EP3574420A1

EP3574420A1 - Method, computing device, and computer program for providing a mounting edge model

Info

Publication number: EP3574420A1
Application number: EP18701185.3A
Authority: EP
Inventors: Claudia NIEUWENHUIS; Oliver Schwarz
Original assignee: Carl Zeiss Vision International GmbH; Carl Zeiss AG
Current assignee: Carl Zeiss Vision International GmbH; Carl Zeiss AG
Priority date: 2017-01-27
Filing date: 2018-01-26
Publication date: 2019-12-04
Also published as: US10859859B2; US20200018994A1; EP3355214A1; WO2018138258A1; CN110447029A; CN110447029B

Abstract

The invention relates to a method for providing a mounting edge model, to a corresponding computer program, and to a corresponding computing device. In order to provide the mounting edge model, a plurality of data sets (20) are first provided, each of which describes a course of a mounting edge. Such data sets can be obtained from tracer data for example. On the basis of the data sets, a parametric mounting edge model (22) is then provided. (Figure 2)

Description

Description Method, computing device and computer program for providing a

Frame edge model

The present application relates to methods and apparatus for providing a frame edge model of a spectacle frame and corresponding computer programs and computing devices. A frame border of a spectacle frame is the part of the

Spectacle frame, in which then a spectacle lens is fitted, wherein a first portion of the frame edge for a left lens and a second portion of the frame edge for the right lens is provided. In the context of this application is under a

The edge of the frame is not just the border of the frame surrounding the lens

Eyeglass frame, but also the corresponding sizes in stunned glasses or partial glasses, e.g. Eyeglass frame inner edge, support disk outer edge or eyeglass lens outer edge. A frame edge model is understood to be a mathematical and / or geometric description of the frame edge of a spectacle frame or at least of the first or second portion of the frame edge.

In order to correctly fit the lenses in a spectacle frame, it is firstly necessary to determine so-called Zentrierparameter so that the optical centers of the glasses can be brought into line with the visual axes of the corresponding eyes, so as. For example, to know the information about the pupil distance and the information about the height of the pupils with respect to the spectacle frame. On the other hand, it is necessary to know the predetermined by the eyeglass frame supporting edge of the lens in which the lens is to be recorded.

In addition to the information of the pupil distance and the information about the height of the pupils in relation to the spectacle frame, the following parameters are subsumed under the term centering parameter: monocular pupillary distance PD, corneal vertex distance HS according to reference point requirement and / or after eye pivot point requirement, monocular center point distance, center point coordinates , Disc distance, centering point decentration, disc height and width, disc center distance, prescription lens a,

Bracket angle ß, grinding height. The centering parameters are determined regularly by an optician. Important

Centering parameters are z. For example, these are defined in the standard EN ISO 13666: 2012 (D / E) and can be determined by an optician and a subject standing or sitting opposite one another, with the test person setting the version of his choice with a pane of glass enclosed in it. The subject is asked to look into the distance, and the optician then draws

Look on the glass or on a contact foil with a cross to see through how he sees it when looking at each other. This cross (centering cross) then determines the position of the optical center of the lens to be inserted into the socket. This procedure is performed individually for each of a subject's eyes. The distance between the centering crosses determined in this way is the pupillary distance PD.

For the Zentrierparameterbestimmung but now automated

Measuring systems used. Such a measuring system is z. In WO 01/84222 A1

described. This system includes a height adjustable mounted on a digital video camera, the lens is arranged together with a mirror and a light source in the region of the front surface of the housing. In particular, the system makes it possible to measure distances and to measure the dimensions required for embedding

Eyeglass lenses are to be considered. In the system there is one with the digital one

Video camera connected computer, taken from the image of a spectacle wearer with a

Spectacle frame and with a measuring bracket fixed to the spectacle frame. Centering parameters for the spectacle frame determined by means of image evaluation.

In some of these measuring systems, one or more images of a person's head, which carries the respective spectacle frame, are recorded simultaneously or sequentially, and the centering parameters are determined by image analysis from the recorded images. The algorithms used are based, among other things, on the captured images

To recognize frame edges of the spectacle frame.

Such devices and methods are described for example in DE 10 201 1 1 15 239 A1. In the approach described there, a tracer record is present for a spectacle frame currently to be measured on a person's head. A tracer is according to the above-mentioned DIN EN ISO 13666 a device for accurate, often three-dimensional measurement of the glass shape of a spectacle frame for the edge of the form of the glass, so a device which measures the frame edge of a spectacle frame. The tracer can thereby determine the tracer dataset either the frame edge of the eyeglass frame or the glass edge of a support disk which is shaped to in the frame edge to be used, measured. In both cases, the tracer record then describes only the frame edge of the spectacle frame, but not other parts of the spectacle frame. Tracer records, after being generated by the tracer, typically contain information identifying the source of the record (ie, the tracer used). However, this information may also be removed as the tracer record is processed. The approach of DE 10 201 1 1 15 239 B4 is then based on finding a contour described by the tracer data record in image recordings. Conventional methods of image analysis such as pattern matching or template matching are used, the latter being described, for example, in the article "Template Matching" on Wikipedia of January 17, 2017. Pattern search, also referred to as pattern-based search or pattern matching, is a designation for symbol-processing Method which identifies corresponding structures in the recorded images on the basis of a predetermined pattern-in this case specified by the tracer data set-Through the use of the tracer data set, the recognition of the frame edge in the images is determined according to the statements of DE 10 201 11 15 239 B4 compared to others

Approaches facilitated. A disadvantage of this approach, however, is that the tracer record and thus the patched glasses must be known to find the frame edge in the captured images. Typically, however, an optician will not generate tracer data from a socket until a customer has already selected the glasses for purchase. Since the generation of the tracer data is time consuming, they become

Usually also generated only when the customer has already left the optician business, otherwise a long wait for the customer would arise. On the other hand, the determination of the centering parameters in the usual workflow is carried out by an optician immediately after the selection of the spectacle frame, so that the customer can already pick up the finished spectacles at the next visit.

