Beschreibung |
Image completion is one of typical tasks in the field of image processing. Among others, algorithms based on matrix completion are popular nowadays for such tasks. A particular advantage of such algorithms is well-established tools for convergence analysis. As the result, stable completion of images can be achieved. However, a natural limitation of classical matrix completion algorithms is the restriction to grey images, since matrices of real numbers can carry only information associated to one real number, i.e. intensity of grey colour. Therefore, the goal of this project is to extend the existing matrix completion algorithms to quaternionic matrices. In this case, completion of coloured images can be done, since a pure quaternion carries information about colours in RGB code. The project will start with analysis of existing algorithms and their implementation in MATLAB. |