Face detection in mid-shot digital images

Abstract

A method for detecting a face in a mid-shot digital image of a person comprises capturing first and second mid-shot digital images of nominally the same scene using different capture settings such that the foreground is differently differentiated from the background in each image, and comparing the first and second images to determine the foreground region of the images. A portion of the foreground region likely to correspond to a face is estimated based upon the geometry of the foreground region.

Description

The present invention relates to a method and system for detecting a face in a digital image, and in particular a method and apparatus for detecting a face in a mid-shot digital image of a person. In this context a mid-shot image of a person is an image having a single human figure in the foreground orientated in a generally upright position.

BACKGROUND OF THE INVENTION

Known face tracking applications for digital image acquisition devices include methods of marking human faces in a series of images such as a video stream or a camera preview. Face tracking can be used to indicate to a photographer the locations of faces in an image or to allow post processing of the images based on knowledge of the locations of the faces. Also, face tracker applications can be used in adaptive adjustment of acquisition parameters of an image, such as, focus, exposure and white balance, based on face information in order to produce improved the quality of acquired images.

A well-known method of fast-face detection is disclosed in US 2002/0102024, hereinafter Viola-Jones. In Viola-Jones, a chain (cascade) of 32 classifiers based on rectangular (and increasingly refined) Haar features are used with an integral image, derived from an acquired image, by applying the classifiers to a sub-window within the integral image. For a complete analysis of an acquired image, this sub-window is shifted incrementally across the integral image until the entire image has been covered.

A number of variants of the original Viola-Jones algorithm are known in the literature, such as disclosed in U.S. patent application Ser. No. 11/464,083 (FN143). However, such face detection applications are computationally expensive.

It is an object of the present invention to provide an alternative and computationally efficient method of face detection in mid-shot digital images.

DISCLOSURE OF THE INVENTION

The present invention provides a method for detecting a face in a mid-shot digital image of a person as claimed in claim 1.

The invention is based upon the recognition that, for mid-shot digital images, a simple geometric analysis of the foreground can locate the face to a high degree of accuracy, thereby dispensing with the need for complex calculations.

If desired, the presence of a face can be confirmed or denied by, for example, looking for a preponderance of flesh tones within the portion of the foreground identified by the inventive method and presumed to include a face, but this is still far less computationally intensive that the prior art techniques.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will now be described by way of example with reference to the accompanying drawings, in which:

FIG. 1 is a block diagram of a digital image acquisition device operating in accordance with an embodiment of the present invention;

FIG. 2 is a flow diagram of face detection software in the image acquisition device of FIG. 1; and

FIG. 3 shows images processed according to two alternate image analysis algorithms which may be used in the geometrical analysis step of FIG. 2.

DESCRIPTION OF THE PREFERRED EMBODIMENT

FIG. 1 is a block diagram of a digital image acquisition device 20 which in the present embodiment is a portable digital camera, and includes a processor 120. It can be appreciated that many of the processes implemented in the digital camera may be implemented in or controlled by software operating in a microprocessor, central processing unit, controller, digital signal processor and/or an application specific integrated circuit, collectively depicted as processor 120. Generically, all user interface and control of peripheral components such as buttons and display is controlled by a microcontroller 122. The processor 120, in response to a user input at 122, such as half pressing a shutter button (pre-capture mode 32), initiates and controls the digital photographic process. Ambient light exposure is monitored using light sensor 40 in order to automatically determine if a flash is to be used. A distance to the subject is determined using a focus component 50 which also focuses the image on image capture component 60. If a flash is to be used, processor 120 causes the flash 70 to generate a photographic flash in substantial coincidence with the recording of the image by image capture component 60 upon full depression of the shutter button. The image capture component 60 digitally records the image in colour. The image capture component preferably includes a CCD (charge coupled device) or CMOS to facilitate digital recording. The flash may be selectively generated either in response to the light sensor 40 or a manual input 72 from the user of the image acquisition device. The high resolution image recorded by image capture component 60 is stored in an image store 80 which may comprise computer memory such a dynamic random access memory or a non-volatile memory. The camera is equipped with a display 100, such as an LCD, for preview and post-view of images.

