The paper discusses the developments and results of the ii-level (representations and algorithms) in the field of computer vision, motivated by the Marr paradigm. The main content of the work is divided into two parts. The first part gives a brief overview of the applied representations and previously obtained results. In particular, a description of the neuromorphic representation (model) of the registered images is given. Based on this representation the coding procedure of the receptive field system of the retina is proposed. The second part is devoted to the synthesis of the procedure for constructing elements of Marr's primary sketch based on the coded data. An algorithm proposed models the functions of simple cells in the lateral geniculate body and in visual area V1. The relationship of the chains of coded pairs ONN and OFF receptive fields with the image contrast contours is discussed. The algorithm of the computer implementation of the contour detection procedure based on the coding data from the ON- and OFF- receptive fields and their subsequent alignment based on the simplest Laplace smoothing method constitute the most important content of the main part.