Image Color Cube Dimensional Filtering and Visualization

We call RGB cube the three dimensional representation of the colors of a digital image in RGB space. It has been observed that 2D structures predominate in the color cube. They are qualitatively explained by two main causes: image blur and texture. In this article a filtering algorithm, the Local Linear Projection (LLP2), is used to smooth the 3D color distribution and to reveal the underlying 2D structures of the color clusters. Source Code The source code (ANSI C), its documentation, and the online demo are accessible at the IPOL web page of this article 1 .

[1]  Jitendra Malik,et al.  A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics , 2001, Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001.

[2]  Jean-Michel Morel,et al.  Scale Space Meshing of Raw Data Point Sets , 2011, Comput. Graph. Forum.

[3]  X. Huo,et al.  LOCAL LINEAR PROJECTION (LLP) , 2002 .

[4]  David Levin,et al.  The approximation power of moving least-squares , 1998, Math. Comput..

[5]  Jean-Michel Morel,et al.  Dimensionality of color space in natural images. , 2011, Journal of the Optical Society of America. A, Optics, image science, and vision.