Film Effect Optimization by Deep Learning and Virtual Reality Technology in New Media Environment

Area of Science:

• Computer Vision
• Artificial Intelligence
• Immersive Media

Background:

• New media technologies have transformed visual expression in film and television.
• Existing virtual reality (VR) films face challenges with immersion, interaction, and simulation fidelity.

Purpose of the Study:

• To optimize film playing effects in a new media environment using deep learning and VR technologies.
• To address limitations in immersion, interaction, and simulation for VR films.

Main Methods:

• Optimized extremum median filter algorithm for image processing (reducing "burr" and low compression).
• Generative Adversarial Network (GAN) for single video image data enhancement.
• Decision tree and hierarchical clustering algorithms for VR image color enhancement.

Main Results:

• Optimized single-frame images show improved contrast (4.21), entropy (8.66), and reduced noise (145.1).
• The proposed GAN improved image quality and diversity compared to supervised GANs, aligning with human subjective evaluations.
• Significant improvement in Frechet Inception Distance (FID) compared to Self-Attention Generative Adversarial Network, indicating precise details and rich textures.

Conclusions:

• The developed system effectively enhances image contrast and reduces dazzling intensity, preserving image details.
• The advanced GAN model generates high-quality, diverse images with superior detail and texture.
• This research offers a valuable framework for enhancing the interactivity, immersion, and simulation of VR films.