Levente Kovacs (1), Tamas Sziranyi (1,2)
(1)University of Veszprem
Department of Image Processing and Neurocomputing
(2)Hungarian Academy of Sciences
Analogical Comp. Lab., Comp. & Automation Research Institute
We present a way of rendering painting-like images and sequences by our Stochastic Painting-based SBR technique. We avoid disturbing artifacts by coding the images in a painting-like way. If we code the resulting stroke sequence, than the level of error comes not from the artifacts but from the painting process. For this reason, if we generate high quality paintings, we can transfer the image without disturbing coding errors. The painting method incorporates some novel properties like dynamic Monte Carlo Markov Chain optimization, multiscale edge gradient following or grayscale stroke templates. The painting technique inherits the properties of the Paintbrush Transformation, like well-defined contours, acceptable distortion and a painting-like view with no fine details below a limit. Our goal is to produce a painting -like output, which contains the stroke-series and the motion data in a losslessly compressed form. This way the painted video contains no compression artifacts (while the painting-like impression remains). The compression scheme of the stroke-series with motion data could also be suitable for compressing painting-like image sequences produced with other painting techniques.