Red Eye False Positive Filtering Using Face Location and Orientation

Abstract

An image is acquired including a red eye defect and non red eye defect regions having a red color. An initial segmentation of candidate redeye regions is performed. A location and orientation of one or more faces within the image are determined. The candidate redeye regions are analyzed based on the determined location and orientation of the one or more faces to determine a probability that each redeye region appears at a position of an eye. Any confirmed redeye regions having at least a certain threshold probability of being a false positive are removed as candidate redeye defect regions. The remaining redeye defect regions are corrected and a red eye corrected image is generated.

Description

BACKGROUND

Redeye is the appearance of an unnatural reddish coloration of the pupils of a person appearing in an image captured by a camera with flash illumination. Redeye is caused by light from the flash reflecting off blood vessels in the person's retina and returning to the camera.

A large number of image processing techniques have been proposed to detect and correct redeye in color images. In general, these techniques typically are semi-automatic or automatic. Semi-automatic redeye detection techniques rely on human input. For example, in some semi-automatic redeye reduction systems, a user manually identifies to the system the areas of an image containing redeye before the defects are corrected.

Many automatic redeye reduction systems rely on a preliminary face detection step before redeye areas are detected. A common automatic approach involves detecting faces in an image and, subsequently, detecting eyes within each detected face. After the eyes are located, redeye is identified based on shape, coloration, and brightness of image areas corresponding to the detected eye locations. In general, face-detection-based automatic redeye reduction techniques have high computation and memory resource requirements. In addition, most of the face detection algorithms are only able to detect faces that are oriented in an upright frontal view; these approaches cannot detect faces that are rotated in-plane or out-of-plane with respect to the image plane.

U.S. Pat. No. 6,407,777 to DeLuca discloses in-camera detection and correction of redeye pixels in an acquired digital image. U.S. Pat. No. 6,873,743 to Steinberg discloses automated real-time detection and correction of redeye defects optimized for handheld devices. US published patent applications 2005/0047655 and 2005/0047656 to Luo et al. disclose detecting and correcting redeye in a digital image and in embedded systems, respectively.

Automatic red eye detection algorithms can sometimes wrongly identify an image region as red eye artefact. Those regions are called False Positives (FP). Although somewhat rare, applying red eye correction to them can result in a visually displeasing image. It is desired to have a technique that reduces the number of false positives by using face location and/or orientation information.

SUMMARY OF THE INVENTION

A method of detecting and correcting a red-eye defect within a digital image is provided. An image is acquired having one or more non red eye defect regions which are red in color. A first stage initial segmentation of candidate redeye regions is performed to determine a first set of one or more confirmed redeye regions designated for correction. A location and orientation of any faces within the image are determined. The first set of confirmed redeye regions is analyzed based on the determined location and orientation of faces to determine a probability that each confirmed redeye region appears at a position of an eye. Any confirmed redeye regions having at least a certain threshold probability of being a false positive are removed from the first set, and a second set is generated. The second set of confirmed red eye regions is corrected and a red eye corrected image is generated which has the second set of confirmed red eye regions corrected therein. The redeye corrected image is electronically stored, transmitted, further processed or edited, or displayed, or combinations thereof.

The performing of the first stage initial segmentation of red eye regions may include pixel analyzing. The performing of the first stage initial segmentation of red eye regions may include falsing and verification filtering. The analyzing and removing may be performed prior to any correcting of the image. The first set of confirmed redeye regions may be initially corrected, and an initial corrected image may be generated prior to the analyzing and removing and the generating of the red eye corrected image. One or more detected faces may include at least one red eye defect such that the second set comprises a non-empty set, or the second set may be empty such that no actual redeye regions are corrected in the image.

An embedded image acquisition and processing system includes an image acquisition subsystem. A red eye filter performs in a first stage an initial segmentation of candidate redeye regions detected within an acquired image to determine a first set of one or more confirmed redeye regions designated for correction. A face location and orientation detector and an analysis filter determine a probability that each confirmed redeye region appears at a position of an eye based on determining face location and orientation information. A processor corrects any red eye defects of the confirmed red eye regions of the first set minus any having at least a certain threshold probability of being a false positive and generates a red eye corrected image. The red eye corrected image is electronically stored, transmitted, further processed or edited, or displayed, or combinations thereof. No redeye defects are corrected when no faces are detected within the image.

One or more storage devices having processor-readable code embodied therein are also provided for programming one or more processors to perform any of the methods described herein alone or in combination with techniques described in references incorporated herein.

