Colour in Context
Computer Vision Center
The segmentation method proposed in this paper is based on the observation that a single physical re°reflectance can have many different image values. We call the set of all these values a dominant colour. These variations are caused by shadows, shading and highlights and due to varying object geometry. There is a family of segmentation methods focused in the detection of these shapes: the physics-based segmentation methods, of which the dichromatic reflection model, introduced in 1985, has been the most relevant. Nonetheless, this model takes to many assumptions, and as a consequence, is not able to find this structures in a theoretical framework, but lacks of adaptability in real images. In our approach, we take the idea that this structures form well-connected ridges in the chromatic histogram.
To capture them, we propose a new Ridge based Distribution Analysis (RAD) to Find the set of ridges representative of the dominant colour. To find them we apply, first, a multilocal creaseness technique, (MLSEC-ST), which has as a main advantatge that is not affectec by little irregularities and gaps in the ridges. Afterwards, we apply a simple ridge extraction algorithm. Finally, a flooding procedure is performed to find the dominant colours in the histogram.
Qualitative results illustrate the ability of our method to obtain excellent results in the presence of shadow and highlight edges. Quantitative results obtained on the Berkeley data set show that our method outperforms state-of-the-art segmentation methods at low computational cost.