Method of automatic identification of ruler scales on images to solve the problem of estimating the geometric dimensions of certain objects

Detection of pixels to real-world unit correspondence on the image is an important step for the further analysis of the geometrical characteristics of the objects on the image. In this paper, I propose an iterative method of automatic identification of the ruler scales on the image based on digit recognition. Experiments showed a mean absolute error of 6.63 px and median absolute error of 4 px on high-resolution images $(3104\mathrm{x}1746\ \text{px})$ compared to the manual measurement results. I focus on the medical domain images, but this method can be used in many others domains like forensic, museum business, etc.