BRIEF DESCRIPTION OF THE DRAWINGS

Statement Regarding Color Drawings

The patent or application file contains at least one drawing that is executed in color, including color photographs. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

FIG. 1A is a block diagram illustrating a first red eye correction method.

FIG. 1B is a block diagram illustrating a second red eye correction method.

FIG. 1C is a block diagram illustrating a third red eye correction method.

FIG. 1D is a block diagram illustrating a fourth red eye correction method.

FIGS. 2A-2B is an image of a person including red eye defects and regions having red color that are not red eye defect regions.

FIGS. 3A-3B is the image of FIG. 2 with the red eye defects and the other regions corrected for color.

FIGS. 4A-4B illustrates the face of the person in FIGS. 2 and 3 being detected and located and having its orientation determined.

FIGS. 5A-5B is the image of FIG. 2 or FIG. 3 with only the actual red eye defects corrected.

DETAILED DESCRIPTION OF THE EMBODIMENTS

A redeye filter process is illustrated in FIG. 1A. An input image is first analyzed by a redeye detection stage 100 at a pixel level 103 and segmented into candidate redeye regions 104. A further series of falsing and verification filters 106 are then applied to the candidate regions and a set of confirmed redeye regions 108 is thereby determined. A correction filter (pixel modifier) 102 is next applied to the confirmed regions and a final image 112, corrected for redeye, is generated.

In embodiments herein, a face detection algorithm is applied either as a component of an advantageous falsing and verification filter 106, or after the set of confirmed red eye regions is initially determined at 108. This face detection algorithm determines location and/or orientation of the face. This provides information as to where specifically that any eyes probably exist and/or where specifically that eyes probably do not exist within the image.

FIG. 1B illustrates a process including analyzing an input image by a red eye detection stage 100 at a pixel level 103 and segmenting into candidate red eye regions 104, as well as applying falsing and verification filters 116 to the candidate regions for determining a confirmed set of red eye regions 108. In this embodiment, the falsing and verification filter 116 including a face detection process that determines location and/or orientation of the face that provides information as to where specifically that any eyes probably exist and/or where specifically that eyes probably do not exist within the image.

FIG. 1C illustrates a process including analyzing an input image by a red eye detection stage 100 at a pixel level 103 and segmenting into candidate red eye regions 104, as well as applying falsing and verification filters 106 to the candidate regions for determining a confirmed set of red eye regions 108. In this embodiment, a face detection process 118 determines location and/or orientation of the face and provides information as to where specifically that any eyes probably exist and/or where specifically that eyes probably do not exist within the image. The face detection process 118 for determining false positives is performed upon confirmation of red eye regions at 108, whereas in FIG. 1D, the face detection process 118 for determining false positives is performed after confirmation of red eye regions 108 and also after correction of the confirmed red eye regions 102.

Referring to FIGS. 2A-2B, a woman wearing red earrings is photographed and red eye artefact is present within the image in both of her eyes. The earrings that she is wearing are similar in color to the red color appearing as photographic artefact in her eyes. While conventional red eye correction algorithms would adjust the color of the eyes as well as the earrings, the present invention serves to distinguish the actual red eye artefact appearing in the woman's eyes from her red earrings. It does this by determining that the woman's eyes probably are at the location of the real red eye artefact based on determining the location and/or orientation of the woman's face, while also determining that the woman's earrings are at a location of a false positive, because the earrings are at respective locations within the digital image relative to her facial position and/or orientation that do not correspond to where her eyes would probably be, based on the determined location and/or orientation of her face.

In the image provided at FIGS. 3A-3B, one can see that an initial red eye auto-detection algorithm correctly found the two actual red-eye regions and corrected them with regard to their color. However, the algorithm also mistakenly found the two earrings and further corrected each of them.

FIGS. 4A-4B illustrate the advantageous use of a face detection algorithm that determines location and orientation of the woman's face. The box shown in FIGS. 4A-4B encloses substantially the woman's face. The red side of the box indicates the top of her head indicating a determination of facial orientation. The algorithm figures this out based on location of features of the woman's face such as eyes, nose, mouth cheeks, chin, eyebrows, hairline, and/or other features of the woman's head or other body parts such as her neck, shoulders, torso and/or legs either appearing or not appearing in certain directions relative to the face within the digital image, and/or the shape of the head, and/or the appearance of ears, among other possible factors (see, e.g., US applications US2006/0204110, PCT/US2006/021393, US2002/0172419 to Lin et al; US2002/0126893 to Held et al US2005/0232490 to Itagaki et al. and US 2004/0037460 to Luo et al., each incorporated by reference).

