Improved Algorithm for Intra-frame Prediction Based on AVS3
Minfeng Chen · 2022
uavs3e is a video encoder based on the AVS3 coding standard. This paper focuses on improving the process of traversing multiple modes at each depth based on the cumbersome block partitioning of uavs3e, first counts the division depths on each LCU (Largest Coding Unit) boundary, and then calculates the image texture change information to derive the threshold value of the texture change degree with high correlation between the division depths on the intersection boundary of two LCUs. And a fast algorithm is proposed to terminate the division depth in advance according to the image information. Experimental results show that using the above optimisation algorithm can save an average 16.24% of coding time while paying only an average 0.53% of performance loss (PSNR-Y) (peak signal to noise ratio-Luminance component).