This article provides a comparative analysis of existing texture synthesis methods. The advantages and disadvantages of the methods are revealed. In our work, we were guided by the results of extensive expert testing, which was attended by more than 25 people on 20 fundamentally different textures. This allowed us to draw conclusions about what types of textures which methods should be used. Our study showed that recently popular neural network and statistical methods are neither the best in terms of synthesis quality nor speed. However, they also generate more diverse textures. On the other hand, the easiest and fastest method for rearranging patches showed the best quality and speed. Thus, pattern synthesis of textures is one of the promising areas of research in which different approaches have different advantages.
Babichev Andrei Yur'evich, andrey.babichev@graphics.cs.msu.ru, кафедра интеллектуальных информационных технологий факультета вычислительной математики и кибернетики Московского государственного университета
