Research on Improved Image Style Transfer Algorithms Based on Deep Learning

This paper addresses common issues in deep learning-based image style transfer algorithms, such as high-frequency noise, low optimization efficiency, and susceptibility to local optima. We propose a multi-resolution optimization style transfer. The method adopts a progressive optimization strategy from low to high resolution, significantly enhancing optimization efficiency while capturing multi-scale style features. Experimental results demonstrate that compared to the original algorithm, the multi-resolution method offers advantages in computational efficiency and style transfer performance. This study provides effective algorithm choices for various application scenarios and offers new perspectives for further advancement of style transfer methods.