Red-eye reduction using facial detection

A computer program product, tangibly encoded on a non-transitory computer-readable medium, operable to cause data processing apparatus to perform operations comprising to

• receive an image depicting a first eye and a second eye corresponding to a human face, and coordinates corresponding to a location of the first eye and the second eye in the image,
• calculate a distance between the first eye and the second eye using the received coordinates,
• obtain a skin tone sample from the image based on the calculated distance and the received coordinates,
• generate a skin tone color region in a color space based on the obtained skin tone sample,
• identify a red-eye pixel from the image associated with the first eye,
• modify the red-eye pixel based on the generated skin tone color region, wherein the operation to modify further comprises operation to:
• identify red-eye pixels from the image associated with the first eye, and
• ascertain a region of the image to modify based on a center of mass of the red-eye pixels and an evaluation, using the generated skin tone color region and an eye-white color region, of pixels from the image at increasing distances from the center of mass, and
• store the image including the modified first eye.

A system for reducing the occurrence of red-eye artifacts comprising:

• a storage device operable to store software instructions;
• a processor operable to interact with the storage device, to retrieve and execute software instructions embodying operations comprising:
• receiving an image depicting a first eye and a second eye corresponding to a human face, and coordinates of the first eye and the second eye in the image;
• calculating a distance between the first eye and the second eye using the received coordinates;
• obtaining a skin tone sample from a rectangular region above and below the first eye, relative to the face, the rectangular region having a size based on the calculated distance;
• generating a skin tone color region in a color space based on the obtained skin tone sample;
• deriving a classification of a pixel corresponding to the first eye based on the generated skin tone color region and a predetermined red-eye color region, wherein the operation of deriving a classification comprises:
• identifying red-eye pixels corresponding to the first eye by comparing pixels in the rectangular region with the generated skin tone color region and the predetermined red-eye color region;
• ascertaining a center of mass of the red-eye pixels;
• determining an iris region for the first eye by evaluating pixels corresponding to the first eye at increasing distances from the center of mass based on the generated skin tone color region and an eye-white color region;
• modifying the pixel based on the derived classification;
• and
• storing the modified first eye using the storage device.