Fast Artistic Style Transfer via Wavelet Transforms
Xinwei Zhang, Xiaoyun Chen · 2023
Artistic style transfer aims to transfer artistic styles to the content images. Although artistic style transfer has received great attention in recent years, most existing style transfer methods are still limited in practical applications due to the slow transfer speed, such as difficulty in efficiently processing images with large scales (e.g., 1024x1024 pixels). To address this issue, we propose a fast artistic style transfer method based on wavelet transforms. Specifically, we utilize the wavelet transform to compress the content image and extract the low frequency component of the content image, which is an approximation of the content image and a quarter the size of the content image. We only stylize the low frequency component of the content image, which can effectively accelerate the style transfer. Experimental results demonstrate that our method can generate high quality stylized images with high efficiency.