Localized image recoloring using ellipsoid boundary function

Abstract

A mechanism for recoloring a color image defines a spatial region within the image (such as the pupil of the eye) a color characteristic of which (e.g. the `red-eye` phenomenon) is to be selectively modified. To determine whether the color of an image sample within a spatial region is to be modified, the value of the chrominance component of each sample within the region is compared with a regularly shaped (e.g. elliptical) chrominance variation discriminator that surrounds a target color of interest (e.g. red, in the case of a `red-eye` correction). The major axis of the ellipse coincides with the saturation direction of the chrominance plane and its minor axis coincides with the hue direction. For each sampling location within the spatial region, the value of the chrominance component is examined to determine whether it falls within the elliptical chrominance discriminator. If the value of the chrominance component falls within the ellipse, it is modified to a destination chrominance value, based upon where the chrominance value of the sampled of interest falls, so as to soften the color at its edges. The luminance component is modified by an offset based upon the difference in the value of the luminance component of the target color and that of the new color. This luminance value difference is weighted in accordance with the product of the previously determined chrominance weighting coefficient and a prescribed relationship between the geometrical location of the luminance component for the sample of interest and the extent of a prescribed range of luminance variation projected from the elliptical discriminator along the luminance axis of the Y,I,Q coordinate system.