An Adaptive Convolution Neural Network-Based Multimedia Generative Adversarial Learning Framework for Style Transfer of Movie Videos
Mengfan He, Xiaoxiao Pang, Yangbo Li · Journal of Circuits Systems and Computers · 2025
Style transfer of movie videos is an important multimedia processing task that can be significant to post-production and visual effect design of movies. The existing methods generally suffer from poor migration performance and high computational complexity when dealing with such tasks. Thus, this paper proposes an adaptive convolutional neural network (CNN)-based multimedia generative adversarial learning framework for this purpose. First, we design an adaptive CNN structure that aims at learning advanced semantic features of videos and being. This is expected to adapt to specific styles of input videos. Second, we introduce generative adversarial networks (GANs) to achieve style transfer between the input video and the target style. The introduction of GANs can more accurately transfer video styles while effectively controlling the balance between image quality and content preservation. After that, we conducted some experiments on a real-world image dataset for performance evaluation. The results show that the proposal has achieved significant improvement in the style transfer tasks of movie videos. Higher transfer performance and lower computational complexity can be achieved by the proposal compared to existing methods.