IMAGE PROCESSING METHOD

• US 20130243322A1
• Filed: 11/13/2012
• Published: 09/19/2013
• Est. Priority Date: 03/13/2012
• Status: Active Grant

• Current Assignee: Hanwha Vision Company Limited (Hanwha Corporation), Korea University Research and Business Foundation
• Original Assignee: Samsung Techwin Corporation (Hanwha Corporation), Korea University Research and Business Foundation
• Inventors: NOH, Seung-in, KO, Han-seok, SONG, Tae-yup, LEE, Young-hyun, KU, Bon-hwa, KIM, Han-jun
• Granted Patent: US 8,908,967 B2
• Time in Patent Office: Days
• US Class Current: 382/171
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

An image processing method of separating an input image into a foreground image and a background image, the method including determining a pixel of the input image as a pixel of the foreground image if a foreground probability value of the pixel of the foreground image determined by using the Gaussian mixture model or the pixel determined to be included in a motion region is greater than a setting threshold.

Claims

1. An image processing method comprising:

(a) determining whether a pixel of an input image is a pixel of a foreground image or a pixel of the background image by using a Gaussian mixture model;

(b) determining whether the pixel of the input image is included in a motion region;

(c) obtaining a foreground probability value of the pixel of the input image according to a foreground probability histogram with respect to a correlation between the input image and a reference image of the input image; and

(d) determining the pixel of the input image as the pixel of the foreground image if the foreground probability value of the pixel of the foreground image determined by using the Gaussian mixture model or the pixel determined to be included in the motion region is greater than a setting threshold.

2. The method of claim 1, wherein the input image comprises a series of frame images of video, and

wherein operations (a) through (d) are repeatedly performed.

3. The method of claim 2, wherein, in operation (c), the reference image is separated into a foreground and a background, and is renewed as a previous image during a process of repeatedly performing operations (a) through (d).

4. The method of claim 1, wherein, in operation (c), the correlation between the input image and the reference image is calculated with respect to the pixel of the input image by using a texture value of the pixel that is a difference between a mean gradation value of the pixel and neighboring pixels and gradation values of the neighboring pixels.

5. The method of claim 1, wherein the input image is an image of an inside of a closed space captured by an image obtaining unit.

6. The method of claim 5, further comprising:

(e), if the pixel determined as the pixel of the foreground image is not connected to an image of a floor of the closed space, setting the pixel determined as the pixel of the foreground image as a pixel of the background image.

7. The method of claim 5, wherein, in operation (c), the correlation between the input image and the reference image is calculated with respect to the pixel of the input image by using a texture value of the pixel that is a difference between a mean gradation value of the pixel and neighboring pixels and gradation values of the neighboring pixels.

8. The method of claim 5, wherein the input image comprises a series of frame images of video, and

wherein operations (a) through (d) are repeatedly performed.

9. The method of claim 8, wherein, in operation (c), the reference image is separated into a foreground and a background, and is renewed as a previous image during a process of repeatedly performing operations (a) through (d).

19. The method of claim 6, wherein the input image is an image of an inside of a closed space captured by an image obtaining unit, and

wherein the method further comprises:

(e), if the pixel determined as the pixel of the foreground image is not connected to an image of a floor of the closed space, setting the pixel determined as the pixel of the foreground image as a pixel of the background image.

20. The method of claim 19, wherein, in operation (c), the correlation between the input image and the reference image is calculated with respect to the pixel of the input image by using a texture value of the pixel that is a difference between a mean gradation value of the pixel and neighboring pixels and gradation values of the neighboring pixels.

10. An image processing method of receiving an image from a surveillance camera for capturing an inside of a closed space having a door to an outside, determining whether each pixel of the received image is a pixel of a foreground image or a pixel of a background image, and separating the received image into the foreground image and the background image, the method comprising:

(a) determining whether a pixel of the received image is a pixel of a foreground image or a pixel of a background image by using a Gaussian mixture model;

(b) determining whether the pixel of the received image is included in a motion region;

(c) obtaining a foreground probability value of the pixel of the received image according to a foreground probability histogram with respect to correlation between the received image and a reference image of the received image;

(d) if the foreground probability value of the pixel of the foreground image determined by using the Gaussian mixture model or the pixel determined to be included in the motion region is greater than a setting threshold, obtaining a foreground-information binary number “

1”

as a first foreground determination value of the pixel of the received image, and, if the foreground probability value is not greater than the setting threshold, obtaining a background-information binary number “

0”

as the first foreground determination value;

(e) determining whether the door of the closed space is open or closed;

(f) if the door of the closed space is open, setting an auxiliary foreground determination value in the same way as the first foreground determination value, and, if the door of the elevator is closed, obtaining a final foreground determination value of a previous image and the auxiliary foreground determination value set based on the first foreground determination value; and

(g) if a result obtained by multiplying the auxiliary foreground determination value and the foreground probability value is greater than a final threshold, obtaining a final foreground determination value as the foreground-information binary number “

1”

, and, if the result is not greater than the final threshold, obtaining the final foreground determination value as the background-information binary number “

0”

.

11. The method of claim 10, further comprising:

(h) if the pixel determined as the pixel of the foreground image is not connected to an image of a floor of the closed space, setting the pixel determined as the pixel of the foreground image as the pixel of the background image.

12. The method of claim 10, wherein the received image comprises a series of frame images of video,

wherein operations (a) through (g) are repeatedly performed.

13. The method of claim 12, wherein, in operation (c), the reference image is separated into a foreground and a background, and is renewed as a previous image during a process of repeatedly performing operations (a) through (g).

14. The method of claim 10, wherein operation (e) comprises:

if a mean foreground probability value of pixels corresponding to top portions of the door of the closed space is greater than a setting mean threshold, determining the door of the closed space to be open.

15. The method of claim 10, wherein, in operation (c), the correlation between the received image and the reference image of the received image is calculated with respect to the pixel of the received image by using a texture value of the pixel that is a difference between a mean gradation value of the pixel and neighboring pixels and gradation values of the neighboring pixels.

16. An image processing method comprising:

(a) determining whether a pixel of an input image is a pixel of a foreground image or a pixel of a background image;

(b) determining whether the pixel is included in a motion region of the input image;

(c) obtaining a foreground probability value of the pixel according to a foreground probability histogram with respect to a correlation between the input image and a reference image of the input image; and

(d) determining that the pixel is included in the foreground image if the foreground probability value of the pixel determined as the pixel of the foreground image or the pixel determined to be included in the motion region is greater than a setting threshold.

17. The method of claim 16, wherein operation (a) is performed using a Gaussian mixture model.

18. The method of claim 16, wherein, in operation (c), the correlation between the input image and the reference image is calculated with respect to the pixel of the input image by using a texture value of the pixel that is a difference between a mean gradation value of the pixel and neighboring pixels and gradation values of the neighboring pixels.