Method of using MPEG-7 standard in object segmentation

• US 20030035479A1
• Filed: 11/21/2001
• Published: 02/20/2003
• Est. Priority Date: 08/14/2001
• Status: Active Grant

• Current Assignee: National Chung Cheng University
• Original Assignee: National Chung Cheng University
• Inventors: Kuo, Chung J., Kan, Ming-Cheng
• Granted Patent: US 6,728,314 B2
• Time in Patent Office: Days
• US Class Current: 375/240.12
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20152 Watershed segmentation
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/155 involving morphological operators
• CPC Class Codes: G06V 10/267 by performing operations on regions, e.g. growing, shrinking or watersheds
• CPC Class Codes: G06V 10/56 relating to colour

Abstract

The invention relates to a method of using MPEG-7 in object segmentation to extract a desired moving or still image object. The method includes extracting the feature of an existing video object Plane (VOP) by MPEG-7 technique and storing it in an MPEG-7 database, segmenting an input image into a plurality of objects by using the watershed process, comparing each of the plurality of objects to the stored object descriptor feature, and extracting the shape and position of the most similar object in the input image.

Claims

1. A method of using MPEG-7 in the object segmentation, comprising the steps of:

extracting the feature of an existing video object Plane (VOP) by the MPEG-7 technique, storing it in an MPEG-7 database and presetting a value;

segmenting an input video image into a plurality of regions by color transformation and the Watershed process;

setting the merge adjustment condition using a merge means and a threshold value to minimize the number of regions segmented;

finding all possible region combinations using a region selector;

finding the highest similarity region combination using a comparator to compare the VOP features stored in the MPEG-7 database to all possible region combinations found by the region selector and outputting the highest similarity region combination as a target VOP, wherein,

the threshold value is reduced to increase the number of regions segmented if the target VOP is greater than the preset value and the threshold value is increased to reduce the number of regions segmented if the target VOP is smaller than the preset value, so that the highest similarity region combination output is more similar to the VOP in the input video image;

repeating the acquisition of highest similarity region combination step until the adjusted threshold value meets any of the preset values and the value zero; and

outputting the highest similarity region combination, after the threshold value adjustment is completed, as the final target VOP.

2. The method as claimed in claim 1, further comprising a MPEG-7 presetting database step, which is implemented by the MPEG-7 descriptors defined in the specification recommendation.

3. The method as claimed in claim 2, wherein the MPEG-7 descriptors comprise the VOP'"'"'s color, shape, and motion.

4. The method as claimed in claim 3, wherein the color comprises color space, dominant color, color histogram and color layout.

5. The method as claimed in claim 3, wherein the shape comprises object bounding box, region-based shape descriptor and contour-based shape descriptor.

6. The method as claimed in claim 3, wherein the motion comprises camera motion, object motion trajectory and parametric object motion.

7. The method as claimed in claim 1, wherein the color transformation technique uses RGB mode to convert the image into a monochrome.

8. The method as claimed in claim 7, wherein the monochrome is gray level.

9. The method as claimed in claim 1, wherein the watershed process comprises:

detecting the region with the smallest gray level as a start of the region expanding to the water top adjacent to a basin, where a separate dam is added to isolate the region from the water of other regions.

10. The method as claimed in claim 1, wherein the region merger is performed by the merge means when the mean difference of the gray level between the regions is smaller than the preset value.

11. The method as claimed in claim 1, wherein the region merger is performed by the merge means when the expected value difference of the gray level between the regions is smaller than the preset value.

12. The method as claimed in claim 1, wherein the using the comparator comprises:

defining a respective similarity matching criteria for each descriptor using the MPEG-7 technique;

comparing the image combined by the merge means to the comparison equation representing the feature of the existing VOP stored in the database; and

outputting the VOP having the most sets with the least comparison difference from the similarity matching criteria for the MPEG-7 descriptors.

13. The method as claimed in claim 12, wherein the comparison equation uses the MPEG-7 descriptor similarity matching criteria equations comprising the following color histogram comparison equation:

dist(A, B)=[hist(A)−

hist(B)]^TW[hist(A)−

hist(B)]

where hist [A] is the VOP'"'"'s color histogram set of the input image, hist [B] the existing VOP'"'"'s color histogram set and W the weight.

14. The method as claimed in claim 13, wherein the weight W depends on the color transformation used.

15. The method as claimed in claim 14, wherein the color transformation mode comprises HSV mode, RGB mode, and YcbCr mode.