competition [6] independently encourage the combination of both, named modedependent transform competition (MDTC). In order to validate the approach of using multiple transforms per IP mode, a first experiment has been carried out, where one additional transform is used to complement the HEVC core transforms.…”

“…Sparse orthogonal transforms, or rate-distortion optimised transforms (RDOTs), were introduced in [7] by Sezer et al and are explained more in detail in [8]. The RDOTs provide better compression gains than the KLT in the context of video coding as reported in [5].…”

“…In this paper, the MDDT concept is extended through the usage of a set of transforms adapted to each IP mode: for a given prediction mode, the encoder selects the best transform in the provided set. The design of multiple transforms and their implementation into HEVC have already been presented in our previous work [5,6], following two different approaches:…”

“…• In [5], it is shown that a transform design method called rate-distortion optimised transform (RDOT) yields to better coding efficiency compared to the traditional KLT design. This method is used to design transforms adapted for each IP mode as in the MDDT system.…”

Transform Coding in the VVC Standard

“…More specifically, in [20], an Enhanced Multiple Transform (EMT) scheme was proposed and included in the Joint Exploration Model, which applies intra prediction mode dependent transform sets with each set consist of multiple candidates of transform kernels. Moreover, multiple non-separable transform schemes have also been proposed in [22] [23]. In [24], a row-column transform scheme is proposed, which approximates non-separable KLT using a separable transform.…”