Reconstruction of shapes of objects from images

Abstract

The present disclosure describes a system and method for transforming a two-dimensional image of an object into a three-dimensional representation, or model, that recreates the three-dimensional contour of the object. In one example, three pairs of symmetric points establish an initial relationship between the original image and a virtual image, then additional pairs of symmetric points in the original image are reconstructed. In each pair, a visible point and an occluded point are mapped into 3-space with a single free variable characterizing the mapping for all pairs. A value for the free variable is then selected to maximize compactness of the model, where compactness is defined as a function of the model's volume and its surface area. “Noise” correction derives from enforcing symmetry and selecting best-fitting polyhedra for the model. Alternative embodiments extend this to additional polyhedra, add image segmentation, use perspective, and generalize to asymmetric polyhedra and non-polyhedral objects.

1 - 20 . (canceled) 
     
     
         21 . A system, comprising a processor and a memory in communication with the processor, the memory storing programming instructions executable by the processor to:
 accept input of a two-dimensional image of a 3-D object that is symmetric about a plane, the 3-D object having a surface;   accept input of three pairs of points, each pair being visible in the image, on the surface of the object, and symmetric about the plane;   create a model of the object by
 reconstructing coordinates in 3-space for each point in the three pairs of points in terms of exactly one free parameter; 
 determining coordinates in 3-space of at least one additional pair of symmetric points on the surface of the object, wherein
 each additional pair includes at least one point that is visible in the image, and 
 each additional pair is specified in terms of the free parameter; and 
 
 selecting a value for the free parameter that at least approximately maximizes a compactness metric as applied to the reconstructed object in 3-space; and 
   provide output that encodes the model.   
     
     
         22 . The system of  claim 21 , wherein the compactness metric is a function of the one free parameter representing the stretch of the model in the dimension that is being reconstructed from the image. 
     
     
         23 . The system of  claim 21 , wherein the compactness metric is the ratio of the volume of the model raised to power m, to the surface area of the model raised to power n (V m /S n ), where m<n. 
     
     
         24 . The system of  claim 23 , wherein m=2, and n=4.8. 
     
     
         25 . The system of  claim 21 , wherein the compactness metric is a function of the surface area of the model. 
     
     
         26 . The system of  claim 21 , wherein the selected value for the free parameter exactly maximizes the compactness metric as applied to the model. 
     
     
         27 . The system of  claim 21 , wherein the programming instructions are further executable by the processor to perform image segmentation on the input image to automatically detect vertices to serve as the three pairs of points. 
     
     
         28 . The system of  claim 21 , wherein the compactness metric is a function of the ratio of
 the total volume raised to power m, of one or more regular polyhedra that collectively circumscribe the object to   the total surface area raised to power n, of the one or more regular polyhedra,   where m<n.   
     
     
         29 . The system of  claim 28 , wherein m=2, and n=4.8. 
     
     
         30 . The system of  claim 21 , wherein the selecting function comprises generating the model from polyhedra, including selecting a best-fit polyhedron using a fit optimization method. 
     
     
         31 . A method of extracting three-dimensional structure of an object from a two-dimensional image of the object, comprising:
 creating a model of the object in 3-space by
 automatically reconstructing coordinates in 3-space for each point in three pairs of symmetric points in terms of a free parameter; 
 automatically determining coordinates in 3-space of at least one additional pair of symmetric points on the surface of the object, wherein
 each additional pair includes at least one point that is visible in the image, and 
 each additional pair is specified in terms of the free parameter; and 
 
 automatically selecting a value for the free parameter that at least approximately maximizes a compactness metric as applied to the model; and 
   generating an output signal that encodes the model.   
     
     
         32 . The system of  claim 31 , wherein the compactness metric is the ratio of the square of the volume of the model to a power of the surface area of the model, where the power is greater than two. 
     
     
         33 . The system of  claim 31 , wherein the compactness metric is a function that varies inversely with the surface area of the model. 
     
     
         34 . The system of  claim 31 , wherein the selected value for the free parameter maximizes the compactness metric as applied to the model. 
     
     
         35 . The system of  claim 31 , wherein the programming instructions are further executable by the processor to perform image segmentation on the input image to automatically detect vertices to serve as the three pairs of points. 
     
     
         36 . The system of  claim 31 , wherein the compactness metric is a function of the ratio of
 the square of the total volume of one or more regular polyhedra that collectively circumscribe the object to   a power of the total surface area of the one or more regular polyhedra,   where the power is greater than two.   
     
     
         37 . The system of  claim 31 , wherein the selecting function comprises generating the model from polyhedra, including selecting a best-fit polyhedron using a fit optimization method. 
     
     
         38 . The system of  claim 31 , wherein the automatically selecting step comprises generating the model from polyhedra, including selecting a best-fit polyhedron using a fit optimization method. 
     
     
         39 . A system for producing a three-dimensional model of an object from a two-dimensional image, comprising a processor and a memory, the memory storing programming instructions executable by the processor to:
 accept input of a two-dimensional image of a 3-D object that is symmetric about a plane, the 3-D object having a surface;   accept input of one or more pairs of points, each pair being visible in the image, on the surface of the object, and symmetric about the plane;   create a model of the object by
 reconstructing coordinates in 3-space for each point in the one or more pairs of points; 
 determining coordinates in 3-space of at least one additional pair of symmetric points on the surface of the object, wherein each additional pair includes at least one point that is visible in the image; and 
   output a signal that encodes the model.   
     
     
         40 . The system of  claim 39 , wherein the programming instructions are further executable by the processor to perform image segmentation on the input image to automatically detect vertices to serve as the one or more pairs of points.