Method and apparatus for providing panoramic view with high speed image matching and mild mixed color blending

• US 20070159524A1
• Filed: 11/02/2006
• Published: 07/12/2007
• Est. Priority Date: 01/09/2006
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Samsung Electronics Co. Ltd.
• Inventors: Kim, Do Kyoon, Konouchine, Anton, Jung, Seok Yoon, Vezhnevets, Vladimir, Lee, Hee Sae, Kim, Keun Ho
• Granted Patent: US 8,189,031 B2
• Time in Patent Office: Days
• US Class Current: 348/36
• CPC Class Codes: G06T 3/4038 Image mosaicing, e.g. composing plane images from plane sub-images

Abstract

An apparatus and method of providing a panoramic image, the method including: preparing images of a plurality of levels, which are made by scaling each of two source images at several rates; computing a Sum of Square Difference (SSD) between two corresponding images in each of the plurality of levels of the two source images to form an image stitched together; and when stitching the two images, the overlapping area is divided into a plurality of parts and each part is assigned a weight which is linearly applied respectively to the divided part of each source image for blending color. The apparatus includes stitching units to stitch source images together at the plurality of levels and blending units to color blend the source images in the overlapped area.

Claims

1. A method of providing a panoramic image, the method comprising:

preparing images of a plurality of levels, which are made by scaling each of two source images at several rates; and

computing a Sum of Square Difference (SSD) with two corresponding images in each of the plurality of levels of the two source images to form a stitched together image.

2. The method of claim 1, wherein:

the SSD with respect to a level having a first number of pixels of the plurality of levels is computed in a first image search range, and

the SSD with respect to a level having a second larger number of pixels of the plurality of levels is computed in a second smaller image search range, so as to increase matching precision.

3. The method of claim 1, wherein the forming of the stitched together image comprises:

computing the SSD with respect to a certain area of the two corresponding images in a level having a first number of pixels of the images of a plurality of levels and detecting a position having the smallest SSD value, so as to first match the two source images; and

computing the SSD with respect to the certain area of the two corresponding images in a level having a second larger number of pixels than the level used in the first matching of the images of the plurality of levels and detecting a position having the smallest SSD value, so as to again match the first matched two images.

4. The method of claim 1, further comprising:

tilting or warping an image before the forming of the stitched together image; and

shear correcting after the forming of the stitched together image, so as to be used in image rendering.

5. A computer readable recording medium in which a program for executing the method of claim 1 is recorded.

6. A method of providing a panoramic image, the method comprising:

detecting a blend area of a certain percent from within a certain range of an overlapped part of two source images; and

generating a blended image by linearly applying a weight to corresponding pixels of the source images in the blend area.

7. The method of claim 6, wherein the weight is linearly applied to the corresponding pixels of the source images in the blend area for each of a plurality of portions which the overlapped part of the two source images is divided into.

8. The method of claim 6, wherein the certain range of the overlapped part is from 25 to 75%.

9. The method of claim 6, wherein the certain percent is 10%.

10. The method of claim 6, further comprising:

tilting or warping an image before the generating of the blended image; and

shear correcting after the generating of the blended image, so as to be used in image rendering.

11. A computer readable recording medium in which a program for executing the method of claim 6 is recorded.

12. A method of providing a panoramic image, the method comprising:

generating a blended image of a first relevant area according to a first weight function linearly applied to two corresponding images in a certain percent area from within a certain range of an overlapped part of two source images;

generating a blended image of a second relevant area according to a second weight function applied to one of the two corresponding images in one area outside the certain percent area within the certain range of the overlapped part of the two source images; and

generating a blended image of a third relevant area according to a third weight function applied to another of the two corresponding images in another area outside the certain percent area within the certain range of the overlapped part of the two source images.

13. The method of claim 12, wherein the first weight function uses linear functions linearly applied to the two corresponding images, an average and a standard deviation of each of the two source images, and an average and a standard deviation of the entire two source images.

14. The method of claim 12, wherein:

the second weight function uses a standard deviation of one of the two source images, the standard deviation of the entire two source images, and a linear functional standard deviation estimated with respect to the relevant area of the two source images to which a weight is applied; and

the third weight function uses a standard deviation of the other of the two source images, the standard deviation of the entire two source images, and a linear functional standard deviation estimated with respect to the relevant area of the two source images to which a weight is applied.

15. The method of claim 12, further comprising:

tilting or warping an image before the generating of the blended images according to the first, second and third weight functions; and

shear correcting after the generating of the blended images according to the first, second and third weight functions, so as to be used in image rendering.

16. A computer readable recording medium in which a program for executing the method of claim 12 is recorded.

17. A stitching apparatus for a panoramic image, the apparatus comprising:

a first stitching unit to compute an SSD between two corresponding images in a level having a first number of pixels in a plurality of levels which are made by scaling each of the two source images at several rates and detecting a position having the smallest SSD value to match the two source images; and

a second stitching unit to compute the SSD between two corresponding images in a level of the images having a second number of pixels larger than in the first stitching unit so as to again match the two source images matched by the first stitching unit.

18. The apparatus of claim 17, wherein an image search range for computing the SSD in the first stitching unit is larger than an image search range for computing the SSD in the second stitching unit.

19. A panoramic image blending apparatus, comprising:

a blend area search unit to search a certain percent of an overlapped part from within an input range as a blend area of two source images, and;

a weight application unit to generate a blended image by linearly applying a weight to two corresponding images in the blend area.

20. A blending apparatus for a panoramic image, the apparatus comprising:

a first blending unit to generate a blended image in a first relevant area according to a first weight function linearly applied to two corresponding images in a certain percent area from within a certain range of an overlapped part of two source images;

a second blending unit to generate a blended image of a second relevant area according to a second weight function applied to one of the two corresponding images in one area outside the certain percent area within the certain range of the overlapped part of the two source images; and

a third blending unit to generate a blended image of a third relevant area according to a third weight function applied to another of the two corresponding images in another area outside the certain percent area within the certain range of the overlapped part of the two source images.