Combination of linear and gradient vector flow field digital mosaic rendering for automated tile placement

Mosaic is a composite image generated by combining sets of small images together through the use of tiles. These tiles tessellate images with the purpose of recreating it in a mosaic form. The traditional method of generating a mosaic requires a design to be intuitively obtained from nature, occurrences or paintings, and a working surface be prepared, before tiles are glued to the surface. The process is time consuming and labour intensive. In recent times, mosaic are being generated digitally through the use of computational tools. The linear mode of rendering mosaic digitally, has been in use for a very long time. Its advantage include better tile coverage area due to lesser space between tiles, but it does not represent curved surfaces properly. This makes the mosaic have an unpleasant physical appearance. This led to the rendering of mosaic in a non-linear form.Gradient Vector Flow (GVF) field algorithm has been employed to tackle the limitations of linear mosaic rendering, because of its advantages of better similarity to a source image, better preservation of fine details, better distribution of grout, and better preservation of an image’s global appearance. However, GVF has a problem of not enough tile coverage area. So an improved hybrid algorithm which combines the linear, and non-linear GVF digital mosaic rendering via image segmentation is proposed, to improve the overall tile coverage area and maintain a good similarity to the source image. There are three main contributions providedin this thesis. Firstly, an improved algorithm using hybrid GVF digital mosaic rendering. Secondly, this thesis evaluates the tile coverage area, and source image similarity of the output digital mosaic obtained from, the hybrid, linear and non-linear GVF digital mosaic rendering, using the percentage tile coverage area and Structural Similarity (SSIM) index respectively. The results show that, the hybrid has a better percentage tile coverage area than the non-linear, but lesser than the linear. Also, the hybrid and non-linear has the best SSIM. Thirdly, the thesis presents the physical tiling of the hybrid, linear and non-linear GVF digital mosaic using a robot. This study can be applied in the field of arts to replicate images