Digital Media Art Style Conversion Algorithm Based on Deep Neural Network

With the rapid development of digital media art, artistic style transfer technology has gradually become a hot research field. However, existing style transfer methods often face the problems of information loss and insufficient style consistency when processing complex images, which limits their effectiveness and flexibility in practical applications. In this paper, a deep neural network-based style transfer algorithm is proposed for digital media arts towards improving quality and accuracy in style transfer. Starting from a deep neural network model, a multi-level feature extraction and reconstruction strategy is adopted, with the source image and the target style image jointly trained to achieve more efficient style fusion. The perceptual loss function was introduced to optimize the quality of the generated image. It underwent rigorous testing and evaluation on multiple public datasets. The style transfer algorithm based on deep neural networks is therefore able to boost artistic expression while conserving high-quality content and style consistency. This is a creation that envisions fresh technical means and approaches for creative digital media art.