System and method of generating projection image with region of interest

1. An image processing system, comprising:

a multiple-lens camera that captures a X-degree horizontal field of view (FOV) and a Y-degree vertical FOV to generate multiple lens images, where X<=360, Y<=180,

a vertex list generator configured to perform first operations comprising: (a) generating a first main vertex list according to a correspondence table; and (b) generating a first region of interest (ROI) vertex list according to the first main vertex list and a position information of a ROI when the ROI overlaps at least one measuring region inside an overlap region in a projection image; and

an image processing apparatus configured to generate the projection image according to the multiple lens images and a second main vertex list related to the first main vertex list in a rendering mode;

wherein the first main vertex list comprises first vertices forming a polygon mesh modeling the projection image, and wherein the first ROI vertex list comprises second vertices forming a transverse segment associated with an intersection between the ROI and the at least one measuring region.

2. The system according to claim 1 , further comprising:

an artificial intelligent (AI) module configured to perform at least one of human detection, human face detection and human head detection using a trained model over the projection image to generate the position information of the ROI.

3. The system according to claim 2 , wherein the AI module is implemented by an AI chip and the AI chip is implemented by a graphics processing unit, an application-specific integrated circuit, or a combination thereof.

4. The system according to claim 1 , further comprising:

a correspondence generator configured to conduct calibration for the first vertices to generate the correspondence table comprising the first vertices, wherein each first vertex provides a mapping between the multiple lens images and the projection image.

5. The system according to claim 1 , wherein the first operation (b) comprises:

performing interpolation over first texture coordinates of multiple first vertices adjacent to the second vertices for each lens image based on first and second destination coordinates of the first and the second vertices to generate second texture coordinates in each lens image for each second vertex;

wherein each first vertex has a first data structure and each second vertex has a second data structure, wherein each first data structure comprises the first destination coordinates in the projection image and the first texture coordinates in each lens image for a corresponding first vertex, and wherein each second data structure comprises the second destination coordinates in the projection image and the second texture coordinates in each lens image for a corresponding second vertex.

6. The system according to claim 5 , wherein the first operation (b) further comprises:

calculating two indices of two adjacent control points and a coefficient blending weight in each lens image for each second vertex according to the second texture coordinates in each lens image for each second vertex to generate the first ROI vertex list;

wherein each second data structure further comprises the two indices of two adjacent control points and the coefficient blending weight in each lens image for the corresponding second vertex.

7. The system according to claim 1 , further comprising:

a stitching decision device configured to perform second operations comprising:

generating a set of optimal warping coefficients for control points in all the measuring regions according to a set of test warping coefficients and accumulated pixel value differences of both all the measuring regions and the transverse segment;

wherein the projection image comprises at least one overlap region and each overlap region comprises multiple measuring regions; and

wherein the second main vertex list is further related to the set of optimal warping coefficients.

8. The system according to claim 7 , further comprising:

a vertex processing device configured to generate a third main vertex list according to the first main vertex list and the set of test warping coefficients, and to generate a second ROI vertex list according to the first ROI vertex list and the set of test warping coefficients in a measuring mode.

9. The system according to claim 8 , wherein the image processing apparatus is further configured to measure the accumulated pixel value differences of the measuring regions and the transverse segment according to the multiple lens images, the third main vertex list and the second ROI vertex list in the measuring mode.

10. The system according to claim 7 , wherein the second operations further comprise:

determining a selected decision group according to a selected control point; and

determining a temporary warping coefficient for the selected control point according to the set of test warping coefficients and local minimums among the accumulated pixel value differences corresponding to each control point in the selected decision group comprising the selected control point and one or two neighboring control points.

11. The system according to claim 10 , wherein the second operation of generating the set of optimal warping coefficients comprises:

when the transverse segment is a single block,

defining the test warping coefficient corresponding to a global minimum among the accumulated pixel value differences of the transverse segment as the temporary warping coefficient of the transverse segment; and

replacing a temporary warping coefficient of a control point inside a first measuring region with the temporary warping coefficient of the transverse segment when the transverse segment either overlaps the first measuring region or occupies more than a half of a second measuring region adjacent to the first measuring region.

12. The system according to claim 10 , wherein the second operation of generating the set of optimal warping coefficients comprises:

when the transverse segment is split into multiple sub-segments overlapping multiple measuring regions,

determining a selected decision group according to a selected sub-segment; and

determining a temporary warping coefficient for the selected sub-segment according to the set of test warping coefficients and local minimums among the accumulated pixel value differences corresponding to each sub-segment in the selected decision group comprising the selected sub-segment and one or two neighboring sub-segments.

13. The system according to claim 12 , wherein the second operation of generating the set of optimal warping coefficients comprises:

when the transverse segment is split into multiple sub-segments overlapping multiple measuring regions,

after the second operation of determining the temporary warping coefficient, replacing a temporary warping coefficient of a control point inside a first measuring region with the temporary warping coefficient of one of the sub-segments when the one of the sub-segments either overlaps the first measuring region or occupies more than a half of a second measuring region adjacent to the first measuring region.