In the case of preview images which are generated in the pre-capture mode 32 with the shutter button half-pressed, the display 100 can assist the user in composing the image, as well as being used to determine focusing and exposure. Temporary storage 82 is used to store one or more of the preview images and can be part of the image store 80 or a separate component. The preview image is preferably generated by the image capture component 60. For speed and memory efficiency reasons, preview images preferably have a lower pixel resolution than the main image taken when the shutter button is fully depressed, and are generated by sub-sampling a raw captured image using sub-sampler software 124 which can be part of the general processor 120 or dedicated hardware or combination thereof. Depending on the settings of this hardware subsystem, the pre-acquisition image processing may satisfy some predetermined test criteria prior to storing a preview image. Such test criteria may be chronological, such as to constantly replace the previous saved preview image with a new captured preview image every 0.5 seconds during the pre-capture mode 32, until the high resolution main image is captured by full depression of the shutter button. More sophisticated criteria may involve analysis of the preview image content, for example, testing the image for changes, before deciding whether the new preview image should replace a previously saved image. Other criteria may be based on image analysis such as the sharpness, or metadata analysis such as the exposure condition, whether a flash is going to happen, and/or the distance to the subject.

If test criteria are not met, the camera continues by capturing the next preview image without saving the current one. The process continues until the final high resolution main image is acquired and saved by fully depressing the shutter button.

Where multiple preview images can be saved, a new preview image will be placed on a chronological First In First Out (FIFO) stack, until the user takes the final picture. The reason for storing multiple preview images is that the last preview image, or any single preview image, may not be the best reference image for comparison with the final high resolution image in, for example, a red-eye correction process or, in the present embodiment, mid-shot mode processing. By storing multiple images, a better reference image can be achieved, and a closer alignment between the preview and the final captured image can be achieved in an alignment stage discussed later.

The camera is also able to capture and store in the temporary storage 82 one or more low resolution post-view images. Post-view images are low resolution images essentially the same as preview images, except that they occur after the main high resolution image is captured.

The image acquisition device 20 has a user-selectable mid-shot mode 30. In mid-shot mode, when the shutter button is depressed the camera is caused to automatically capture and store a series of images at close intervals so that the images are nominally of the same scene. A mid-shot mode face detecting processor 90 analyzes and processes the stored images according to a workflow to be described. The processor 90 can be integral to the image acquisition device 20—indeed, it could be the processor 120 with suitable programming—or part of an external processing device 10 such as a desktop computer. As will be described, the particular number, resolution and sequence of images, whether flash is used or not, and whether the images are in or out of focus, depends upon the particular embodiment. However, in this embodiment the processor 90 receives a main high resolution image from the image store 80 as well as a low resolution post-view image from the temporary storage 82.

Where the mid-shot mode face detecting processor 90 is integral to the image acquisition device 20, the final processed image may be displayed on image display 100, saved on a persistent storage 112 which can be internal or a removable storage such as CF card, SD card or the like, or downloaded to another device, such as a personal computer, server or printer via image output means 110 which can be connected via wire, fiber, or other data transmission means including wireless means. In embodiments where the processor 90 is implemented in an external device 10, such as a desktop computer, the final processed image may be returned to the image acquisition device 20 for storage and display as described above, or stored and displayed externally of the camera.

FIG. 2 shows the workflow of a first embodiment of mid-shot mode processing according to the invention.

First, mid-shot mode is selected, step 200. Now, when the shutter button is fully depressed, the camera automatically captures and stores two digital images:

• • a main, high pixel resolution, flash image (image A), step 202.
  • a post-view, low pixel resolution, non-flash image (image B), step 204.

The post-view image B is captured immediately after the main image A, so that the scene captured by each image is nominally the same. If desired image A could be non-flash and image B taken with flash. The important thing, for this embodiment, is that one of them is taken with flash and one without. Normally, for a mid-shot image of a person, the main image A would be the flash image but this will depend on other lighting. An example of a mid-shot image A is shown in FIG. 3(a)—the post-view image B will be substantially the same but of lower resolution.

Steps 200 to 204 just described necessarily take place in the image acquisition device 20. The remaining steps now to be described are performed by the mid-shot processor 90 and can take place in the camera or in an external device 10.

Images A and B are aligned in step 206, to compensate for any slight movement in the subject or camera between taking these images. Techniques for aligning images in this way are well-known. Then, step 208, the images A and B are matched in pixel resolution by up-sampling image B and/or down-sampling image A. Again, this is well-known in the art.

