Method of realism assessment of an image composite

• US 8,373,721 B2
• Filed: 12/14/2009
• Issued: 02/12/2013
• Est. Priority Date: 12/14/2009
• Status: Active Grant

• Current Assignee: National Taiwan University (Government of The Republic of China)
• Original Assignee: National Taiwan University (Government of The Republic of China)
• Inventors: Wong, Bing-Yi, Chen, Homer H., Liang, Chia-Kai, Lin, Tai-Hsu
• Primary Examiner(s): Sajous, Wesner
• Application Number:
  US12/637,608
• Publication Number:
  US 20110141353A1
• Time in Patent Office: 1,156 Days
• Field of Search: 345418-419, 345/428, 345/581, 345589-591, 345592-593, 345600-604, 345/606, 345610-611, 345/619, 345/624, 345629-630, 345/640, 345/643, 345/660, 345/670, 345472-4721, 345/690, 345547-549, 345/555, 348254-257, 348/263, 348552-553, 348/557, 348/560, 348/565, 348577-586, 348/591, 348598-600, 348630-631, 348/655, 348659-661, 348/668, 348671-672, 348674-657, 348/708, 358515-518, 358/522, 358/525, 358/540, 358447-448, 358/450, 358/453, 382/162, 382166-167, 382168-169, 382/173, 382/232, 382/235, 382/254, 382/274, 382/276, 382/251, 382282-285, 382293-296, 382298-300, 382/305
• US Class Current: 345/591
• CPC Class Codes: G06T 11/001 Texturing; Colouring; Generation of texture or colour inpainting G06T5/77

Abstract

In a method of realism assessment of an image composite, the foreground and the background are transformed into a color space, such as YCbCr, followed by projecting the foreground and the background to a subspace represented by the axes representing chromatic information. The image composite is assessed in the projected subspace, according to linearity of color distributions of the foreground/the background, or according to distance between the color distributions of the foreground/the background and a center of the transformed color space.

Claims

1. A method of realism assessment of an image composite, comprising an image processor for performing the following steps:

providing the image composite having a foreground and a background;

transforming the foreground and the background image composite into a color space, wherein the transformed color space has a first axis representing luminance information, and a second axis and a third axis representing chromatic information;

projecting the foreground and the background images to a subspace represented by the second axis and the third axis; and

assessing the image composite, in the projected subspace, according to a linearity of color distribution of the foreground image and a linearity of color distribution of the background image, and further according to a relationship between the color distribution of the foreground image and the color distribution of the background image.

2. The method of claim 1, wherein the foreground image is an inserted object of the image composite.

3. The method of claim 1, wherein the transformed color space is YCbCr color space, and the first axis is Y axis, the second axis is Cb axis, and the third axis is Cr axis.

4. The method of claim 1, wherein the transformed color space is YUV color space or L*a*b* color space.

5. The method of claim 1, wherein the image composite is assessed as realistic when the foreground and the background images have linear color distributions that are approximately consistent or when neither the foreground nor the background image has a linear color distribution, and wherein the image composite is assessed as unrealistic when the foreground and the background images have linear color distributions that are substantially inconsistent.

6. The method of claim 1, further comprising a step of:

firstly rotating color distribution of the foreground image and/or the background image around the first axis, such that the color distribution of the foreground image is approximately aligned or parallel with the color distribution of the background image in the projected subspace.

7. The method of claim 6, further comprising a step of:

secondly rotating the color distribution of the foreground image and/or the background image around the second axis or the third axis, such that the color distribution of the foreground image is approximately aligned or parallel with the color distribution of the background image in a subspace represented by the first axis and the third axis or in a subspace represented by the first axis and the second axis.

8. A method of realism assessment of an image composite, comprising an image processor for performing the following steps:

providing the image composite having a foreground and a background;

transforming the foreground and the background image composite into a color space, wherein the transformed color space has a first axis representing luminance information, and a second axis and a third axis representing chromatic information;

measuring distances between color distributions of the foreground and the background images respectively and a center of the transformed color space as grayness; and

assessing the image composite, in the transformed color space, according to the measured grayness of the foreground image and the grayness of the background image.

9. The method of claim 8, wherein the foreground image is an inserted object of the image composite.

10. The method of claim 8, wherein the transformed color space is YCbCr color space, and the first axis is Y axis, the second axis is Cb axis, and the third axis is Cr axis.

11. The method of claim 8, wherein the transformed color space is YUV color space or L*a*b* color space.

12. The method of claim 8, wherein the grayness is measured as negative of a distance from the center of the transformed color space to a nearest point of the color distribution of the foreground or the background image.

13. The method of claim 8, wherein the image composite is assessed as unrealistic when a hollow region exists between the color distribution of the foreground image and the color distribution of the background image, and otherwise the image composite is assessed as realistic.

14. The method of claim 8, further comprising a step of:

firstly rotating color distribution of the foreground image and/or the background image around the first axis, such that the color distribution of the foreground image is approximately aligned or parallel with the color distribution of the background image in a subspace represented by the first axis and the third axis or in a subspace represented by the first axis and the second axis.

15. The method of claim 14, further comprising a step of:

secondly rotating the color distribution of the foreground image and/or the background image around the second axis or the third axis, such that the color distribution of the foreground image is approximately aligned or parallel with the color distribution of the background image in the subspace represented by the first axis and the third axis or in the subspace represented by the first axis and the second axis.

16. A method of realism assessment of an image composite, comprising an image processor for performing the following steps:

providing the image composite having a foreground and a background;

shrinking the foreground and the background image composite to obtain local color distributions of the foreground and the background images; and

determining color similarity between the shrunk foreground image and the shrunk background image by histogram distance;

wherein the determined color similarity S is expressed as

17. The method of claim 16, wherein the foreground image is an inserted object of the image composite.

18. The method of claim 17, wherein the shrunk background image is a band region outside a boundary of the inserted object, and the shrunk foreground image is a band region inside the boundary of the inserted object.

19. The method of claim 16, further comprising a step of:

adjusting histogram of the shrunk foreground image and the shrunk background image such that distribution of the shrunk foreground image and distribution of the shrunk background image are approximately matched.

20. The method of claim 19, wherein the histogram adjustment is performed by:

transforming, channel-by-channel, color of each pixel of the shrunk foreground image in CIE L*a*b* space by using the following equation,

c′

=(σ

_B/σ

_F)(c−

μ

_F)+μ

_B,

where c denotes a value of the pixel in an input image for a given channel, c′

denotes a modified pixel value, μ

denotes mean, σ

denotes standard deviation, subscript B stands for the shrunk background image, and subscript F stands for the shrunk foreground image.