Thus, therefore, for the implementation of the method of DE 10 201 1 1 15 239 B4 the usual workflow of the optician must be changed, which also means waiting times for the customer when visiting the optician. Moreover, this approach is also because of that

cumbersome, as this every optician to each version must have the respective Tracer record in stock or must create before the determination of Zentrierparameter. Finally, the tracer datasets are at least preferably required as three-dimensional datasets (ie, datasets describing the three-dimensional shape of the rim of the frame), while some conventional tracers only generate two-dimensional datasets describing the shape of the rim of the frame in a plane. Szu-Hao Huang et al., "Human-centric design personalization of 3D glasses frame in markerless augmented reality," Advanced Engineering Informatics 26 (2012) pp. 35-45 describe a method of virtual fitting a spectacle frame Spectacle frame used by means of a

Main component analysis is created on a variety of stored spectacle acquisition models. The spectacle frame models are present as polygonal meshes which correspond to CAD (Computer Aided Design) datasets of spectacle frames. US 2015/055085 A1 discloses a virtual fitting on of glasses, to which a model of the glasses is used. Captured head images are analyzed with a principal component analysis.

Wu Chenyu et al., "Automatic Eyeglass Removal from Face Images", IEEE Trans, on Pattern Intelligence and Machine Intelligence, Vol. 26, No. 3, pp. 322-336 discloses a method for removing a pair of spectacles from a captured image of a head. It will be

various learning methods and a parametric model used for a spectacle frame. The parameters of the parametric model used there, which describe the shape of the version, are assumed to be normally distributed.

Diana Borza et al., "Eyeglasses Lens Contour Extraction from Facial Images Using an Efficient Shape Description", Sensors 2013, 13, pp. 13638-13658 discloses a method of accurately matching a position of spectacles and the size and shape of a spectacle frame from an image capture A 2D model of a spectacle frame based on Fourier descriptors is used.

Against this background, it is generally an object of the present application to provide methods, computing devices and computer programs for providing a frame edge model, wherein the frame edge model is to cover a plurality of spectacle frames, so that, for example, for the above-explained image analysis for the purpose of

Zentrierparameterbestimmung not a separate Tracer record must be created for each spectacle frame.

In the above-cited publication by Szu-Hao Huang et al. lie models of

Spectacle frames, and from these, a model of an entire spectacle frame is obtained by a principal component analysis. Such a model is explained for the above Fassungsranderkennung only conditionally suitable because in the model, for example, data regarding the eyeglass temple are included, which have nothing to do with the detection of the frame edge, but still need to be processed with. It is therefore a first object of the invention to provide a method for providing a

To create a margin model, which is used for the above identification of a

Border edge as described in DE 10 201 1 1 15 239 B4 is better suited than the model, which the method of Szu-Hao Huang et al. supplies. For this purpose, a computer-implemented method according to claim 1 is provided according to a first aspect of the invention. Dependent claims 2-12 define others

Embodiments of the method of the first aspect.

In the above-cited paper by Wu Chenyu et al. is based on a Gaussian distribution of parameters. However, shapes of real frame margins do not necessarily vary according to a Gaussian distribution. If a Gaussian distribution is assumed in the identification of the edge of the frame according to DE 10 201 1 1 15 239 B4, this leads to the result that the method results in a solution which is not of sufficient quality, since many forms of

Frame margins are not included in the model and thus can not be detected.

It is therefore a second object of the invention to provide a method for providing a

To create a perimeter border model that is better suited for identifying a perimeter border than a model with a fixed Gaussian distribution of parameters.

For this purpose, a computer-implemented method according to claim 13 is provided according to a second aspect of the invention. Dependent claims 14-18 define others

Embodiments of the second aspect. The method for providing a model of a spectacle frame according to Szu-Hao Huang et al., Which uses inter alia a principal component analysis, is relatively complicated mathematically. Proceeding from this, it is a third object of the present invention to propose a simplified method for providing a frame edge model. For this purpose, a computer-implemented method according to claim 19 is provided according to a third aspect of the invention. Dependent claims 20-22 define others

Embodiments of the third aspect. In the method according to Szu-Hao Huang et al. It is assumed that 3D data from spectacle frames are in a form suitable for principal component analysis. When data describing the frame edge of an eyeglass frame comes from a variety of sources, such as various meters, e.g. with different tracers from different socket manufacturers or opticians, the data are in various formats and may also be noisy due to the measurement process, i. Have fluctuations that are due to the measurement.

Based on this, it is a fourth object of the present invention to provide a method for providing a frame edge model in which data from different sources and / or noisy data are better processable.

For this purpose, a computer-implemented method according to claim 23 is provided according to a fourth aspect of the invention. Dependent claims 24-32 define others

Embodiments of the fourth aspect.

Further subclaims define image analysis methods that use a frame edge model that has been constructed according to one of the above aspects. In addition, a

Computer program according to claim 40 by performing one of the above defined

Method and a corresponding computing device according to claim 41 provided.

The frame edge models provided with the above methods can be used, in particular, for centering spectacle lenses, for enclosing spectacle lenses and for producing spectacle lenses. For this, methods according to claims 42, 43 and 44 are provided.

According to the invention, a method is provided for providing a frame edge model

provided, comprising:

Providing a plurality of data sets, each describing a course of a frame edge of a respective eyeglass frame, and

Providing a parametric frame edge model based on the plurality of

Records. The course of the frame edge for a spectacle lens or for both spectacle lenses (left and right spectacle lens) is to be understood as a profile of a frame edge, ie for one or both of the sections mentioned above. A parametric rim model is a rim model that has one or more modifiable parameters. By changing the parameter or the parameters, the geometry of the frame edge described by the frame edge model then changes, for example with regard to size or shape or with regard to the alignment of first and second sections for the two spectacle lenses. This is similar to parametric modeling in CAD (Computer Aided Design) technology (see http://www.blien.de/ralf/cad/db/para_mod.htm Stand

27.01.2017), wherein in the present application the parameters are not limited to dimensions and generally changes in geometry e.g. in terms of size and / or shape by changing the parameters are possible. Parameters of such a model are thus variables by means of which the frame edge described by the frame edge model can be changed. A parametric frame edge model in the sense used here can also be a so-called parameter-free model. Parameter-free models are described, for example, in the Wikipedia article "Parameter-Free Statistics." These are

Models whose model structure is not determined a priori, but is determined based on the large number of data sets. They also have parameters, but the nature and number of these parameters are also not fixed a priori. They are also applicable if the necessary prerequisites for certain parametric models (e.g., for Gaussian distributions) in the data are not met.

