Detecting And Correcting Peteye

Abstract

Peteye is the appearance of an unnatural coloration (not necessarily red) of the pupils in an animal appearing in an image captured by a camera with flash illumination. Systems and methods of detecting and correcting peteye are described. In one aspect a classification map segmenting pixels in the input image into peteye pixels and non-peteye pixels is generated based on a respective segmentation condition on values of the pixels. Candidate peteye pixel areas are identified in the classification map. The generating and the identifying processes are repeated with the respective condition replaced by a different respective segmentation condition on the pixel values.

1 - 22 . (canceled) 
     
     
         23 . A method of processing an input image, comprising:
 segmenting pixels in the input image into an animal-fur color class and a non-animal-fur color class;   identifying in the input image candidate peteye pixel areas corresponding to respective clusters of pixels in the non-animal-fur color class;   selecting ones of the identified candidate peteye pixel areas as detected peteye pixel areas; and   recoloring ones of the pixels in the detected peteye pixel areas.   
     
     
         24 . The method of  claim 23 , wherein the segmenting comprises segmenting into the animal-fur color class pixels having values corresponding to one of multiple specified gray levels, a specified brown color range, and a specified flesh color range, wherein one of the gray levels is white. 
     
     
         25 . The method of  claim 24 , further comprising segmenting into a second animal-fur color class pixels having values corresponding to one of the multiple specified gray levels excluding white, the specified brown color range, and the specified flesh color range. 
     
     
         26 . The method of  claim 23 , wherein the segmenting comprises segmenting into the animal-fur color class pixels having a chroma value below a chroma level threshold. 
     
     
         27 . The method of  claim 23 , wherein the segmenting comprises segmenting into the animal-fur color class pixels having a chroma value below a chroma level threshold, a luminance value below a luminance threshold, and a hue value outside a hue range defined by a lower hue threshold and an upper hue threshold. 
     
     
         28 . The method of  claim 23 , wherein the segmenting comprises segmenting into the animal-fur color class pixels having a chroma value below a chroma level threshold, a luminance value above a luminance threshold, and a hue value within a hue range defined by a lower hue threshold and an upper hue threshold. 
     
     
         29 . The method of  claim 23 , further comprising classifying pixels in the detected peteye areas selected during the given iteration into one of a peteye pixel class and a non-peteye pixel class, and desaturating ones of the input image pixels classified in the peteye pixel class. 
     
     
         30 . A method of processing an input image, comprising:
 segmenting pixels in the input image into peteye pixels and non-peteye pixels based on a mapping of the input image pixels into a one-dimensional luminance space;   identifying in the input image candidate peteye pixel areas based on the segmented peteye pixels;   selecting ones of the identified candidate peteye pixel areas as detected peteye pixel areas; and   recoloring ones of the pixels in the detected peteye pixel areas.