A Study on the Improvement of CartoonGAN Image Quality for Nonlinear Edge Reinforcement using SRCNN Quality Enhancement Processing

With the recent development of image processing and deep learning technology, research and application in the fields of ultra-high-resolution image generation and cartoon-style conversion are expanding further.Among the ultra-resolution image generation models, the CartoonGAN model is specialized in generating cartoon-style images, but there are limitations in terms of resolution improvement.This study proposes an innovative method of additionally applying a Super-Resolution Convolutional Neural Network (SRCNN) to overcome these limitations of CartoonGAN and improve the resolution and quality of images.Through this, it is expected to expand the applicability in the art and design fields and develop the performance of deep learning-based image generation technology to the next level.The detailed structure and algorithm process of the CartoonGAN model and the SRCNN model were briefly understood, and the algorithm proposed in this study was verified through experiments based on this understanding.As a result of the experiment, it was confirmed that the highresolution image using SRCNN showed improvement in terms of quality through the values of the image quality evaluation metrics PSNR, MSE, and SSIM.The results of this study demonstrate that CartoonGAN image generation with SRCNN provides high resolution and quality compared to the existing CartoonGAN, and presents a new technical contribution to cartoon-image generation.