Application for raster image vectorisation

The aim of the project was to implement an application for detecting edges in raster images and vectorising them. The result of the processing is an SVG file containing a set of “path” objects corresponding to the individual edges and described by third-order (cubic) Bézier curves.
The image processing parameters can be adjusted by the user individually for each image in order to obtain satisfactory results (e.g. the most accurate possible representation of objects, or the preparation of simplified graphics for further processing in a graphics editor).
The edge detection process is based on binarisation of the input image and extraction of the points describing the edges of the resulting shapes. Distorted edges are smoothed. For vectorisation, the edges are divided into sections without corners (continuity of the first two derivatives), which are then split into segments approximated by cubic Bézier curves or straight lines.
The application was developed in Java using the OpenCV library (image processing) and SVGSalamander (SVG graphics handling).