METHOD AND APPARATUS FOR COMPOSING IMAGE

• US 20110200273A1
• Filed: 01/27/2011
• Published: 08/18/2011
• Est. Priority Date: 02/16/2010
• Status: Active Grant

• Current Assignee: INHA Industry Partnership Institute, Samsung Electronics Co. Ltd.
• Original Assignee: INHA Industry Partnership Institute, Samsung Electronics Co. Ltd.
• Inventors: Park, In-Kyu, Lee, Man-Hee, Cho, Sung-Dae, SINGHAL, Nitin
• Granted Patent: US 8,638,979 B2
• Time in Patent Office: Days
• US Class Current: 382/284
• CPC Class Codes: G06T 3/16 Spatio-temporal transformations, e.g. video cubism
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]
• CPC Class Codes: G06V 10/422 for representing the structure of the pattern or shape of an object therefor

Abstract

Provided are a method and apparatus for combining and composing a plurality of images into one image. An image processing unit calculates importance levels of the plurality of input images, respectively, according to selection criteria, selects at least two important images from among the plurality of input images according to the calculated importance levels, determines Regions Of Interest (ROIs) of the important images, selects the ROIs in one canvas, arranges corresponding important images at positions corresponding to the selected ROIs, and composes the arranged important images into one image.

Claims

1. A method for composing a plurality of images into one image in an image processing apparatus, the method comprising the steps of:

calculating importance levels of a plurality of input images, respectively, according to selection criteria, and selecting at least two important images from among the plurality of input images according to the calculated importance levels;

determining Regions Of Interest (ROIs) of the important images;

selecting the ROIs in one canvas; and

arranging corresponding important images at positions corresponding to the selected ROIs, and composing the arranged important images into one image.

2. The method of claim 1, wherein the selection criteria comprises at least one selected from the group consisting of an entropy of each image, a similarity level between images, a number of facial regions included in each image, and sizes of the facial regions included in each image.

3. The method of claim 1, wherein selecting important images comprises:

calculating similarity levels between each input image and each of remaining input images with respect to each of the plurality of input images when the selection criteria include a similarity level of each image;

determining a final similarity level of a corresponding input image using the calculated similarity levels;

selecting an image having a lowest similarity level among the plurality of input images as an important image; and

selecting an image having a lowest similarity level, in comparison with a previously selected important image, as a next important image.

4. The method of claim 1, wherein selecting important images comprises:

detecting sizes of facial regions included in each of the plurality of input images when the selection criteria include sizes of facial regions included in each image; and

determining importance levels of the input images in proportion to the sizes of the facial regions.

5. The method of claim 1, wherein each ROI is determined according to a position of a facial region included therein and entropy distribution of a related important image.

6. The method of claim 5, wherein determining ROIs comprises:

partitioning a corresponding important image into a plurality of blocks, and calculating an entropy corresponding to each of the plurality of blocks;

setting a value of a block in which a facial region is located to a settable maximum entropy value when the corresponding important image includes the facial region;

setting a block having a largest entropy value among the plurality of blocks as an initial ROI; and

continuously expanding an ROI in a direction showing a highest entropy increase rate in comparison with a number of pixels, among four directions of above, below, right, and left, in which the ROI is expandable, with the initial ROI as a center, until a final ROI includes information of a threshold value.

7. The method of claim 1, wherein selecting the ROIs in one canvas comprises:

partitioning the one canvas using a K-means clustering scheme, to arrange height-to-width ratios of bounding boxes of clusters and height-to-width ratios of the ROIs, respectively, in order of size;

selecting the ROIs in the clusters in regular sequence.

8. The method of claim 1, wherein selecting the ROIs in one canvas is performed in such a manner as to repeat a step of selecting initial positions of the ROIs in the one canvas, and a step of changing positions of the ROIs based on the initial positions and expanding sizes of the ROIs until boundaries of the ROIs are adjacent to each other.

9. An apparatus for composing a plurality of images into one image, the apparatus comprising:

a memory unit for storing a plurality of input images; and

an image processing unit for calculating importance levels of the plurality of input images, respectively, according to selection criteria, selecting at least two important images from among the plurality of input images according to the calculated importance levels, determining Regions Of Interest (ROIs) of the important images, selecting the ROIs in one canvas, arranging corresponding important images at positions corresponding to the selected ROIs, and composing the arranged important images into one image.

10. The apparatus of claim 9, wherein the selection criteria comprises at least one selected from the group consisting of an entropy of each image, a similarity level between images, a number of facial regions included in each image, and sizes of the facial regions included in each image.

11. The apparatus of claim 9, wherein the image processing unit calculates similarity levels between each input image and each of remaining input images with respect to each of the plurality of input images when the selection criteria include a similarity level of each image, determines a final similarity level of a corresponding input image using the calculated similarity levels, selects an image having a lowest similarity level among the plurality of input images as an important image, and selects an image having a lowest similarity level, in comparison with the selected important image, as a next important image.

12. The apparatus of claim 9, wherein the image processing unit detects sizes of facial regions included in each of the plurality of input images when the selection criteria include sizes of facial regions included in each image, and determines importance levels of the input images in proportion to the sizes of the facial regions.

13. The apparatus of claim 9, wherein each ROI is determined according to a position of a facial region included therein and entropy distribution of a related important image.

14. The apparatus of claim 13, wherein the image processing unit partitions a corresponding important image into a plurality of blocks, calculates an entropy corresponding to each of the plurality of blocks, sets a value of a block in which a facial region is located to a settable maximum entropy value when the corresponding important image includes the facial region, sets a block having a largest entropy value among the plurality of blocks as an initial ROI, and continuously expands an ROI in a direction showing a highest entropy increase rate in comparison with a number of pixels, among four directions of above, below, right, and left, in which the ROI is expandable, with the initial ROI as a center, until a final ROI includes information of a threshold value.

15. The apparatus of claim 9, wherein the image processing unit partitions the one canvas using a K-means clustering scheme, arranges height-to-width ratios of bounding boxes of clusters and height-to-width ratios of the ROIs, respectively, in order of size, and selects the ROIs in the clusters in regular sequence.

16. The apparatus of claim 9, wherein the image processing unit repeats to select initial positions of the ROIs in the one canvas, to change positions of the ROIs based on the initial positions, and to expand sizes of the ROIs until boundaries of the ROIs are adjacent to each other.