To achieve low-powervideo communication in Internet of Things, this study presents a new deep learning-based fast transcoding algorithm from distributed videocoding (DVC) to high efficiency videocoding (HEVC). The ...
详细信息
Due to power consumption restrictions, low-power H.264 encoders cannot take advantage of the variable block sizes available in H.264 motion estimation. This work presents two methods to determine a block-size partitio...
详细信息
Due to power consumption restrictions, low-power H.264 encoders cannot take advantage of the variable block sizes available in H.264 motion estimation. This work presents two methods to determine a block-size partition without an initial search. With both of these methods, computationally burdensome Lagrange optimization is not required. The methods are derived from a cellular nonlinear network (CNN) segmentation algorithm and, along with the partition, indicate early termination of motion estimation and the skip modes of H.264. Both methods achieve better rate-distortion performance when compared to motion estimation with only 16 x 16 sized blocks. The 16 x 16 only case is descriptive of a low-power case where the variable block sizes cannot be used. For low bitrates, both methods achieve equivalent performance when compared to Lagrange optimization. Also presented are the computational complexity of the methods and the power consumption when implemented with existing CNN hardware. (c) 2007 Elsevier B.V. All rights reserved.
Aiming at the low-powervideo communication with low latency at resource-constrained terminals in the Internet of Things, a machine learning-based fast transcoding from distributed videocoding (DVC) to high-efficienc...
详细信息
Aiming at the low-powervideo communication with low latency at resource-constrained terminals in the Internet of Things, a machine learning-based fast transcoding from distributed videocoding (DVC) to high-efficiency videocoding (HEVC) is put forward. In order to accelerate the transcoding, the DVC decoding information is efficiently exploited to reduce HEVC encoding complexity at both levels of coding unit (CU) and prediction unit (PU). First of all, the predictions of CU partition and PU modes are regarded as two binary classification tasks. Subsequently, the initial features are extracted from the DVC decoding information and the support vector machine (SVM)-recursive feature elimination algorithm is adopted to select feature vectors to construct the training data which is used to train the SVM classifiers for CU and PU, respectively. By means of the top-down division prediction method, the CU partition is first determined by the trained SVM classifier. For the CUs which are not further split, the PU modes will be predicted to terminate the quad-tree coding process of HEVC in advance, so that HEVC encoding complexity is reduced. Experimental results show that the proposed algorithm can reduce 57.64% computational complexity on average with Bjontegaard delta bit-rate 2.43%. In terms of transcoding time efficiency, the proposed algorithm outperforms the state-of-the-art fast DVC to HEVC transcoding algorithms based on machine learning.
暂无评论