Line removal and object detection in an image
View Patent ↗In one aspect, lines in image data of an event are automatically found and repaired. For example, the event may be a sporting event which is played on a field, and the line segment is a field line on the field which may be obscured by a player, game ball or other object. The line segment is automatically detected in a mask image, and a portion of the line segment which is occluded by the object is automatically determined, and the object is automatically removed. The line segment can also be repaired. Optionally, a virtual viewpoint of the event is provided from the image, with the line repaired and the object removed. In another aspect, an object in an image of an event is automatically located by detecting blobs in the image which meet at least one specified criterion, such as size, aspect ratio, density or color profile.
1. A method for automatically finding and repairing lines in image data of a live event, comprising:
obtaining an image of the live event from a camera, the live event includes a scene having at least one line segment which includes a portion which is occluded by an object, the object having a width which is greater than a width of the portion;
forming a mask image from the image obtained from the camera;
automatically detecting the at least one line segment in the mask image;
automatically determining that the portion of the at least one line segment is occluded by the object;
automatically removing the object;
automatically repairing the portion of the at least one line segment which is occluded by the object to provide a repaired portion of the at least one line segment; and
depicting at least one virtual viewpoint of the live event which is rotated by an angular offset from a viewpoint of the camera, the at least one virtual viewpoint of the live event uses texture from the image of the live event and includes the repaired portion of the at least one line segment.
2. The method of claim 1 , wherein:
the automatically detecting comprises obtaining points from vertical and horizontal scanning of segments in the mask image, finding center points of the segments of the mask image from intersections of the points from the vertical and horizontal scanning, applying a Hough transform using the center points, and finding at least one straight line segment based on the Hough transform.
3. The method of claim 2 , wherein:
the automatically detecting comprises discarding center points of the at least one straight line segment, and chaining together remaining center points which are on a continuous curve and which have a tangent angle which is less than a threshold angle.
4. The method of claim 1 , wherein:
the automatically determining comprises determining an average width of a plurality of detected line segments in the mask image, comparing a width of the object to the average width, and determining that the width of the object exceeds the average width by a margin.
5. The method of claim 1 , further comprising:
combining data of a textured 3d model of the event and data from the image obtained from the camera, with the object removed and the portion repaired, to depict the at least one virtual viewpoint of the event.
6. The method of claim 1 , wherein:
the portion of the at least one line segment which is occluded by the object is next to a portion of the at least one line segment which is not occluded by the object; and
the at least one virtual viewpoint of the live event includes the portion of the at least one line segment which is not occluded by the object.
7. The method of claim 6 , wherein:
the automatically repairing includes filling in pixel data of the portion of the at least one line segment which is occluded by the object using the portion of the at least one line segment which is not occluded by the object.
8. The method of claim 6 , wherein:
the live event is a sporting event which is played on a field, the at least one line segment is a field line on the field, and the object is a participant in the sporting event.