Cornsweet surfaces for selective contrast enhancement

• A novel image parametrisation approach based on B-spline surfaces.
• Selective countershading guided by image-content and artistic input allowing for strong contrast enhancements that better respect the edge structure.
• Psychophysical experimentation evaluating the visual quality of our results.

A typical goal when enhancing the contrast of images is to increase the perceived contrast without altering the original feel of the image. Such contrast enhancement can be achieved by modelling Cornsweet profiles into the image. We demonstrate that previous methods aiming to model Cornsweet profiles for contrast enhancement, often employing the unsharp mask operator, are not robust to image content. To achieve robustness, we propose a fundamentally different vector-centric approach with Cornsweet surfaces. Cornsweet surfaces are parametrised 3D surfaces (2D in space, 1D in luminance enhancement) that are extruded or depressed in the luminance dimension to create countershading that respects image structure. In contrast to previous methods, our method is robust against the topology of the edges to be enhanced and the relative luminance across those edges. In user trials, our solution was significantly preferred over the most related contrast enhancement method.

Figure optionsDownload high-quality image (86 K)Download as PowerPoint slide