Complexity reduction for 3D-HEVC depth maps intra-frame prediction using simplified edge detector algorithm

Depth modeling modes complexity control system for the 3D-HEVC video encoder

“…In our previous work [12], we demonstrated that the highest difference of the four corner samples of a block (called Dmax) contains significant information to classify this block as homogeneous or edge. Motivated by this analysis, the Simplified Edge Detector (SED), whose flowchart is presented in Fig.…”

“…It is notable that if SED uses a fixed threshold as proposed in [12], then a variable number of DMMs should be skipped for different types of videos since each video has a different quantity of borders and details as exemplified in Fig. 3.…”

“…Otherwise, the encoding block is classified as a homogeneous region and the DMMs evaluations are skipped. The SED classification proposed in [12] is based on a fixed threshold value according to the video resolution and block size. The thresholds were defined in an offline statistical analysis regarding the resolutions 1024×768 and 1920×1080.…”

“…Saldanha et al [11] propose to compute the encoding block border gradient and specifically reduce the complexity of DMM-1 algorithm by only evaluating wedgelets starting in a high gradient value. In our previous work [12], a Simplified Edge Detector (SED) algorithm is proposed to skip DMM evaluations when the encoding block is classified as homogeneous. This classification is performed by analyzing the highest difference of the four corners samples of the block.…”

“…Therefore, this paper presents the first complexity control system for the 3D-HEVC standard that targets only the DMMs processing. In this proposed solution, the SED algorithm [12] is applied to decide for an encoding depth block if DMM evaluations should be skipped according to a target computational complexity.…”