System and method for stitching images

Patent metadata

Field	Value
Publication number	US-10489885-B2
Application number	US-201815921646-A
Country	US
Kind code	B2
Filing date	Mar 14, 2018
Priority date	Mar 14, 2017
Publication date	Nov 26, 2019
Grant date	Nov 26, 2019

Abstract

An image stitching method includes: obtaining a first and a second images; selecting a first and a second search regions from the first and second images, respectively, the first and second search regions at least partially overlapping with each other along a first direction; dividing each of the first and second search regions into a plurality of equally-sized pixel blocks; moving the first and second images toward each other along the first direction using the pixel blocks as a step size, until each of the first and second search regions has been traversed; after each movement, determining a pixel difference value representing a difference between all overlapping pixel blocks of the first search region and the second search region; determining an optimal block-matching position for the first and second images based on the pixel difference value; and stitching the first and second images according to the optimal block-matching position.

First claim

What is claimed is:

1. An image stitching method, comprising:
obtaining a first source image and a second source image;
selecting a first search region from the first source image and a second search region from the second source image, the first search region and second search region at least partially overlapping with each other along a first direction;
dividing each of the first and second search regions into a plurality of pixel blocks, the pixel blocks in the first and second search regions having equal sizes;
moving the first source image and the second source image toward each other along the first direction using the pixel blocks as a step size, until each of the first and second search regions has been traversed;
after each movement, determining a pixel difference value representing a difference between all overlapping pixel blocks of the first search region and the second search region;
determining an optimal block-matching position for the first and second source images based on the pixel difference value; and
stitching the first and second source images according to the block-matching optimal position.

2. The method according to claim 1 , further comprising:
moving the first source image and the second source image toward each other along a row direction of the pixel blocks;
after each movement, determining the pixel difference value by averaging block pixel difference values of all overlapping pixel blocks, each of the block pixel difference value being a difference between average pixel values of two overlapping pixel blocks.

3. The method according to claim 2 , wherein when the first and second search regions have equal sizes, the following formula is used to compute the average value of the block pixel difference values:

Vb
=

∑

i
=
1

M

⁢

∑

j
=
1

N

⁢

(

Lij
-
Rij

)

×

(

Lij
-
Rij

)

n
×
M

;

wherein Vb is the average value of the block pixel difference values, M is the number of rows of pixel blocks in the search regions, N is the number of columns of pixel blocks in the search regions, Lij is the average value of all pixel values in the pixel block at row i and column j of the first source image, Rij is the average value of all pixel values in the pixel block at row i and column j of the second source image, and n is the movement as measured by the number of pixel block columns.

4. The method according to claim 2 , wherein the selection of the optimal block-matching position for the first source image and the second source image based on the pixel difference value comprises:
when the average value of the block pixel difference values is at its minimum, determining the relative position of the first source image and second source image as the block-matching optimal position.

5. The method according to claim 2 , wherein the stitching of the first source image and the second source image according to the optimal block-matching position comprises:
moving the first source image and the second source image toward each other along the row direction with the pixel as the step size in the block-matching optimal position, and computing the difference value of the pixel values of all overlapping pixels after each movement;
when the difference value of the pixel values is at its minimum, determining the relative position of the first source image and second source image as the pixel-matching optimal position; and
stitching the first source image and the second source image according to the pixel-matching optimal position.

6. The method according to claim 5 , wherein the following formula is used to compute the difference value of the pixel values:
Vp=Σ x=1 H Σ y=1 L ( Lxy−Rxy )×( Lxy−Rxy );

wherein Vp is the difference value of the pixel values; H is the number of rows of all overlapping pixels in the block-matching optimal position, L is the number of columns of all overlapping pixels in the block-matching optimal position, Lxy is the pixel value of the pixel at row x and column y of the first source image, and Rxy is the pixel value of the pixel at row x and column y of the second source image.

7. The method according to claim 1 , wherein after the obtainment of the first source image and the second source image to be stitched, the method further comprises:
converting the first source image and second source image into images in the HSV format, the pixel difference value being the hue difference value.

8. The method according to claim 5 , wherein after the obtainment of the first source image and the second source image to be stitched, the method further comprises:
computing the hue difference value and the gradient difference value of all pixels in the first source image and the second source image; and
computing the sum value of the hue difference value and gradient difference value of all pixels, and selecting the pixels corresponding to the minimum sum value as the seam position.

9. The method according to claim 8 , wherein the stitching of the first source image and the second source image according to the optimal block-matching position comprises:
stitching the first source image and the second source image in the pixel-matching optimal position according to the seam position.

10. The method according to claim 9 , wherein the stitching of the first source image and the second source image in the pixel-matching optimal position according to the seam position comprises:
in the regions between the seam position and the pixel-matching optimal position, selecting pixels corresponding to the first source image and the second source image, respectively, for the stitching; and
along the seam position, using linear blending to stitch pixels corresponding to the first source image and the second source image, in order to complete the stitching of the first source image and the second source image.

11. The method according to claim 8 , wherein before the stitching of the first source image and the second source image according to the block-matching optimal position, the method further comprises:
converting the format of the first source image and the second source image to YCrCb;
after the format conversion, computing the average of luminance values in the first source image and the second source image at t

Assignees

• Shanghai Xiaoyi Tech Co Ltd

Inventors

• Ding Dajun
• Zhao Lili

Classifications

• G06T3/4038 Primary

  Image mosaicing, e.g. composing plane images from plane sub-images · CPC title

• H04N5/265 Primary

  Mixing · CPC title

• H04N5/2624

  for obtaining an image which is composed of whole input images, e.g. splitscreen · CPC title

• G06V10/25

  Determination of region of interest [ROI] or a volume of interest [VOI] · CPC title

• G06V10/758

  Involving statistics of pixels or of feature values, e.g. histogram matching · CPC title