| Nowadays,various electronic products come into view,and their display screens have different requirements for video quality and resolution.In this context,the scalable video coding technology designed for multi-resolution video transmission has become a hotspot.On the one hand,in the process of video storage or transmission,in order to make the encoded video meet certain quality requirements,it needs to be encoded several times to obtain an appropriate bitrate,which takes much time and calculation.On the other hand,in the statistical multiplexing of TV broadcasting,bitrate allocation is used to improve the channel utilization rate.Existing bitrate allocation methods are mainly based on complex rate-distortion theory,which can obtain accurate bitrates,however,the high accuracy is usually excessive.For these two reasons,this paper proposes a quasi-quantized bitrate estimation model for scalable video coding in multi-resolution video transmission systems.Through this lightweight model,an appropriate coding bitrate is obtained,which can save the computational cost.The main work of this paper includes the following three parts.(1)A detailed experimental comparison of the coding efficiency of HE VC simulcast and SHVC is carried out.It is worth noting that the coding efficiency of HEVC is higher than that of SHVC in some sequences.A hypothesis is proposed that SHVC are more efficient than HEVC simulcasting in encoding videos of high motion.The further designed frame extraction coding experiments confirms this hypothesis.(2)A video dataset is constructed for scalable video coding bitrate estimation.Based on the big data of audience rating,various of video programs are selected from mainstream media,and 50 video sequences are obtained.The bitrate data under the objective video quality evaluation is obtained by SHVC,which lays a data foundation for the training and testing of the scalable video coding bitrate estimation model.(3)Based on the multiple linear regression model,a scalable video coding bitrate estimation model is proposed.Some features are extracted from the spatial and temporal information of the video content,which are used to compose the explanatory variables of the model.After correlation analysis,stepwise regression analysis and multicollinearity detection,the final explanatory variables that have a significant impact on the bitrate are filtered.The obtained scalable video coding bitrate estimation model has a fitting degree of 79.7%,and achieves an ideal result in random cross-validation,which has certain practical value. |
