Digital media image enhancement and real-time rendering algorithms for virtual reality scenes

Focusing on the declining image quality and the constrained rendering performance in virtual reality (VR) system, this article proposes a collaborative optimization algorithm of digital media image enhancement and real-time rendering for VR scenes. By constructing a deep learning model based on U-Net structure and incorporating attention mechanism, the degraded images such as blurred and low illumination are successfully enhanced. The average peak signal-to-noise ratio (PSNR) is 29.7 dB, the structural similarity index (SSIM) is 0.823, and the perceived image quality evaluation (LPIPS) is reduced to 0.194, which is obviously superior to the traditional methods. At the same time, a dynamic rendering strategy combining Asynchronous Time Warping (ATW) and Space-Time Warping (STW) is designed to achieve an average frame rate of 92 FPS on the HTC Vive Pro 2 device, and the frame delay is controlled within 8.3 ms. Experiments show that the proposed method can improve the image quality, alleviate the performance bottleneck caused by high-resolution rendering to some extent, and take into account the image quality and interactive fluency.