Human society is characterized by the living together of various cultures, and as a result of globalization, cities are increasingly being built similarly. How can this be avoided? urban culture memes in a multicultural setting are important for the growth of a city. The majority of earlier evaluations of urban style relied on straightforward classification tasks. Without taking into account the most important visual components of cities as a whole in order to achieve the visual aspects of cities. In order to do this, we use the dictionary learning method to extract visual social media twitter and memes from the image data of ten representative urban areas from around the world. We then use memetic similarity to quantify the symmetric similarities and differences between cities and sparse representation to interpret the causes of these similarities and differences. The experimental findings show that visual memes and media thread from twitter have some restrictions across cities, i.e., the components of urban style are very similar while the logical combinations of visual memes change greatly, which accounts for the variations in urban style across cities.