Fast Content Adaptive Representation Selection in HEVC-Based Video Coding for Streaming Applications

Preparing the video content for adaptive video streaming and ensuring the best possible quality of experience (QoE) to end-users while minimizing the server memory consumption and needed computational power presents a challenging task. This paper presents a novel solution and notation of a model for the content-adaptive estimation of the bitrate for a certain spatial resolution for encoding with the High Efficiency Video Coding (HEVC) encoder. Compared to available solutions, the idea was to simplify the process of determining the bitrate ladder that takes into account the video content. Contrary to other solutions, the proposed solution does not require precoding or high computational power, thus it minimizes latencies so it can be used for real-time applications. The proposed model requires only the values of spatial and temporal information to determine the optimal bitrate ladder for a certain video sequence. The proposed solution consists of a representation selection model and a bitrate ladder model. The representation selection model is used to select the optimal bitrates for switching to higher spatial resolution. Pearson Linear Correlation Coefficient (PLCC) between the values estimated by the model and the measured values calculated for all video sequences used in the development of the model is 0.963 and for the test sequence is 0.975, confirming the accuracy of the model. In addition, the bitrate ladder model was developed to determine the bitrates for the desired quality levels. For the test sequence, the proposed model achieves a PLCC value of 0.9175, and the mean difference between estimated and measured values is 1.826 %.