The Art of Scale-Space. J. Andrew Bangham, Stuart Gibson and Richard Harvey School of Information Systems, University of East Anglia, Norwich, NR4 7TJ, UK. Email:{a.bangham@uea.ac.uk} Abstract Artists pictures rarely have photo-realistic detail. Tools to create pictures from digital photographs might, therefore, include methods for removing detail. These tools such as Gaussian and anisotropic diffusion filters and connected-set morphological filters (sieves) remove detail whilst maintain- ing scale-space causality, in other words new detail is not created using these operators. Non-photorealistic rendering is, therefore, a potential application of these vision techniques. It is shown that certain scale-space filters preserve the appropriate edges of retained segments of interest. The resulting images have fewer extrema and are perceptually simpler than the original. A second artistic goal is to accentuate the centre of attention by reducing detail away from the centre. The process also removes the detail providing perceptual cues about photographic texture. This allows the ‘eye’ to readily accept al- ternative, artistic, textures introduced to further create an artistic impression. Moreover, the edges bounding segments accurately represent shapes in the original image and so provide a starting point for sketches. 1 Introduction A photographer tends to choose uncluttered backgrounds and make careful use of focus to direct attention. Of course, lens blurring is both easy and effective for it exploits the natu- ral and powerful way in which the brain rejects non-foveated regions of a scene (they are simply out-of-focus). The technique finds its way into rendering, digital art, advertising, and video through the Gaussian blur filter widely used to de-focus background material. But the method is rarely used by painters. Rather, they direct attention by selecting detail and manipulating textures and geometry. By contrast, a painter starts with a blank canvas, adds paint and the more skilled knows when to stop. It is the progressive addition of detail that characterizes the process of pro- ducing representational art in which only some detail directly represents that in the orig- inal scene. It difficult to capture representational detail manually from three-dimensional (3D) scenes onto two-dimensional (2D) canvases, but this does not satisfactorily explain why trained artists limit the amount of detail they use. After all two dimensional, pho- tographic quality, images have been traced for over five centuries by those projecting images onto surfaces using concave mirrors and lens [10]). But the evidence from the resulting pictures suggests that artists pick only those details that resonate with their artis- tic interpretation. They choose to ignore some objects and lots of detail. Painting is not photography. BMVC 2003 doi:10.5244/C.17.58