Kabushiki Kaisha Toshiba

Minh-Tri Pham

Bjorn Stenger

Frank Perbet

Oliver Woodford

Roberto Cipolla

Atsuto Maki

Sam Johnson

Riccardo Gherardi

Pablo Alcantarilla

In a method of generating a three dimensional representation of a human body shape from a depth image of a clothed human subject is disclosed. The method comprises identifying at least one sample from a plurality of pre-calculated representative samples, each pre-calculated representative sample comprising a three dimensional representation of a human body shape and a corresponding depth map, by comparing the depth maps of the pre-calculated representative samples with the depth image of the clothed human subject; determining a neighbourhood of the data space of possible depth images of unclothed human body shapes from the depth map of the at least one representative sample; generating a parametric model for the human body shape from a neighbourhood of the data space of human body shapes, the neighbourhood of the data space of human body shapes corresponding to the neighbourhood of the data space of possible depth images of unclothed human body shapes; and generating the three dimensional representation of the human body shape by fitting the parametric model to the depth image of the clothed human subject.

Methods and systems for generating a three dimensional representation of a human body shape

In one embodiment a method of generating a three dimensional model of a subject is disclosed. The method comprises receiving point cloud data for a subject; receiving user inputs indicating a plurality of cubes, and a grouping of the cubes into a plurality of blocks to form a representation of a class of subjects; generating a first mesh comprising a plurality of quadrilaterals by subdividing patches corresponding to faces of blocks of the plurality of blocks; fitting the first mesh to the point cloud data to generate a fitted mesh; iteratively generating further meshes, each comprising a plurality of quadrilaterals by subdividing patches of the fitted mesh from the previous iteration and fitting the further mesh to the point cloud data; and outputting as the three dimensional model of the subject the iteratively generated fitted mesh.

Method and systems for generating a three dimensional model of a subject by iteratively generating meshes

A method of object recognition and/or registration includes receiving a point cloud, arranging the points of the point cloud into a hierarchical search tree, and determining geometric information of the points located within a region, by identifying a highest level tree nodes where all of descendent leaf nodes are contained within the region and selecting the leaf nodes for the points where no sub-tree is entirely contained within the region, such that the points falling within the region are represented by the smallest number of nodes and performing statistical operations on the nodes representing the points in the region. The geometric information includes descriptors of features in the point cloud. The method further includes comparing the feature descriptors with a database of feature descriptors for a plurality of objects.

Image analysis method by analyzing point cloud using hierarchical search tree

According to one embodiment, a method of generating a three dimensional representation of a subject from a depth image, comprises comparing a depth image of the subject with a plurality of representative images, wherein each representative image is associated with a respective parametric model of a subject; identifying a representative image of the plurality of representative images as a closest representative image to the depth image of the subject; selecting the parametric model associated with the closest representative image to the depth image; and generating a three dimensional representation of the subject by fitting the selected parametric model to the depth image of the subject.

Methods and systems for generating a three dimensional representation of a subject

A method of comparing two object poses, wherein each object pose is expressed in terms of position, orientation and scale with respect to a common coordinate system, the method comprising: where X is the object pose of one object and Y is the object pose of the other object, s(X) and s(Y) are scalar functions representing the scale of the object poses X and Y respectively, R(X) and R(Y) are matrices expressing the rotation of object poses X and Y respectively, t(X) and t(Y) are vectors expressing the translation of object poses X and Y respectively, and &#x3c3;, &#x3c3;and &#x3c3;are weighting factors for d, dand drespectively.

Image processing method and system

An object location method includes: analyzing data including plural objects each including plural features, and extracting the features from the data; matching features stored in a database with those extracted from the data, and deriving a prediction of the object, each feature extracted from the data providing a vote for at least one prediction; expressing the prediction to be analyzed in a Hough space, the objects to be analyzed being described by n parameters and each parameter defining a dimension of the Hough space, n is an integer of at least one; providing a constraint by applying a higher weighting to votes which agree with votes from other features than those votes which do not agree with votes from other features; finding local maxima in the Hough space using the weighted votes; and identifying the predictions associated with the local maxima to locate the objects provided in the data.

Object location method and system

A method dividing an image into plural superpixels of plural pixels of the image. The method calculates an initial set of weights from a measure of similarity between pairs of pixels, from which a resultant set of weights is calculated for pairs of pixels that are less that a threshold distance apart on the image. The calculation calculates a weight for a pair of pixels as the sum over a set of third pixels of the product of initial weight of the first pixel of the pair of pixel with the third pixel and the weight of the third pixel with the second pixel. Each weight is then subjected to a power coefficient operation. The resultant set of weights and the initial set of weights are then compared to check for convergence. If the weights converge, the converged set of weights is used to divide the image into superpixels.

Growing community of inventors

Cambridge, United Kingdom

Minh-Tri Pham