This article is devoted to the search for qualitative methodological approaches in visual cultural studies. The relevance of the analysis of the qualitative side of the image is dictated by the disproportionate trend in quantitative research (using machine vision, neural networks and big data technologies). Modern neural networks successfully cope with formal computational operations, however, they have a number of objective limitations in the qualitative analysis of images. So obvious is their potential in generating images, but analytical tools at the time of writing are practically not presented for use by researchers. At the same time, visual studies are in dire need of methodological techniques for comparing the visual and identifying the internal (semiotic, phenomenological) connection of images, taking into account the emergent nature of the media. This approach, in our opinion, has a long-term perspective of development in synthesis with quantitative methods of analysis. The article discusses the appropriateness of highlighting metonymic relationships as the basis for a technique in relation to the visual analysis of an image. This technique is specifically considered on the example of a situation of practical image analysis and, in our opinion, can be successfully implemented as a working tool for a researcher of visual culture.