By providing the parametric frame edge model based on the plurality of datasets, a plurality of datasets are incorporated into the model. This causes the parametric frame edge model to be applicable to a variety of different spectacle frames, even to spectacle frames that are not directly described by a record of the plurality of datasets. The parametric frame edge model can then be adapted to the frame edge of a specific spectacle frame by the parameter or parameters of the parametric frame edge model.

According to the first aspect of the invention, optionally also with other aspects of the invention, tracer data records of different spectacle data of spectacle frames are used, which were determined via a corresponding multiplicity of spectacle frames. Using Tracer datasets provides data only for the required frame margins, and not for other parts of a frame such as brackets. Unlike the in the publication Szu-Hao Huang et al. In models used, tracer records contain only sequences of points in space or in the plane that describe the socket outline. They are not meshes, as in the above-cited publication Szu-Hao Huang, which also contains edges and surfaces in addition to points. This can simplify the creation of the frame border model as well as a subsequent image analysis, since the

Frame contour does not have to be calculated first from a model of an entire spectacle frame. In addition, an optician often does not have polygon meshes of spectacle frames as in the method of Szu-Hao Huang et al. before, for example in the form of CAD datasets, on the eyeglass frames, since these are the trade secret of the frame manufacturers. On the other hand, most opticians have a tracer that can easily retrieve a variety of tracer data sets. Thus, the method according to the first aspect of the invention can also be easily performed by an optician without relying on the provision of records by third parties such as socket manufacturers.

In other embodiments of the second, third and fourth aspects may alternatively be CAD data (Computer Aided Design) of the spectacle frames of

Spectacle manufacturers are used, which were used by the spectacle makers to manufacture or design the spectacle frames. Preferably, the data sets are present as 3D data records, i. they describe the spatial course of the respective border of the frame. This results in more accurate models, since all three spatial directions can be determined. However, it is also possible to use 2D datasets which describe the course of the frame margins only in one plane. In particular, by using the plurality of data sets, it can be achieved that the parametric frame edge model does not allow "untypical" frame margins, e.g., no frame margins, the cut-off lenses, non-symmetrical ones

Eyeglass lenses, or lenses with glass edges, which have a bulge or invagination, so a high curvature correspond.

The data sets are preferably present as 3D data sets, i. Data sets that describe the course of the respective border of the frame in three dimensions. In particular, the

Datasets as point clouds with a variety of measurement points, which on the

At the edge of the frame are present. If only 2D datasets are present (ie, descriptions of the frame edge in a plane), for example, because a tracer provides only 2D datasets, then by approximating eyeglass lenses bordered by the border of the frame, as Planes or paraboloid or similar 3D models, from which a three-dimensional dataset can be generated. For example, in one embodiment typical values for centering parameters such as the pretilt angle (according to 5.18 of DIN EN ISO 13666: 2012) and the mounting disk angle (according to 17.3 of DIN EN ISO 13666: 2012) can be determined for this purpose. This determination can be made from ordering statistics for spectacle lenses by averaging, medians, or similar characteristics of pretilt angle distributions and

Socket angle can be determined over a variety of lenses. Alternatively, it is also possible to simply assume typical values for these parameters. In addition, a symmetry of two levels, paraboloid or similar 3D models in which the lenses are, to each other to be accepted. With pre-tilt angle,

Bracket angle and the assumption of symmetry then yields the location of the two planes (e.g., defined by normal vectors on the planes and points of appeal for the plane)

Normal vectors) into which the 2D data sets are then projected. Even if 3D data sets and corresponding 3D frame edge models are preferred, the method according to the invention also makes it possible to provide a 2D model based on 2D data sets. This may also be beneficial in some applications. Thus, 2D datasets are smaller (less data) than corresponding 3D datasets, and because of their lower dimensionality, 2D models have fewer parameters in most cases, so optimization methods using the 2D model, eg, when the model approaches a frame edge in an image is adjusted as described later, have a lower complexity. For some applications, such as identifying a frame border in a single image, only one 2D model is needed. Providing the parametric frame edge model may, in one variant, include calculating the parametric frame edge model from the plurality of datasets. In this case, the large number of data sets is used as the basis for a new creation of a parametric frame edge model. An example of this is machine learning based on the large number of data records. Machine learning is outlined in the Wikipedia article "Machine Learning" of January 17, 2017, and is a generic term for artificially generating knowledge from experience. In the present example, the plurality of data sets represents experience, and the parametric perimeter model represents the generated knowledge dar. The mentioned Wikipedia article also gives an overview of various approaches and algorithms that can be used here. One possible approach for calculating the parametric frame edge model from the plurality of datasets is to use a principal component analysis. The Principal Component Analysis (PCA) is a multivariant statistics method that generates a large number of so-called "main components", the data sets being then calculated as the sum of an average value and a value

Linear combinations of these main components can be displayed. For this purpose, for example, the mean value of the plurality of data sets is first determined and subtracted from the data records. From the result, the main components are determined. The main components are orthogonal to each other. This means that the representation of each record is unique, i. that the respective linear combination is unique

Having coefficients. These coefficients of such a linear combination of

Principal components then represent the parameters of the parametric frame edge model. An advantage of the principal component analysis is also that it provides the dimension of the

Model space can be minimized so that as little information is lost. The dimension of the model space is equal to the number of parameters of the model and in the case of the principal component analysis the number of main components (one coefficient for each

Main component). For example, principal component analysis is described in the article "Principal Component Analysis" on Wikipedia of January 17, 2017. The principal component analysis can also be used to remove noise or other artifacts similar to filtering as described below

Principal components resulting from the above principal component analysis describe typical curves of barrel margins, their linear combinations as

Approximations of the original ones defined by the plurality of data sets

Can serve frames edges. The coefficients of the main components for a concrete

The border of the frame is determined by projecting a curve defined by the respective edge of the frame onto the main components. By omitting major components with little importance (small coefficient, corresponds to a small variance of the respective

Record the respective frame edge along this main component), the noise components can be eliminated. Therefore, the representation of the frame edges by means of principal component analysis by omitting such main components and a

Fulfill filter function.

