Multiple-Frame Quality Compensation Encoding Method Based on H.266 / VVC

This research focuses on the multi-frame quality compensation coding method based on H.266/VVC. By leveraging technologies such as optical flow algorithms and convolutional neural networks, the video sequence passes through a preprocessing module to ensure its consistency. The multi-frame quality compensation network contains sub-modules like multi-frame quality compensation, original frame guidance, and rate-distortion decision-making, and is driven by a composite loss function that combines pixel mean square error and perceptual loss. 500 diverse videos were collected and 15,000 frames were extracted to construct a dataset. Based on the H.266/VVC standard, the encoding parameters were precisely adjusted. With the industry-standard H.265/HEVC as the baseline for comparison, experiments were carried out on a high-performance GPU workstation equipped with an NVIDIA GeForce RTX 3090 and 32GB of memory, using the PyTorch deep learning framework. The results show that this method has advantages in both image quality and coding efficiency. Ablation experiments quantified the effectiveness of each module. Although there are challenges in handling ultra-high-resolution videos, it is expected to promote the development of the next-generation video coding standard and empower fields such as multimedia communication.