14. The system according to claim 7 , wherein the set of test warping coefficients are associated with an offset that a lens center of the multiple lenses is separated from the camera system center of the multiple-lens camera.

15. An image processing method, applied to an image processing system comprising a multiple-lens camera that captures a X-degree horizontal field of view (FOV) and a Y-degree vertical FOV to generate multiple lens images, the method comprising:

obtaining a first main vertex list according to a correspondence table;

obtaining a first region of interest (ROI) vertex list according to the first main vertex list and a position information of a ROI when the ROI overlaps at least one measuring region inside an overlap region in a projection image; and

generating the projection image according to the multiple lens images and a second main vertex list related to the first main vertex list;

wherein the first main vertex list comprises first vertices forming a polygon mesh modeling the projection image, and wherein the first ROI vertex list comprises second vertices forming a transverse segment associated with an intersection between the ROI and the at least one measuring region, where X<=360, Y<=180.

16. The method according to claim 15 , further comprising:

performing at least one of human detection, human face detection and human head detection using a trained model over the projection image to generate the position information of the ROI prior to the step of obtaining the first ROI vertex list.

17. The method according to claim 15 , further comprising:

conducting calibration for the first vertices to obtain the correspondence table defining first vertex mapping between the multiple lens images and the projection image prior to the step of obtaining the first main vertex list.

18. The method according to claim 15 , wherein the step of obtaining the first ROI vertex list further comprises:

performing interpolation over first texture coordinates of the first vertices adjacent to the second vertices for each lens image based on first and second destination coordinates of the first and the second vertices to generate second texture coordinates in each lens image for each second vertex;

wherein each first vertex has a first data structure and each second vertex has a second data structure, wherein the first data structure comprises the first destination coordinates in the projection image and the first texture coordinates in each lens image for a corresponding first vertex, wherein the second data structure comprises the second destination coordinates in the projection image and the second texture coordinates in each lens image for a corresponding second vertex.

19. The method according to claim 18 , wherein the step of obtaining the first ROI vertex list further comprises:

calculating two indices of two adjacent control points and a coefficient blending weight in each lens image for each second vertex according to the second texture coordinates in each lens image for each second vertex;

wherein the second data structure further comprises the two indices of two adjacent control points and the coefficient blending weight in each lens image for the corresponding second vertex.

20. The method according to claim 15 , further comprising:

obtaining a set of optimal warping coefficients for control points in all the measuring regions according to a set of test warping coefficients and accumulated pixel value differences of both all the measuring regions and the transverse segment;

wherein the projection image comprises at least one overlap region and each overlap region comprises multiple measuring regions; and

wherein the second main vertex list is further related to the set of optimal warping coefficients.

21. The method according to claim 20 , wherein the step of obtaining the set of optimal warping coefficients comprises:

obtaining a third main vertex list according to the first main vertex list and the set of test warping coefficients;

obtaining a second ROI vertex list according to the first ROI vertex list and the set of test warping coefficients; and

measuring the accumulated pixel value differences of all the measuring regions and the transverse segment according to the multiple lens images, the third main vertex list and the second ROI vertex list.

22. The method according to claim 20 , wherein the step of obtaining the set of optimal warping coefficients comprises:

determining a selected decision group according to a selected control point; and

determining a temporary warping coefficient for the selected control point according to the set of test warping coefficients and local minimums among the accumulated pixel value differences corresponding to each control point in the selected decision group comprising the selected control point and one or two neighboring control point.

23. The method according to claim 22 , wherein the step of obtaining the set of optimal warping coefficients further comprises:

when the transverse segment is a single block,

defining the test warping coefficient corresponding to a global minimum among the accumulated pixel value differences of the transverse segment as the temporary warping coefficient of the transverse segment; and

replacing a temporary warping coefficient of a control point inside a first measuring region with the temporary warping coefficient of the transverse segment when the transverse segment either overlaps the first measuring region or occupies more than a half of a second measuring region adjacent to the first measuring region.

24. The method according to claim 22 , wherein the step of obtaining the set of optimal warping coefficients further comprises:

when the transverse segment is split into multiple sub-segments overlapping multiple measuring regions,

determining a selected decision group according to a selected sub-segment; and

determining a temporary warping coefficient for the selected sub-segment according to the set of test warping coefficients and local minimums among the accumulated pixel value differences corresponding to each sub-segment in the selected decision group comprising the selected sub-segment and one or two neighboring sub-segments.

25. The method according to claim 24 , wherein the step of obtaining the set of optimal warping coefficients further comprises:

when the transverse segment is split into multiple sub-segments overlapping multiple measuring regions,

after the step of determining the temporary warping coefficient, replacing a temporary warping coefficient of a control point inside a first measuring region with the temporary warping coefficient of one of the sub-segments when the one of the sub-segments either overlaps the first measuring region or occupies more than a half of a second measuring region adjacent to the first measuring region.

26. The method according to claim 20 , wherein the set of test warping coefficients are associated with an offset that a lens center of the multiple lenses is separated from the camera system center of the multiple-lens camera.