Next, step 210, the flash and non-flash images A and B are used to construct a foreground map, step 210. A foreground map is a set of data defining those regions of the aligned images which belong to the foreground of the images. FIG. 3(b) represents the foreground map for the image of FIG. 3(a), although it is to be understood that the map is not necessarily produced as a visible image. The foreground map locates the foreground subject within the boundaries of the overall image.

In this embodiment, steps 206 to 210 may be carried out in accordance with the method disclosed in U.S. patent application Ser. No. 11/217,788 and PCT Application No. PCT/EP2006/005109 (Ref: FN122), which is hereby incorporated by reference.

Finally, step 212, the portion of the foreground region likely to correspond to a face is identified by analysis of the size and shape of the foreground region. It will be appreciated that such a simple geometric approach to face detection can be used where the approximate size and shape of the subject is known in advance, as is the case for a mid-shot of a single human figure. Two algorithms for detecting the face region will now be described, with reference to FIGS. 3(c) and 3(d).

First, and common to both algorithms, the orientation of the foreground subject in the image relative to the camera is determined, as disclosed in International Patent Application No. PCT/EP2006/008229 (Ref: FN119), which is hereby incorporated by reference. This method is based on the observation that in a normally orientated camera for a normally orientated scene, the close image foreground, in this case, the subject, is at the bottom of the image and the far background is at the top of the image. Alternatively, the orientation of the subject in the image may be ascertained using motion sensors as is well known in the art.

In the first algorithm, FIG. 3(c), the width of the body is estimated using a band of N pixels in the lower (with respect to orientation) part of the image and bounding it to the left and the right. Then this rectangle is grown upwardly until it reaches the shoulders (rectangle 300). The line of the shoulders is determined by computing line-by-line the ratio of foreground and background pixels. When the ratio of background pixels reaches a certain threshold and does so for a number of consecutive lines, the first line encountered is taken to be the one with the shoulders. From this rectangle 300 the position and size of the face area is estimated (rectangle 302) and verified to be fully covered by foreground.

In the second algorithm, FIG. 3(d), the rectangle 304 bounding the foreground is calculated. Then, using the orientation information, the top portion 306 of the bounding rectangle 304 is selected, where the head is assumed to be. For this purpose the general position of the head is computed from a head/body ratio, loosened such that it will contain the head no matter what position it is in (e.g. straight, bent, and so on). The top of the rectangle 306 is coincident with the top of the rectangle 304 and extends down ⅜ of the height of the latter. It is also narrower, being ¾ the width of the rectangle 304 and centred within it.

Now the top rectangle 306 is reduced in width to include only the face. First, the bounding box of the foreground in the previously found rectangle 306 is computed by shrinking rectangle 306 until it contains only foreground pixels. This bounding box may contain hands or false positives from the background which are filtered by selecting the largest rectangle 308 that is the right shape/size and made only of foreground. More particularly, the height of rectangle 308 is computed using body proportions. The face height is estimated to be 2/7th of the height of rectangle 304. The vertical displacement between 308 and 306 is the presumed forehead height.

Variations of the foregoing embodiment are possible. For example, image B could be a preview image rather than a post-view image. Alternatively, both images A and B could be low resolution pre- and/or post-view images, and the foreground map derived therefrom used to identify the face region in a third, high resolution main image. In such a case all three images, i.e. images A and B and the main image, will need to be nominally the same scene. In another embodiment image B could have the same high resolution as the main image A. This would avoid the need to match image resolution at step 208.

In a further embodiment, where the use of flash for one of the images is not desirable, foreground/background separation may be carried out in accordance with the method disclosed in PCT Application No. PCT/EP2006/008229 (Ref: FN119). In this case, the main image A is taken with the foreground more in focus than the background and the other image B is taken with the background more in focus than the foreground. Using the focused and non-focused images, foreground/background separation is carried out to identify the mid-shot subject.

The invention is not limited to the embodiment(s) described herein but can be amended or modified without departing from the scope of the present invention.

Claims

1. A method for detecting a face in a mid-shot digital image of a person, the method comprising: using a lens, image sensor and processor of a digital image acquisition and processing apparatus,capturing first and second mid-shot digital images of approximately a same scene that is known to include at least a face and a portion of a torso of a human person, including using different capture settings such that the foreground is differently differentiated from the background in each image,comparing the first and second images to determine the foreground region of the images, andidentifying a portion of the foreground region likely to correspond to a face based upon matching one or more known face-to-torso geometric relationships to one or more geometric relationships between regions of the foreground region of the first and second images.

