Method and device for three-dimensional reconstruction of a scene

A method for three-dimensional reconstruction of a scene comprising at least one object, the method comprising the steps of:

• providing at least one camera for recording a plurality of images of the scene including the object,
• recording a first sequence of first images of the scene from a first perspective relative to the scene, and
• recording a second sequence of second images of the scene from a second perspective relative to the scene, the first and second perspectives being different from one another,
• determining a plurality of first image areas within the first images and determining a plurality of second image areas within the second images,
• identifying a plurality of correspondences between the first and second image areas, and
• reconstructing the scene based on the correspondences between the first and second image areas,
• wherein the correspondences are identified by matching a parameterized function to each image area in order to obtain a plurality of first and second function parameters representing the first and second image areas, and by comparing respective first and second function parameters,
• wherein the first and second sequences each comprise a plurality of first and second images so that a spatial position of the image areas and any movement over time of the image areas are used in order to identify the correspondences, and
• wherein a differential volume between surfaces determined from the first and second image areas is used as a similarity measure for comparing the respective first and second function parameters.

A method for three-dimensional reconstruction of a scene comprising at least one object, the method comprising the steps of:

• providing at least one camera for recording a plurality of images of the scene including the object,
• recording a first sequence of first images of the scene from a first perspective relative to the scene, and
• recording a second sequence of second images of the scene from a second perspective relative to the scene, the first and second perspectives being different from one another,
• determining a plurality of first image areas within the first images and determining a plurality of second image areas within the second images,
• identifying a plurality of correspondences between the first and second image areas, and
• reconstructing the scene based on the correspondences between the first and second image areas,
• wherein the correspondences are identified by matching a parameterized function to each image area in order to obtain a plurality of first and second function parameters representing the first and second image areas, and by comparing respective first and second function parameters,
• wherein the first and second sequences each comprise a plurality of first and second images so that a spatial position of the image areas and any movement over time of the image areas are used in order to identify the correspondences, and
• wherein a Euclidian distance between function parameters representing first and second image areas is used as a similarity measure for comparing the respective first and second function parameters.

A device for three-dimensional reconstruction of a scene comprising at least one object, the device comprising at least one camera for recording a plurality of images of the scene including a first sequence of first images of the scene from a first perspective relative to the scene, and a second sequence of second images of the scene from a second perspective relative to the scene, the first and second perspectives being different from one another, comprising at least one evaluation unit for determining correspondences between the first and second images, and comprising at least one calculating unit for reconstructing the object, wherein the evaluation unit is configured for determining a plurality of first image areas within the first images and for determining a plurality of second image areas within the second images, the evaluation unit further being configured for identifying the plurality of correspondences between the first and second image areas, and wherein the calculating unit is configured for reconstructing the scene based on the correspondences between the first and second image areas, wherein the correspondences are identified by matching a parameterized function to each image area in order to obtain a plurality of first and second function parameters representing the first and second image areas, and by comparing respective first and second function parameters, wherein the first and second sequences each comprise a plurality of first and second images so that a spatial position of the image areas and any movement over time of the image areas are used in order to identify the correspondences, and

• wherein the function parameters are determined by linear optimization using a computation path based on physically motivated simplifying assumptions.