In the example of FIGS. 4A-4B, one of the earrings is within the face detection box (i.e., the right earring), while the other earring is outside the face detection box (i.e., the left earring). The red left earring is deemed to have lower probability of being red eye artefact, because it is outside the box, than features such as the woman's eyes and the right earring, because these are inside the box. In addition, both of the woman's red earrings are located near sides of the box, while her eyes are detected within the box at a location that is more probably where her eyes would be than at the sides of the box. Thus, both earrings now have reduced probability of being red eye artefact than the woman's red eyes, possibly with the left earring having the lower probability of the two earrings. Using orientation detection, the red eye artefact within each of the woman's eyes is determined to be at a location that is probably where her eyes would be and thus are deemed to be red eye artefact to be corrected, while the red earrings are determined to be at a location that is probably not where her eyes would be and thus deemed not to be red eye artefact.

The red side of the box that indicates the “top” of the woman's head or face, or the “up” side of her face may be obtained by applying the face detection algorithm on the image as indicated, and also by using face tracking which uses multiple preview images or other reference images such as postview images or images captured simultaneously with the main image of FIGS. 2A-2B.

In the described example, the position of every detected red eye region is analyzed in respect to the face location. Two correct red eyes are found to be positioned inside the face region and in agreement with the orientation and are therefore considered red eye candidates. The left earring region is found to be inside the face box but, because of orientation mismatch, is classified as a false positive and filtered out. The right earring region is found to be outside the face box, but very close to the face regions. In this case it is considered that the likelihood of another face being present is minimal and therefore the region is classified as a false positive and removed. If a second face is detected within the image, then the red eye process can proceed with respect to that face, and any red eye artefact can be removed in that second process. Other possible red eye regions that are located far enough from the face are unaffected and therefore remain red eye candidates.

FIGS. 5A-5B illustrate the corrected image of the woman having actual red eye artefact corrected and the false positive earrings returned to their initially captured red color. In the examples of FIGS. 2A-5B, the horns being worn by the woman, perhaps at some Holiday party, were not initially detected as red eye artefact. If they were initially deemed to be red eye artefact, then the same process as with the earrings would have identified these as false positives as well.

While an exemplary drawings and specific embodiments of the present invention have been described and illustrated, it is to be understood that that the scope of the present invention is not to be limited to the particular embodiments discussed. Thus, the embodiments shall be regarded as illustrative rather than restrictive, and it should be understood that variations may be made in those embodiments by workers skilled in the arts without departing from the scope of the present invention as set forth in the appended claims, and structural and functional equivalents thereof.

In addition, in methods that may be performed according to preferred embodiments herein and that may have been described above, the operations have been described in selected typographical sequences. However, the sequences have been selected and so ordered for typographical convenience and are not intended to imply any particular order for performing the operations, except for those where a particular order may be expressly set forth or where those of ordinary skill in the art may deem a particular order to be necessary.

In addition, all references cited herein as well as the background, invention summary, abstract and brief description of the drawings, as well as U.S. Pat. Nos. 6,407,777, 7,315,631 and 7,336,821, and US published patent applications nos. 2005/0041121, 2005/0031224, 2005/0140801, 2006/0204110, 2006/0093212, 2006/0120599, 2007/0110305 and 2006/0140455, and PCT/EP2006/008358, and U.S. patent application Nos. 60/773,714, 60/804,546, 60/865,375, 60/865,622, 60/829,127, 60/829,127, 60/821,165, 60/892,882, 60/945,558, 60/915,669 10/772,767, 11/554,539, 11/464,083, 11/462,035, 11/282,954, 11/027,001, 10/842,244, 11/024,046, 11/233,513, 11/753,098, 11/753,397, 11/766,674, 11/769,206, 11/772,427 and 11/460,218, are all incorporated by reference into the detailed description of the preferred embodiments as disclosing alternative embodiments.

