A three-dimensional reconstruction and visual optimization method for theatrical styling based on an improved generative adversarial network

The digitalization of theatrical styling refers to the process of converting physical theatrical elements, including costumes, makeup, and props, into high-fidelity 3D digital models. Traditional methods suffer from low efficiency in reconstructing high-fidelity 3D models, poor real-time performance in capturing complex theatrical styling features, and obvious step-shading errors in 3D reconstruction, which cannot meet the practical demands of efficient digital creation and visual optimization in theatrical design and production. This article presents a 3D reconstruction and visual optimization method for theatrical styling based on an improved generative adversarial network (GAN). The method incorporates a hybrid quantum compression classical GAN model with a matrix product operator, enhancing both the efficiency and quality of data generation by optimizing the quantum generator structure. Additionally, the model accelerates convergence and reduces the number of parameters by integrating classical network characteristics. The improved GAN architecture eliminates the batch normalization layer, reducing computational complexity, while the new 3D reconstruction algorithm and its parallax map optimization method significantly improve reconstruction accuracy and algorithmic efficiency. Experimental results demonstrate that the proposed hybrid quantum compression classical GAN model achieves stable loss function convergence at 2,800 and 4,500 rounds for generating the numbers 0 and 3, respectively. Furthermore, on the Flickr2K dataset, the Peak Signal-to-Noise Ratio (PSNR) of the proposed 3D reconstruction and visual optimization model reaches 27.142, and the Structural Similarity Index (SSIM) reaches 0.835, outperforming state-of-the-art models including Bicubic, Fast Super-Resolution Convolutional Neural Network (FSRCNN), Super-Resolution Generative Adversarial Network (SR-GAN), and Enhanced Super-Resolution Generative Adversarial Network (ESRGAN), indicating higher reconstruction accuracy and better structural similarity preservation.