A proposed invention relates to the field of information technologies and can be used in information technologies, which relates to computer vision, for comparison and search of similar images and their fragments by the content in large databases/pictures banks. The claimed method for completely automatic detection of similar images and images fragments in large data storages or data banks comprising image content processing, stages of construction to represent numerical vector image content. Each image is processed in the data store to determine the content that is treated as geometric shapes, and numerical vector representing the image built based on the geometric parameters of defined geometric shapes. A numeric vector which consists of geometric parameters of image figures, measured at original size of geometric shape, and after normalization, is the vector that defines the image with accuracy and uniqueness, and which allows comparison and detection of similar images.