Alternatively, an Independent Component Analysis (ICA) may be used, as described, for example, in the article "Independence Analysis" on Wikipedia of January 20, 2017. This type of analysis yields Also components that as a linear combination adds up to a mean value

Can represent records. However, the components are not orthogonal here, so there are at least partially several possibilities (i.e., multiple sets of coefficients) to represent a data set. Again, the coefficients of the linear combination, the

Parameters of the parametric borderline model.

In the second aspect of the invention, optionally also in the first or fourth aspect of the invention, providing the parametric frame edge model comprises calculating probability distributions for parameters of a predetermined parametric one

Perimeter border model based on the large number of data records. In this variant, therefore, no parametric frame edge model is created completely new, but at first a parametric frame edge model is specified. A simple example of such a model is two planes symmetric to a median plane, each plane

approximately describes the position of a spectacle lens. Instead of the levels, higher-order forms are also possible, for example square or cubic splines with associated parameters. Splines are functions which are piecewise composed of quadratic, cubic or functions of even higher order such that the transitions are continuous or continuously differentiable. Splines are described, for example, in the Wikipedia article "Spline" of Jan. 17, 2017. In addition, the frame margins can then be further described by parametric geometric shapes within the planes, a simple example of which are ellipses, for example, by the length of the major semiaxes , usually denoted by a, and the small semiaxes, usually denoted by b, can be characterized as parameters, where the major axis formed by the major semiaxes, the axis passing through the foci of the ellipse, and the minor axis, which is formed of two small semi-axes, is perpendicular to it.

Probability distributions can then be created for these parameters. For this purpose, the predetermined parametric frame edge model can for example be adapted (fitted) to each of the data records in order to obtain respective parameter values for the data record. Such fitting can be done, for example, with the least squares method (see Wikipedia article "Least Squares Method" of January 17, 2017) or with other conventional Fit methods, thus obtaining parameter values for the plurality of records. which can then be statistically evaluated.

For example, mean and standard deviation can be specified for each of the parameter values. By this procedure, any distributions of the Parameter values are collected and analyzed, and there is no restriction on a given distribution such as a Gaussian distribution. The statistical evaluation can be done in a simple manner by means of a histogram, as described for example in the Wikipedia article "Histogram", as of 1 December 1, 2017.

These probability distributions may then be used in a later use of the model to more quickly adapt the parametric frame edge model to a particular frame edge (eg, frame edge identified in an image) by using fitting methods that take into account the statistical distribution. Such optimization methods, the

Include probability distributions over the parameters to be optimized include, for example, Markov Chain Monte Carlo (MCMC) methods. Such optimization methods are described, for example, in Christopher Bishop, "Pattern Recognition and Machine Learning," Springer, 2006, ISBN 10: 0-387-31073-8, which estimate optimal parameters of a model by randomly choosing the individual parameters according to their probability distribution and then how well the resulting concrete model (ie the model with the so chosen

Parameters) on the data. This is repeated very often, e.g. more than 10,000 times or more, with the number of repetitions also depending on the particular method chosen and its implementation. As a result one obtains a probability distribution over model parameter sets (ie over all parameters together), from which e.g. the parameter set with the highest probability can be selected as "best fit", i.e. as a fitted model, or alternatively, for example, the mean or median of the distribution: In particular, probability distributions of parameters of the above-mentioned parameter-free models can also be determined in this way.

In the third aspect of the invention, optionally also in the first or fourth aspect of the invention, providing the parametric frame edge model comprises providing the plurality of datasets or a plurality of datasets derived from the plurality of datasets as the model, one parameter of the parametric

Perimeter frame then indicates a record of the plurality of records or the plurality of other records. In this case, the data records or other data records derived therefrom are used directly as a model, and a parameter indicates a respective data record, for example a number of the data record. In this way no elaborate processing of the data sets, for example by means of principal component analysis, is necessary in order to create the model. In such a parametric frame edge model For example, to adapt to a frame edge found in an image, the data set closest to the frame edge in the image can then be selected (again, for example, by means of the least squares method).

Derived additional records are records that are derived from the multitude of records using mathematical operations such as those explained below

Transformations are obtained.

In the fourth aspect of the invention, optionally also in the other aspects of the invention, providing the parametric frame edge model includes transforming the plurality of datasets. In this way, a plurality of transformed data sets are obtained. A transformation of a data record is to be understood as meaning the application of an operation to the data record which makes the data record in another form, i. a respective further data set, transferred, wherein the information of the data set about the geometry of the frame edge remains at least approximately. Such transformations can serve in particular for normalization or reparameterization in order to bring the multiplicity of data records into a format corresponding to one another, as it were. A

Reparameterization is a change of parameters and / or values (e.g., coordinates) that describe the frame edge, such as a change of a used coordinate system or a change of a base such as a coordinate. a reference point of a used

Coordinate system. Specific examples are explained below.

For this purpose, the transformation in a first variant of the fourth aspect of the invention, optionally also in other aspects of the invention, comprise a coordinate transformation. In a coordinate transformation, data present in a first coordinate system are transferred to a second coordinate system. By means of such coordinate transformations, data sets present in different first coordinate systems can be transformed into a common second coordinate system, and / or the data records can be converted into a coordinate system which is more suitable for later processing. Thus, in Cartesian coordinates ((x, y) in 2D or (x, y, z) in 3D) existing data sets in polar coordinates (φ, r) (in 2D) or spherical coordinates (φ, ψ, r) (in 3D ) are transferred. In addition, by means of a coordinate transformation, the coordinate systems of the data sets can be brought to mutually corresponding reference points. For example, the center of a portion of the frame edge can serve as a reference point for a spectacle lens, wherein the center of gravity can be used in particular as the center of gravity. In other embodiments, a midpoint between the

The peripheral edge portions serve as a reference point for the two lenses. In addition, by means of the coordinate transformation, a rotation to align the data sets in the same direction. Thus, for example, a horizontal direction, which connects the centers according to the box system (5.4 DIN EN ISO 13666 from 10/2013) of right and left lenses in the respective spectacle frame, be defined as an x-axis, and directions perpendicular thereto as y- and z -Axis. For example, an average of the two normals of the

Eyeglass lenses approximating planes as z-axis (for records that describe both portions of the frame edge, for one of the lenses, such as the left) and a direction perpendicular thereto are defined as the y-axis, the lenses are defined by the records. Rotations then align all the records of the plurality of records into that coordinate system.