2. The method of claim 1, wherein the different capture settings comprise taking one image using a flash and taking the other image without using a flash.

3. The method of claim 1, wherein the different capture settings comprise taking one image with the foreground more in focus than the background and taking the other image with the background more in focus than the foreground.

4. The method of claim 1, wherein the first and second images have different pixel resolutions.

5. The method of claim 4, further comprising matching the pixel resolutions of the first and second images prior to the comparing.

6. The method of claim 1, wherein the first and second images are captured by a digital camera.

7. The method of claim 6, wherein the first image is a relatively high resolution main image, and wherein the second image is a relatively low resolution pre- or post-view version of the first image.

8. The method of claim 6, wherein the first and second images are relatively low resolution pre- and/or post-view versions of a higher resolution main image of said scene also captured by the camera.

9. A digital image acquisition and processing apparatus comprising: means for capturing first and second mid-shot digital images of approximately a same scene that is known to include at least a face and a portion of a torso of a human person, including using different capture settings such that the foreground is differently differentiated from the background in each image,means for comparing the first and second images to determine the foreground region of the images, andmeans for identifying a portion of the foreground region likely to correspond to a face based upon matching one or more known face-to-torso geometric relationships to one or more geometric relationships between regions of the foreground region of the first and second images.

10. One or more non-transitory processor-readable media having code embedded therein for programming a processor to perform a method for detecting a face in a mid-shot digital image of a person, wherein the method comprises: using a lens, image sensor and processor of a digital image acquisition and processing apparatus,capturing first and second mid-shot digital images of approximately a same scene that is known to include at least a face and a portion of a torso of a human person, including using different capture settings such that the foreground is differently differentiated from the background in each image,comparing the first and second images to determine the foreground region of the images, andidentifying a portion of the foreground region likely to correspond to a face based upon matching one or more known face-to-torso geometric relationships to one or more geometric relationships between regions of the foreground region of the first and second images.

11. The one or more processor-readable media of claim 10, wherein the different capture settings comprise taking one image using a flash and taking the other image without using a flash.

12. The one or more processor-readable media of claim 10, wherein the different capture settings comprise taking one image with the foreground more in focus than the background and taking the other image with the background more in focus than the foreground.

13. The one or more processor-readable media of claim 10, wherein the first and second images have different pixel resolutions.

14. The one or more processor-readable media of claim 13, wherein the method further comprises matching the pixel resolutions of the first and second images prior to the comparing.

15. The one or more processor-readable media of claim 10, wherein the first and second images are captured by a digital camera.

16. The one or more processor-readable media of claim 15, wherein the first image is a relatively high resolution main image, and wherein the second image is a relatively low resolution pre- or post-view version of the first image.

17. The one or more processor-readable media of claim 15, wherein the first and second images are relatively low resolution pre- and/or post-view versions of a higher resolution main image of said scene also captured by the camera.

18. A digital image acquisition and processing apparatus, comprising: a lens;an image sensor;a processor; anda processor-readable medium having code embedded therein for programming the processor to perform a method for detecting a face in a mid-shot digital image of a person, wherein the method comprises: capturing first and second mid-shot digital images of approximately a same scene that is known to include at least a face and a portion of a torso of a human person, including using different capture settings such that the foreground is differently differentiated from the background in each image,comparing the first and second images to determine the foreground region of the images, andidentifying a portion of the foreground region likely to correspond to a face based upon matching one or more known face-to-torso geometric relationships to one or more geometric relationships between regions of the foreground region of the first and second images.

19. The apparatus of claim 18, wherein the different capture settings comprise taking one image using a flash and taking the other image without using a flash.

20. The apparatus of claim 18, wherein the different capture settings comprise taking one image with the foreground more in focus than the background and taking the other image with the background more in focus than the foreground.

21. The apparatus of claim 18, wherein the first and second images have different pixel resolutions.

22. The apparatus of claim 21, wherein the method further comprises matching the pixel resolutions of the first and second images prior to the comparing.

23. The apparatus of claim 18, wherein the first and second images are captured by a digital camera.

24. The apparatus of claim 23, wherein the first image is a relatively high resolution main image, and wherein the second image is a relatively low resolution pre- or post-view version of the first image.

25. The apparatus of claim 23, wherein the first and second images are relatively low resolution pre- and/or post-view versions of a higher resolution main image of said scene also captured by the camera.