Based on the user-specified parameters, the vector image is rescaled and divided into smaller same-size rectangular cells for matching. The image-matching phase represents the crux of our project, requiring a lot of careful consideration. Observations show that curvy fonts retain their shape better after the thinning process, which tends to produce better results. As part of our experimentation, we have explored a range of fonts, including simple and minimalistic options like JetBrains Mono and JuliaMono, as well as more intricate and curvy fonts like Courier, CutiveMono, and SimSun. Any of the popular fonts in everyday life will work as long as they come in the form of Scalable Vector Graphics. The algorithm performs optimally with monospace fonts due to their fixed aspect ratio (character’s height / character’s width). The banquet in raster format before processing the banquet in lines after processingĪ set of font images for the 95 ASCII characters is necessary to generate our ASCII art. The monk in raster format before processing the monk in lines after processing This aligns the stroke width of the pictures with the characters, ensuring a proper match. Before this stage, the images are thinned to achieve a uniform stroke width of one pixel. The raster image processing phase presents some challenges as we had to balance collecting reasonable curves and maintaining as many details from the input as possible. This transformation minimizes distortion during the subsequent process of redrawing the images to pixel art according to size. Most of our inputs come in the form of raster graphics, which represent images as grids of tiny square pixels. Although raster graphics is a very popular format, it has a significant drawback when it comes to retaining the shape of the images during resizing. Therefore, the initial task involves converting the inputs into vector images, which are made of points, lines, and curves based on mathematical equations.

This project involves the implementation of the paper “Structure-based ASCII Art” by Xuemiao Xu, Linling Zhang, and Tien-Tsin Wong, published in ACM Transactions on Graphics (SIGGRAPH 2010 issue), Vol. Additionally, we will discuss the challenges encountered and our ongoing efforts to optimize the algorithm.

Throughout this blog post, we will explore the various stages involved in this project, including processing pixel images, preparing font images, and the crucial image-matching phase. Users can upload their images and specify the parameters, including the font, font size, and the number of characters for the output’s width, to get the generated ASCII art. ASCII art can be divided into two types: tone-based, which decides the output characters based on the density of the image, and structure-based, which determines the result by matching characters with the outline shape of the image.Īs one of our two research projects this summer, we are working on the implementation of structure-based ASCII art generation. The input picture can be anything from a complex image to simple emoticons. Ilinkin’s lab! I am Ha, a rising sophomore, and this summer I am working on a project that involves creating ASCII art.ĪSCII art is a form of digital art that uses ASCII characters to produce images.

Easy ascii art ten lines