HEVC-based three-layer texture and depth coding for lossless synthesis in 3D video coding

Abstract

Efficient lossless coding of a texture image and its corresponding depth map is important to perform accurate view synthesis in 3D applications. In this paper, a novel HEVC-based three-layer texture and depth coding method is proposed for lossless synthesis in 3D video coding. The proposed method performs lossy and lossless texture coding in the first and second layers, respectively. A quantization parameter (QP) in the first lossy coding layer is adaptively selected for each block, based on a relationship between the first and second layers. In the third layer, the lossy depth coding is performed by using synthesis-based depth coding and texture-based depth intra prediction mode. The synthesis-based depth coding technique adopts the lossy coding but guarantees zero synthesis distortion. The texture-based depth intra prediction mode performs the depth prediction by using the associated texture information. Experimental results demonstrate that the proposed method obtains higher coding performance than conventional lossless coding methods.

Appendix: Derivation of offset QP

To derive the offset in Eq. (6), QStep in Eq. (4) is replaced with QStep^’ defined by Eq. (3) as follows:

$$ {\displaystyle \begin{array}{l} QP=6\cdot {\log}_2\left( QSte{p}^{\prime}\right)+4\\ {}\kern0.96em =6\cdot {\log}_2\left( QSte p\cdot \frac{QSte{p}^{\prime }}{ QSte p}\right)+4\\ {}\kern1.08em =6\cdot {\log}_2\left( QSte p\right)+6\cdot {\log}_2\left(\frac{QSte{p}^{\prime }}{ QSte p}\right)+4\end{array}} $$

(11)

Eq. (11) can be represented for the (k-1)^th block as follows:

$$ 6\cdot {\log}_2\left({ QSte p}_{k-1}\right)+4={QP}_{k-1}-6\cdot {\log}_2\left(\frac{QSte{p}_{k-1}^{\prime }}{{ QSte p}_{k-1}}\right) $$

(12)

Once an optimum QP is found for a block, it can be applied to the following blocks, i.e., QSteps of the (k-1)^th can be used for the k^th block. Therefore, QStep_k-1 in the left side of Eq. (12) can be replaced with QStep_k for the optimum QP as follows:

$$ 6\cdot {\log}_2\left({ QSte p}_k\right)+4={QP}_{k-1}-6\cdot {\log}_2\left(\frac{QSte{p}_{k-1}^{\prime }}{{ QSte p}_{k-1}}\right) $$

(13)

According to the relation between QP and QStep in Eq. (4), it can be simplified as follows:

$$ {QP}_k={QP}_{k-1}-6\cdot {\log}_2\left(\frac{QSte{p}_{k-1}^{\prime }}{{ QSte p}_{k-1}}\right) $$

(14)