In addition, in a second variant of the fourth aspect of the invention, optionally also in other aspects of the invention, the transformation may also include resampling of the frame margins defined by the datasets, in which the geometries defined by the datasets are in polar coordinates, for example The reference point can be selected as explained above, in this way the data sets are unified in terms of their representation, which is the following further

Processing, for example, the application of machine learning method as explained above, simplified. Such resampling is an example of reparameterization.

In a third variant of the fourth aspect of the invention, optionally in other aspects of the invention, the transformation comprises filtering the data records with a low-pass filter in order to smooth out the frame edge geometries defined by the data sets. Filtering is understood to mean a transformation of data sets in the above-mentioned sense by which noise or other measurement artifacts that occurred during the creation of the data sets are reduced or eliminated, ie. a partial aspect of the signal is partially or completely suppressed. Examples of such filtering include Gaussian filtering (see the Wikipedia article "Gaussian Filter" as of January 18, 2017) or Fourier filtering. <br /> <br /> In a Fourier filtering, a Fourier transform is made on the data sets from the spatial domain in performed the frequency space, and the

Low-pass filtering then attenuated high-frequency components of the resulting frequency spectrum, ie reduced. For this Fourier filtering, two dimensions of the data of points of the data sets (for example, x and y coordinates of points of the data sets) can be represented as complex numbers by the x coordinate indicating the real part and the y coordinate the imaginary part). For three-dimensional datasets, the third dimension (in In this case the z-coordinate) are filtered separately. In this case, the periodicity of the Fourier base can be utilized for the Fourier transformation. Also a Gaussian filtering in

Frequency space is possible, where the 3 dimensions can be filtered independently and the periodicity can be exploited by a periodic

Continued at the beginning and at the end.

After that, an inverse transformation (inverse Fourier transformation) can take place. In other embodiments, a further processing, for example, the above-mentioned calculation of the model in the frequency space and take place, for example, only after a back transformation.

Filtering can then provide the parametric

Margin frame model can be facilitated because such measurement artifacts or noise then need not be considered in the frame edge model.

In a fourth variant of the fourth aspect of the invention, optionally in other aspects of the invention, the transformation comprises an approximation of the data sets by parametric curves. Parametric curves are curves that are described by parameters. An example of such parametric curves are the already mentioned splines. These curves are matched to each of the data sets in such embodiments, again using the least squares method, for example, so that ultimately each data set is then characterized by the respective parameters, such as spline coefficients. This leads to a smoothing, as smooth curves are defined by the datasets defined points.

In this way, the data sets are reduced to a representation with fewer degrees of freedom (just the parameter values), which can simplify computational effort in subsequent processing to provide the parametric frame edge model. Such fitting of parametric curves is another example of one

Re-parameterization.

An example of such subsequent processing is image analysis. For such an image analysis, a method is provided according to the invention, comprising:

Providing at least one image of at least a portion of a head carrying a spectacle frame, Providing a parametric frame edge model created by one of the methods discussed above, and

Analyze the at least one image using parametric

Frame edge model.

The analysis of the image is to be understood as a processing of the image which helps to ultimately obtain information from the at least one image.

An example of such information is the detection of the frame edge in one or more images, which are preferably taken from different directions. This detection is also called segmentation. Segmentation is generally understood as the identification of specific image components. It is considerably facilitated if there is a frame edge model whose parameters have to be adapted, since only special meaningful curves of the curve can be represented in the picture. Without such a model it is difficult to find a curve that matches the characteristics of a

Spectacle edge or two lens edges is sufficient, e.g. the symmetry of the two glasses detected in the image is naturally given in a 3D frame model. In contrast, it is difficult to achieve the symmetry of the two lenses in an image segmentation process, because perspective distortions in the image of the 3D frame edge into the image lead to asymmetrical 2D contours. Centering parameters can then be determined by the detection of the edge of the frame, as explained in the introduction with reference to DE 10 201 11 15999 B4.

Another example of information is therefore the concrete provision of the

Parameters of the parametric frame edge model, which are detected by a detection of the

Fassungsrandes can be obtained in one or more images and adjusting the frame edge model to the detected frame edge. When using multiple images, they are preferably taken from multiple directions, which increases the accuracy of the detection of the frame edge. From these parameters, further parameters can be calculated, which include a translation, rotation, scaling, etc. of the

Frame border model to adapt to the picture describe. If the concrete parameters of the frame border model are known, the position of the frame in the room is determined.

According to the above-described definition of the frame border also for stunned glasses and spectacles with partial versions with the location of the frame edge and a representation of an edge of a corresponding to be inserted into the socket lens is known and can thus be determined. Thus, also such a representation of an edge of a spectacle lens as described above for the detection of the frame edge can be determined.

By the representation of the edge of a spectacle lens, the invention means a

Representation of the base of the spectacle lens specified in the standard EN ISO 13666: 2012 (D / E) in section 13.1.

A representation of the edge of a spectacle lens is a data set from which the three-dimensional course of the wearer's wear rim facing away from the spectacle wearer can optionally be unambiguously determined on the basis of additional variables describing the spectacle lens. A representation of the edge of a spectacle lens can, for. B. be surrounded by the projection of the edge of the lens in the image plane of an image sensor of an image pickup device surface in which the lens is imaged for image capture.

In the case of full rim eyeglasses, the wearing edge of a spectacle lens corresponds to one

Bnllenfassungsinnenrand. In the case of a partial rim eyeglass, the edge of a spectacle lens which corresponds to the inner edge of the eyeglass frame and the outer edge of the eyeglass lens which is not connected to the eyeglass lens frame are to be understood as the support rim of a spectacle lens. The supporting edge of a spectacle lens in rimless glasses is the glass outer edge.

