초록 ▼

To compose an image from a plurality of partially overlapping images, both positional movement and perspective transformation are performed using pyramidal data structures. After the positional movement is determined for a unit block of one partially overlapping image to match a corresponding block,

To compose an image from a plurality of partially overlapping images, both positional movement and perspective transformation are performed using pyramidal data structures. After the positional movement is determined for a unit block of one partially overlapping image to match a corresponding block, a set of perspective transformation parameters is determined based upon the positional movement. The partially overlapping images are composed into a single image while the partially overlapping images efficiently undergo the perspective transformation according to the above determined parameters.

대표청구항 ▼

What is claimed is: 1. A method of composing partially overlapping images, comprising: inputting partially overlapping images of at least a common object; dividing at least two of the overlapping images into a predetermined number of blocks; determining a positional relation between a correspondi

What is claimed is: 1. A method of composing partially overlapping images, comprising: inputting partially overlapping images of at least a common object; dividing at least two of the overlapping images into a predetermined number of blocks; determining a positional relation between a corresponding pair of the blocks in the two overlapping images based upon following equations, where c0 through c7 are perspective parameters, xshiftN and yshiftN are perspective transformation parameters, and pN and qN are positional coordinates; determining the perspective transformation parameters based upon the positional relation and an approximation for the above equations, the approximation being x shiftN≈c0pN+c1 qN+c2-c6pN xshiftN-c7qNxshift N; and y shiftN≈c3pN+c4 qN+c5-c6pN yshiftN-c7qNyshiftN composing the overlapping images according to perspective transformation based upon the perspective transformation parameters. 2. The method of composing partially overlapping images according to claim 1 wherein the positional relation is determined based upon an amount of parallel movement necessary to match the corresponding pair of the blocks. 3. The method of composing partially overlapping images according to claim 1 wherein the positional relation is determined based upon affine transformation. 4. The method of composing partially overlapping images according to claim 1 wherein the positional relation is determined based upon correlational coefficients representing the overlapping images. 5. The method of composing partially overlapping images according to claim 1 wherein the perspective transformation parameters are determined based upon a method of least squares. 6. The method of composing partially overlapping images according to claim 1 wherein the inputted partially overlapping images are organized into a pyramidal data structure having a predetermined number of hierarchical layers prior to said dividing at least two of the overlapping images, the hierarchical layers including at least a minimal layer and a maximal layer, said determining the positional relation being sequentially refined from the minimal layer to the maximal layer. 7. The method of composing partially overlapping images according to claim 6 wherein the pyramidal data structure is generated by selecting one of a predetermined set of criteria for reducing information from the maximal layer towards the minimal layer. 8. The method of composing partially overlapping images according to claim 7 wherein the one of the criteria is selected based upon pixel value characteristic of the overlapping images. 9. The method of composing partially overlapping images according to claim 8 wherein the pixel value characteristics include a black-pixel image on a white background and a white pixel image on a black background. 10. A system for composing partially overlapping images, comprising: an image input unit for inputting partially overlapping images of at least a common object; a block dividing unit connected to said image input unit for dividing at least two of the overlapping images into a predetermined number of blocks; a positional relation determination unit connected to said block dividing unit for determining a positional relation between a corresponding pair of the blocks in the two overlapping images based upon following equations, where c0 through c7 are perspective parameters, xshiftN and yshiftN are perspective transformation parameters, and pN and qN are positional coordinates; a perspective transformation parameter determination unit connected to said positional relation determination unit for determining the perspective transformation parameters based upon the positional relation and an approximation for the above equations, the approximation being x shiftN≈c0pN+c1 qN+c2-c6pN xshiftN-c7qNxshiftN y shiftN≈c3pN+c4 qN+c5-c6pN yshiftN-c7qNyshift N; and a composing unit connected to said perspective transformation parameter determination unit for composing the overlapping images according to perspective transformation based upon the perspective transformation parameters. 11. The system for composing partially overlapping images according to claim 10 wherein said positional relation determination unit determines an amount of parallel movement necessary to match the corresponding pair of the blocks. 12. The system for composing partially overlapping images according to claim 10 wherein said positional relation determination unit determines the positional relation based upon affine transformation. 13. The system for composing partially overlapping images according to claim 10 wherein said positional relation determination unit determines the positional relation based upon correlational coefficients representing the overlapping images. 14. The system for composing partially overlapping images according to claim 10 wherein said perspective transformation parameter determination unit determines the perspective transformation parameters based upon a method of least squares. 15. The system for composing partially overlapping images according to claim 10 further comprising a pyramidal data structure generation unit connected to said image input unit for generating a pyramidal data structure having a predetermined number of hierarchical layers based upon the inputted partially overlapping images, the hierarchical layers including at least a minimal layer and a maximal layer, said positional relation determination unit sequentially refining the positional relation from the minimal layer to the maximal layer. 16. The system for composing partially overlapping images according to claim 15 wherein said pyramidal data structure generation unit generates the pyramidal data structure by selecting one of a predetermined set of criteria for reducing information from the maximal layer towards the minimal layer. 17. The system for composing partially overlapping images according to claim 16 wherein said pyramidal data structure generation unit selects the one of the criteria based upon pixel value characteristic of the overlapping images. 18. The system for composing partially overlapping images according to claim 17 wherein the pixel value characteristics include a black-pixel image on a white background and a white-pixel image on a black background. 19. A storage medium storing computer executable instructions for composing partially overlapping images, the computer executable instructions performing the steps of: inputting partially overlapping images of at least a common object; dividing at least two of the overlapping images into a predetermined number of blocks; determining a positional relation between a corresponding pair of the blocks in the two overlapping images based upon following equations, where c0 through c7 are perspective parameters, xshiftN and yshiftN are perspective transformation parameters, and pN and qN are positional coordinates; and determining the perspective transformation parameters based upon the positional relation and an approximation for the above equations, the approximation being x shiftN≈c0pN+c1 qN+c2-c6pN xshiftN-c7qNxshiftN y shiftN≈c3pN+c4 qN+c5-c6pN yshiftN-c7qNyshift N; and composing the overlapping images according to perspective transformation based upon the perspective transformation parameters. 20. The storage medium storing computer executable instructions according to claim 19 wherein the perspective transformation parameters are determined based upon a method of least squares. 21. The storage medium storing computer executable instructions according to claim 19 wherein the inputted partially overlapping images are organized into a pyramidal data structure having a predetermined number of hierarchical layers prior to said dividing at least two of the overlapping images, the hierarchical layers including at least a minimal layer and a maximal layer, said determining the positional relation being sequentially refined from the minimal layer to the maximal layer. 22. The method of composing partially overlapping images according to claim 1 where the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c1pN+c0 qN+c3. 23. The method composing partially overlapping images according to claim 1 wherein the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c3pN+c4 qN+c5. 24. The system for composing partially overlapping images according to claim 10 wherein the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c1pN+c0 qN+c3. 25. The system for composing partially overlapping images according to claim 10 wherein the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c3pN+c4 qN+c5. 26. The storage medium storing computer executable instructions according to claim 19 wherein the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c1pN+c0 qN+c3. 27. The storage medium storing computer executable instructions according to claim 19 wherein the approximation is made as follows: x shiftN≈c0pN+c1 qN+c2 y shiftN≈c3pN+c4 qN+c5.