Artistic Image Style Transfer Based on Laplacian Pyramid Network
Shenzhi WANG, Qiu Chen · International Symposium on Affective Science and Engineering · 2023
As a very creative research direction in the field of deep learning, style transfer research has been attracted a lot of attention in the past two years, which can transform ordinary photos into artistic paintings and generate interesting pictures. Existing methods can be divided in two categories: arbitrary style transfer and single style transfer. The representative method of single style transfer and method LapStyle synthesizes stylized image in a progressive procedure by using AdaIN, an arbitrary style transfer method. Inspired by it, this paper proposes a novel arbitrary style transfer framework named LapAttN to optimize the single style transfer, which combines the advantages of LapStyle and AdaAttN. The experimental results on the COCO dataset show that the image quality of proposed method is superior to other SOTA algorithms.