METHOD OF CONSTRUCTING 3D CLOTHING MODEL BASED ON A SINGLE IMAGE

摘要

A method of constructing 3D clothing model based on single image, estimating a 3D model of human body of an inputted image and constructing 3D clothing plane according to the clothing silhouette of the inputted image. The method includes utilizing the 3D clothing plane and the 3D model of human body to generate a smooth 3D clothing model through a deformation algorithm. A decomposition algorithm of intrinsic image is utilized along with a shape-from-shading algorithm to acquire a set of detail information of clothing from the inputted image. A weighted Laplace editing algorithm is utilized to shift the acquired detail information of clothing to the smooth 3D clothing model to yield a final 3D clothing model. A 3D clothing model is used to generate the surface geometry details including folds, wrinkles.

主权项

1. A method of constructing 3D clothing model based on single inputted image, characterized in that it comprises the steps as follows:
estimating a 3D model of human body of and constructing a 3D clothing plane according to the clothing silhouette of the inputted image; utilizing a method of semi-automatic 3D pose estimation method to estimate 3D pose(s) of human body of the inputted image according to 2D joint(s) of human body in the inputted image specified by user; estimating the 3D model of human body in the inputted image through a method of deformable template according to a human contour line specified in the inputted image by user and the estimated 3D pose(s) of human body; specifying the clothing silhouette by user in the inputted image; utilizing the estimated 3D model of human body to divide the clothing silhouette specified by user into silhouette edge and boundary outline; calculating the projection area of 3D model of human body on the image and the skeleton(s) related to clothing area; calculating a directed surface for each skeleton related to the clothing area through 3D coordinate(s) of skeleton joint(s) and a relative rotation matrix; calculating an intersection line between adjacent directed surfaces which is used to calculate an internal cutting line of the clothing area of the inputted image; utilizing the internal cutting line to divide the clothing area of the inputted image into different parts of which each corresponding to one directed surface; projecting the outline and the internal cutting line of each part of the clothing area of the inputted image on the corresponding directed surface to form 3D clothing area of each part accordingly; triangulating the 3D clothing area of each part and utilize the apex(s) of the common internal cutting lines between different 3D clothing areas to form an initial 3D clothing plane; duplicating the initial 3D clothing plane; utilizing the apex(s) of the silhouette edge on the initial 3D clothing plane as the common apex(s) to combine these initial 3D clothing planes before and after duplication to form a final 3D clothing plane; utilizing the final 3D clothing plane and the 3D model of human body to yield a smooth 3D clothing model through a deformation algorithm; utilizing a Laplace deformation algorithm to deform the final 3D clothing plane under a set of constraints of outline apex(s) and the 3D model of human body to initialize a 3D clothing model; then calculating the tension degree of each apex of the initial 3D clothing model through calculating the shortest distance between the apex(s) of silhouette edge and the surface of the 3D model of human body; utilizing the rotating surface as the deformation constraint to initialize again the 3D clothing model in the area whose tension degree of apex(s) is marked loose to yield a smooth 3D clothing model; utilizing the intrinsic image decomposition algorithm and the shape-from-shading algorithm to acquire the detail information of clothing from the inputted image; shifting the acquired detail information of clothing to the smooth 3D clothing mode through the weighted Laplace editing algorithm to yield a final 3D clothing model; utilizing the intrinsic image decomposition algorithm to decompose the inputted image into a shading diagram and a reflectogram; utilizing the shading diagram and the shape-from-shading algorithm to calculate a relative depth value corresponding to each pixel point; establishing a corresponding relation between the clothing area of the inputted image and the final 3D clothing plane, and calculating the change of the relative depth of each apex on the final 3D clothing plane according to the depth value corresponding to the pixel point of the clothing area; renewing the final 3D clothing plane according to the calculated change of the relative depth to yield a 3D clothing detail plane; calculating separately the Laplace coordinates of the apex(s) on the final 3D clothing plane and the 3D clothing detail plane; utilizing these two Laplace coordinates of the apex(s) to calculate the detail information of surface geometry of the apex(s) on the 3D clothing; utilizing the Laplace coordinates of the apex(s) on the smooth 3D clothing model and the calculated detail information of surface geometry to calculate the shortest distance between the apex(s) and the apex(s) of silhouette edge so as to build a weighted Laplace deformation energy function; and utilizing a linear optimization algorithm to minimize the deformation energy function mentioned as above to transfer the detail information of clothing surface geometry to the smooth 3D clothing mode to yield a final 3D clothing model.