From this, the frame-specific centering parameters can then be determined, e.g. the pre-tilt or the bracket angle. If, in addition, the position of the eyes in the room is known, the centering parameters can be completely determined.

Accordingly, a method for centering a spectacle lens in a

Spectacle frame provided, characterized in that in a step (i) to the

Spectacle lens centering parameters are determined, wherein determining the centering parameters comprises determining the representation of the edge of the spectacle lens with the method as described above and in a step (ii) the spectacle lens is centered with the centering parameters determined in step (i) in the spectacle frame.

Furthermore, a method for inserting a spectacle lens into a spectacle frame is provided, characterized in that in a step (i) to the spectacle lens

Centering parameters are determined, wherein determining the centering parameters, determining the representation of the edge of the spectacle lens with the method as described above includes; and in a step (ii) grinding the spectacle lens based on the centering parameters determined in step (i) for an arrangement in the spectacle frame.

Finally, a method for producing a spectacle lens, characterized by the method step einschieifen the spectacle lens as described above.

Thus, by means of the frame edge models according to the invention, centering, bending and production of a spectacle lens can take place without having to change the workflow of the optician, as is the case with the initially cited DE 10 201 1 1 15 239 B4.

After the detection of the frame border in the image by means of the frame border model also eyeglass properties can be determined from the image as information, for. B. color, shape or size of the glasses. If the specific parameters of the 3-D frame model have been known or previously estimated, camera parameters from the identification of the frame edge in one or more images can also be obtained as information, e.g. the extrinsic parameters (one location) of a used calibrated camera or the intrinsic and extrinsic

Camera parameters of an uncalibrated camera together. This may be particularly useful in applications where mobile devices such as smartphones or tablets are used for image capture whose location in space is not fixed. Alternatively, both the parameters of the 3D frame model and the camera parameters can be determined simultaneously, for example by solving an optimization problem. Among the camera parameters of a camera are extrinsic parameters, ie parameters that do not arise from the properties of the camera itself, such as the relative orientation in space, ie the relative orientation of the optical axis of the camera to other objects, as well as the intrinsic parameters of the camera An example of such intrinsic parameters are parameters that define how the coordinates of a point in space in a coordinate system referenced to the camera are converted into the coordinates of the pixel of that point located in the image plane of the image sensor. A detailed description of the calibration of cameras, in which such camera parameters are determined, can be found for. See, for example, on page 8 in the textbook "Multiple View Geometry in Computer Vision" by Richard Hartley and Andrew Zisserman, 2nd Edition, Cambridge University Press 2004. Information obtained can also be used to modify or improve the image. Thus, after determining the parameters of the frame border model, artifacts may be removed from the image, eg, a shadow cast on the head by the eyeglasses in the image, by projecting the frame edge upside down depending on the position of a given one

Light source and subsequent inpainting (repairing the texture) in this area. Likewise, the frame can be removed from the images by projecting the frame model into the images according to the camera imaging properties and removing the resulting contour from the images and repairing it with inpainting. Using the parametric boundary model in the analysis can simplify the analysis, examples of which are discussed below.

The provision of the at least one image may, for example, comprise taking one or more images with one or more cameras arranged in a fixed geometry. This allows for multiple cameras with fixed geometry or a single camera that performs a given (and thus known) or any motion relative to the head to pick up the head from different directions

in particular a triangulation of the head, i. determining a three-dimensional contour of the head. Such image acquisition is described for example in the aforementioned DE 10 201 1 1 15 239 A1.

For image analysis using the parametric frame model, there are

various possibilities. Thus, the parametric frame model can be used to segment the at least one image, in particular to identify the

Fassungsrandes in the at least one picture to facilitate.

It should be noted that the segmentation of the frame edge has two main problems. On the one hand, edges of the frame may be bad or even unidentifiable (eg rimless frames or frames with colors that are similar to skin color). This can lead to edges being only partially identified. On the other hand, edges in the image can be incorrectly identified as border edges (eg if the outer edge of the glasses shows the hair or the background, strong edges to the skin color of the face, which are easily confused with the edge of the frame). In order to close incomplete edges correctly and / or to exclude false edges, the parametric frame model can be used, since the parametric frame model does not allow for untypical frame margins (eg cut glasses, not symmetrical glasses, glass edges with bulge or

Invagination, so with high curvature, etc.)

In such a segmentation, in one embodiment, in the case of a 2D frame edge model, the space of possible solutions to that in the image may increase

identifying frame border on the through the parametric frame model

restricted possibilities, i. it is searched from the outset only for frame margins in the image, which are describable by the parametric frame model. In a 3D frame edge model, in one embodiment, camera calibration data is used. The camera calibration data describes a mapping rule of (three-dimensional) points in space to an image captured by a camera used to capture the at least one image. Using this camera calibration data, the parametric frame edge model can be projected into the image and used to identify edges of the frame border. A projection is understood to be an image of a 3D model in a 2D image. If no calibration data should be present, in some embodiments, the frame edge model may be projected into the at least one image, e.g. by the

above-mentioned mapping rule is optimized or estimated in a first step. The at least one image may also comprise a plurality of images taken from different directions, e.g. as stereo image recording. Using triangulation, a 3D image can be calculated therefrom according to computation steps known from stereoscopic image recording, in which the 3D frame edge model can then be used to identify the frame edge. Identifying the frame edge in the image will then also reveal the parameters of the frame perimeter frame model for that frame edge, e.g. when segmenting and determining the parameters in one

common optimization process takes place. The parametric rim model with the parameters thus determined can then be used e.g. be used to determine Zentrierparametern, as will be explained briefly below.

Lying as explained above probability distributions for the parameters of

parametric edge model, these can also be used, for example, in the image analysis of an average of the parameter

is assumed. This value can then be improved sequentially, for example within an MCMC method, in which a parameter set corresponding to the distributions of the individual parameters is randomly selected and then the distance of the images projected into the image Sizing edges with the image detected in the image, for example by means of an edge detector edges is calculated. This procedure is repeated very frequently, leading to a distribution over parameters with associated distances. At the end the parameter set with the smallest distance value can be selected as best.

Alternatively, the mean of the parameter may be used as an initial value in an optimization problem, e.g. within a gradient descent process.

