New Approaches to Artistic Synthesis Relying on Visual Perception

Artists often take advantage of the limitations of the human visual system to create art that provides new experiences for the observers. Similarly, in this thesis, we explore new computational artistic compositions that create new visual experiences by relying on the limitations of the human visual system. We focus on two characteristics of the human visual system: One is the ambiguity that arises due to the differences between the eye projection space and the real physical space, and the second is the property of the integration of temporal information.

For the projection-space ambiguity, we propose two frameworks that enable two different artistic representations. First, we show how to create high reliefs that represent full 3D forms in a very limited space. The main challenges are to preserve the relationships of scene parts and the fine details of the shape. Second, we develop an approach for creating sculpture paintings where paintings and sculptures are blended within the same compositions. The main challenge is to ensure continuity between the 2D and 3D parts of the scene geometry. They create plausible effects for the viewer by showing a new relationship of 2D and 3D parts at each different angle.

For the temporal-space ambiguity, we propose a new stream of frames called -tempocode- that encodes the hidden information in video frames that show artistic dither matrices. In order to achieve this, we develop a visual-masking method that hides an input image, both spatially and temporally. Our masking function creates temporal and spatial variations on the frequency bands of the original input signal, with the constraint that the integrations of the final signal and the original signal are the same. The hidden information cannot be perceived by the human eye but it can be easily decoded by a long exposure photograph with a camera.