Representing a region of a color image using a space-color separable

Abstract

A representation of a color image region is called a space-color separable (SCS) model because the spatial structure of the image region is represented separately from that of the color-space structure. The representation uses two functions, one to represent the region's spatial structure and another to represent its color-space structure. In one implementation, the spatial structure is represented by polynomials of arbitrary degree to map image locations to the parameter values of the color-space model which in turn maps these parameter values to a zero-, one-, two-, or three-dimensional linear subspace of color space. The representation is resolution-independent and can accommodate both transformations of image coordinates and transformation of color coordinates. In particular, the polynomial-linear implementation can easily accommodate both affine transformations of image coordinates and linear transformations of color coordinates. This flexibility makes it useful for a variety of applications in illustration and image editing systems. The invention combines several features: representing smooth shading or textures in color images; separating the spatial structure of the colors in the image from its color space structure; and enabling transformations of images and their colors by transforming the SCS model itself.