To improve the identification of the frame border can be at another

Embodiment, a deviation of a frame edge determined from the at least one image is also included in the parametric frame edge model,

For example, in the form of a penalty term in a cost function for determining the

Frame border in the at least one picture. The farther the larger a currently identified frame edge is from the parametric frame edge model, the larger it becomes. This therefore constitutes an addition in existing optimization methods, for example in an MCMC method or gradient descent method, graph cut method, Markov Random Field optimization method and the like. such

Optimization techniques are described, for example, in Christopher Bishop, "Pattern Recognition and Machine Learning," Springer, 2006, ISBN 10: 0-387-31073-8.

For this, in one embodiment, possible frame margins are mathematically, e.g. Binary function on the image with an argument of the function describing, for example, points as a result of points or also as a spectacle lens to be inserted into the spectacle frame. between 0 (outside the lens) and 1 (inside the lens). The amount of possible frame margins represents a Fassungsrand Funktionsschar dar. To the

In order to identify the frame border or its shape in the at least one image, a cost function is defined which uses eg the edges in the image or the colors in the image and assigns a value to each possible border of the frame. The minimum of this cost function (minimized by the marginal margins feature set - the cost function is thus a function of functions or frame margins) is then the best frame constraint function that fits the assumptions (edges and color, etc.) in the cost function. In addition to the other terms in the cost function (edge, color, symmetry, etc.), one can then add a penalty to the cost function that determines the deviation of the frame edge to be identified in the at least one image (in the mathematical representation) from all possible contours. which can describe the parametric socket edge model. A penalty is generally an additional term in the cost function, which is the "cost", ie the Value of the function, increased depending on certain circumstances, in this case depending on a deviation of the frame edge from the parametric frame edge model. That is, solutions for the frame edge to be identified that are far from the space of the frame margins that can be written by the parametric frame model have a high cost in optimizing the frame edge to be identified (ie optimizing a frame edge as well as possible) identify the at least one image so that the identified frame edge corresponds as closely as possible to the actual frame edge). This Strafterm you can to all sorts

Add cost functions that optimize the frame edge to be identified in the at least one image.

In another possibility of the analysis, the frame edge is identified in the image by means of conventional methods, for example with the methods mentioned in the introduction to the description of DE 10 201 1 1 15 239 A1. At this identified frame edge, the parametric frame edge model is then adjusted, i. the parameters of the parametric

Frame border models are optimized in order to achieve the greatest possible match of the parametric frame edge model with the frame edge identified in the at least one image. This adaptation can be done by the least squares method or another conventional optimization method. In this application, it is easier to determine the parameters of the parametric frame edge model than in the case where the parametric frame edge model is used for segmentation as described above, since only the parameters are found and the frame edge has already been identified in the image. After determining the parameters, the parametric frame edge model can then be used for further evaluations. For example, a position of the eyes can also be determined from the at least one image or with a distance measurement to the camera, and the centering parameters mentioned in the introduction can then be determined from the position of the eyes and the parametric frame edge model. Since these centering parameters are simple geometric quantities, they then simply result from the geometric description of the frame edge, which is given by the parametric frame edge model, and the position of the eyes for this purpose. Some centering parameters, such as

Dimensions of a rectangle formed by the horizontal and vertical tangents at the outermost edges of the spectacle lens (box system according to 5.1 DIN EN ISO

13666: 2012), even without the position of the eyes, only from the parametric Margin frame model are determined, since they depend only on the dimensions and the shape of the frame edge.

The adaptation of the parametric frame edge model to a frame edge identified in an image may, in the case where the parametric frame edge model includes the plurality of data sets or other data sets derived from the plurality of data sets, also be the best matching one of the plurality of data sets or other records. For this purpose, for example, the deviations from the identified frame edge are calculated for all data sets of the plurality of data records or the plurality of further data records and the data record with the smallest

Deviations selected. The number of this record then represents the specific parameter.

The above methods are typically performed computer-implemented. Accordingly, according to another aspect of the present invention, a

A computer program provided with a program code which, when executed on one or more processors of a computing device, one of the above

performs described method. Finally, a corresponding computing device having a processor and a memory in which a corresponding computer program is stored is also provided. The memory can be in particular a physically tangible memory. The

Computing device also has a data input to receive the records. The data input can be a network connection to receive the data records via a network such as the Internet, or a recording for a storage medium to receive the data stored on the storage medium records, such as a USB interface for receiving a memory stick or an optical Drive (DVD or CD drive) for holding an optical storage medium. For further explanation, the invention is explained in more detail by means of embodiments with reference to the accompanying drawings. Show it:

1 shows a flow chart of a method according to an exemplary embodiment, FIG. 2 shows a flow chart of a method according to a further exemplary embodiment, 3 is a block diagram of a computing device according to an embodiment,

4 is a diagram illustrating a data set, and FIG. 5 is a diagram illustrating a parametric model.

FIG. 1 shows a flow chart, which generally shows the sequence of a method according to an exemplary embodiment. In a step 10, the method includes providing a plurality of data records each having a course of a frame edge of a

Mark eyeglass frame. As previously explained, these records may include tracer data or CAD data. In a step 11, the method then comprises providing a parametric frame edge model based on the data records provided in step 10, wherein the possibilities already explained can be used for the provision. FIG. 2 shows a more detailed flowchart of a method according to FIG

Embodiment. In the embodiment of FIG. 2, in a step 20 data records, for example tracer data, are provided as training examples. In a step 21, preprocessing, i. transforming the data records into another format, for example by coordinate transformation or reparameterization as already described.

In step 22, based on the preprocessed data sets, a parametric

Fassungsrandmodell, in this case, a 3D model, generated, for example by

Principal component analysis. As indicated by steps 23 to 26, this model can then be used in various ways. In step 23, for example, a

Segmentation of one or more captured images of at least a portion of a head (especially the eye area) that supports a spectacle frame using the SD model. Here, as described above, the parametric frame edge model is used to facilitate identifying a frame edge in captured images.

