Countless fragments of existing architectural photography are merged into multilayered shapes. The resulting collages introduce a third abstract point of view next to the original ones of architect and photographer.
Digital scans of analogue architectural photography form tiny pieces of a large resulting puzzle. The original pictures are being analysed and categorised according to their vanishing-points and shapes. Based on this analysis, slices are being extracted from the source image. These slices retain the information of their position corresponding to their original vanishing-point and thus form a large pool of pieces, ready to be applied to new perspectives and shapes.
Using the extracted image segments, it is now possible to form collages of originally different pictures with a new common perspective. In order to compose a collage, a perspective-grid is defined and a lining of matching image segments is being applied. The segments are not altered to match the frame but fitting ones are chosen from the sheer mass of possible pieces. By defining additional keywords which describe the content of the original photographs, the selection of segments used for the final composition can be influenced. Thus a contextual layer is added through the semantic linking with the source material.
The recompositions mix and match the views and perspectives of both the architect and the photographer with a third, newly chosen frame. The resulting fine-art prints are entirely unique each time.
Looking at hundreds of architecture photographs we noticed a strong bias regarding image composition. Perspective foreshortening and vanishing lines dominate the overall impression of the image. This realisation lead to experiments in automated extraction of said features in the image data.
The medium used for the result – for now a printed image – is just one of the many possibilities. The method also bears the potential for further experimentation and can be considered a work in progress.
|building:||Radio Bremen||source images by Klaus Frahm|