Claims

1. A method of detecting and correcting a red-eye defect within a digital image, comprising: acquiring an image including one or more non red eye defect regions having a red color;performing in a first stage an initial segmentation of candidate redeye regions to determine a first set of one or more confirmed redeye regions designated for correction;analyzing pixels in a vicinity of at least one of the one or more confirmed red eye regions of the first set;analyzing the first set of confirmed redeye regions based on the analyzing of said pixels in said vicinity of said at least one of said one or more confirmed red eye regions of said first set;removing from the first set any confirmed redeye regions having at least a certain threshold probability of being a false positive based on said analyzing of said first set of confirmed red eye regions, and thereby generating a second set;correcting the second set of confirmed red eye regions and generating a red eye corrected image which has the second set of confirmed red eye regions corrected therein; andelectronically storing, transmitting, further processing or editing, or displaying the red eye corrected image, or combinations thereof.

2. The method of claim 1, wherein the performing of the first stage initial segmentation of red eye regions comprises pixel analyzing.

3. The method of claim 2, wherein the performing of the first stage initial segmentation of red eye regions comprises falsing and verification filtering.

4. The method of claim 1, wherein the analyzing and removing are performed prior to any correcting of the image.

5. The method of claim 1, further comprising initially correcting the first set of confirmed redeye regions and generating an initial corrected image prior to the analyzing and removing and the generating of said red eye corrected image.

6. The method of claim 1, wherein the one or more faces further include at least one red eye defect such that the second set comprises a non-empty set.

7. The method of claim 1, wherein the second set comprises an empty set such that no actual redeye regions are corrected in the image.

8. An embedded image acquisition and processing system, comprising. (a) an image acquisition subsystem;(b) a red eye filter that performs in a first stage an initial segmentation of candidate redeye regions detected within an acquired image to determine a first set of one or more confirmed redeye regions designated for correction;(c) a pixel analyzer configured to analyze pixels in a vicinity of at least one of the one or more confirmed red eye regions of the first set;(d) an analysis filter that determines a probability that each confirmed redeye region appears at a position of an eye based on results of applying the pixel analyzer to the at least one of the one or more confirmed red eye regions of the first set;(e) a processor for correcting the red eye defects of the confirmed red eye regions of the first set minus any having at least a certain threshold probability of being a false positive based on said results of applying said pixel analyzer and generating a red eye corrected image; and(f) wherein the red eye corrected image is electronically stored, transmitted, further processed or edited, or displayed, or combinations thereof.

9. The system of claim 8, wherein the performing of the first stage initial segmentation of red eye regions comprises pixel analyzing.

10. The system of claim 9, wherein the performing of the first stage initial segmentation of red eye regions comprises falsing and verification filtering.

11. The system of claim 8, wherein the analyzing and removing are performed prior to any correcting of the image.

12. The system of claim 8, wherein the processor further for initially correcting the first set of confirmed redeye regions and generating an initial corrected image prior to the analyzing and removing and the generating of said red eye corrected image.

13. The system of claim 8, wherein the system is configured such that no redeye defects are corrected when no faces are detected within the image.

14. One or more storage devices having processor-readable code embodied therein for programming one or more processors to perform a method of detecting and correcting a red-eye defect within an acquired digital image that includes one or more non red eye defect regions having a red color, the method comprising: performing in a first stage an initial segmentation of candidate redeye regions to determine a first set of one or more confirmed redeye regions designated for correction;analyzing pixels in a vicinity of at least one of the one or more confirmed red eye regions of the first set;analyzing the first set of confirmed redeye regions based on the analyzing of said pixels in said vicinity of said at least one of said one or more confirmed red eye regions of said first set;removing from the first set any confirmed redeye regions having at least a certain threshold probability of being a false positive based on said analyzing of said first set of confirmed red eye regions, and thereby generating a second set;correcting the second set of confirmed red eye regions and generating a red eye corrected image which has the second set of confirmed red eye regions corrected therein; andelectronically storing, transmitting, further processing or editing, or displaying the red eye corrected image, or combinations thereof

15. The one or more storage devices of claim 14, wherein the performing of the first stage initial segmentation of red eye regions comprises pixel analyzing.

16. The one or more storage devices of claim 15, wherein the performing of the first stage initial segmentation of red eye regions comprises falsing and verification filtering.

17. The one or more storage devices of claim 14, wherein the analyzing and removing are performed prior to any correcting of the image.

18. The one or more storage devices of claim 14, further comprising initially correcting the first set of confirmed redeye regions and generating an initial corrected image prior to the analyzing and removing and the generating of said red eye corrected image.

19. The one or more storage devices of claim 14, wherein the one or more faces further include at least one red eye defect such that the second set comprises a non-empty set.

20. The one or more storage devices of claim 14, wherein the second set comprises an empty set such that no actual redeye regions are corrected in the image.