In step 24, as described in an optimization process for identifying a frame edge in one or more images, a penalty is created based on a deviation of a currently identified frame edge from the parametric frame edge model. In step 25, as already explained, the data records or the

preprocessed datasets serve as a library. In step 26, a frame edge is first identified in one or more images, and the parameters of the parametric frame border models are then adapted to the identified frame border. Steps 23 to 26 may ultimately serve to determine centering parameters as described. FIG. 3 shows a block diagram of a computing device 30 according to FIG

Embodiment. The computing device 30 includes a processor 31 and a memory 32, such as a random access memory (RAM) or a read-only memory (ROM), in which a program code is stored. When the program code stored in the memory 32 is running on the processor 31, one or more of the methods described above, in particular the method of FIG. 1 or FIG. 2, are executed. For this purpose, the computing device 30 via a data input 33 corresponding

Records such as the described tracer data or CAD data supplied. At one

Data output 34 is then output a corresponding parametric frame edge model, which was determined as described. It should be noted that data input 33 and data output 34 can also be physically identical, for example by a

Network interface or the like may be formed. As such a computing device 30, for example, a commercially available computer can be used, which then

is programmed accordingly. FIG. 4 shows a plurality of data records 40 in a 2D representation

Illustration of the concepts explained above. In Fig. 4, each record corresponds to a closed line indicating the portion of a frame edge for a spectacle lens. From this variety of lines, which for example tracer data from measurements

represent different spectacle frames, then as described a parametric frame edge model can be created.

FIG. 5 shows a simple diagram for illustrating a parametric one

Frame edge model. In the example of FIG. 5, an ellipse 50 serves as a model, and the parameters are the large semiaxis a and the small semiaxis b of the ellipse 50. In one of the variants already described, the ellipse 50 can serve, for example, as the predetermined parametric frame edge model 50. The ellipse 50 may then be adapted to each data set, for example to each of the curves of the data set 40 of FIG. 4, so as to obtain for each data set a value for the major semiaxis a and a value for the minor semiaxis b. From the values thus determined, a probability distribution for the parameters a and b can then be created, for example in the form of a histogram or in the form of mean value and standard deviation. It should be noted that the ellipse 50 in this case was chosen only as a simple example, and even more complex geometric shapes are possible. In general, it should be noted that the illustrated and described

Embodiments are for illustrative purposes only and are not to be construed as limiting.

Some embodiments are defined by the following clauses:

Clause 1. A method of providing a frame border model comprising:

Provide a variety of records, each one a course of a

Describe the frame edge of a respective spectacle frame, and

Providing a parametric margin model based on the plurality of data sets.

Clause 2. Procedure according to clause 1, wherein providing the parametric

Frame edge model, calculating the parametric frame edge model from the

Variety of records includes.

Clause 3. The method of clause 2, wherein the computing includes machine learning on the plurality of data sets.

Clause 4. Procedure according to clause 2 or 3, whereby the calculation is a

Principal component analysis and / or independence analysis of the plurality of

Includes records. Clause 5. Procedure according to any one of clauses 1-4, whereby the provision of the

parametric borderline model computing

Probability distributions for parameters of a given

A frame border model based on the large number of data sets. Clause 6. A method according to any of clauses 1-5, wherein providing the

parametric framing model providing the plurality of data sets or a plurality of others derived from the plurality of data sets

Comprising a parameter of the parametric frame edge model, a record of the plurality of data sets or the plurality of others

Indicates records. Clause 7. The method of any of clauses 1-6, wherein providing the parametric frame edge model comprises transforming the plurality of data sets.

Clause 8. The procedure of Clause 7, wherein transforming a

Coordinate transformation and / or scanning of curves defined by the data sets at equal angular intervals and / or filtering of the data sets and / or an approximation of the data sets by parametric curves comprises.

Clause 9. The procedure of clause 8, wherein the coordinate transformation is a

Transformation to corresponding reference points of a frame edge described by the respective data set and / or a similar alignment of the

Includes records.

Clause 10. Image analysis method comprising:

Providing at least one image of at least part of a head with a spectacle frame,

Providing a parametric margin model created by the method of one of clauses 1-9, and

Analyze the at least one image using parametric

Frame edge model.

Clause 1 1. The method of clause 10, wherein analyzing the image comprises identifying a frame edge in the at least one image with the aid of the

parametric frame edge model.

Clause 12. The method of Clause 1 1, wherein identifying the frame edge comprises an optimization process with a penalty depending on a deviation of a frame edge to be optimized from the parametric frame edge model and / or identifying the frame edge based on probability distributions of parametric frame edge model parameters.

Clause 13. The method of clause 10, wherein analyzing comprises identifying a

Frame border in which at least one image and adjusting parameters of the parametric frame edge model based on the identified frame edge comprises. Clause 14. A computer program having a program code which, when executed on a processor, performs the method of any of clauses 1-13. Clause 15. A computing device comprising:

a processor,

a data input for receiving a plurality of data records, each describing a profile of a frame edge of a respective spectacle frame, and a memory, wherein in the memory the computer program is stored according to clause 14.

Claims

claims

A computer-implemented method of providing a frame edge model (50), comprising:

Providing a plurality of data sets (40), each having a course of a

Describe the frame edge of a respective spectacle frame, and

Providing a parametric frame edge model (50) based on the plurality of

Records

characterized in that

the multitude of records are tracer records.

The method of claim 1, wherein providing the parametric

A frame boundary model comprises calculating the parametric frame edge model from the plurality of data sets (40).

The method of claim 2, wherein the computing includes machine learning on the plurality of data sets (40).

Method according to claim 2 or 3, wherein the calculating a

Principal component analysis and / or independence analysis of the plurality of

Includes data sets (40).

The method of any one of claims 1-4, wherein providing the parametric frame edge model (50) comprises calculating probability distributions for parameters of a given frame edge model based on the plurality of data sets (40).

The method of claim 1, wherein providing the parametric frame edge model comprises transforming the plurality of datasets to generate a plurality of further datasets, wherein each dataset of the plurality of further datasets is based on a dataset of the plurality of datasets.

The method of claim 6, wherein the transforming comprises a coordinate transformation and / or a sampling of curves defined by the data sets in the same Angular distances and / or filtering of the data sets and / or an approximation of the data sets by parametric curves comprises.

The method of claim 7, wherein the coordinate transformation transforms to corresponding reference points of one through the respective data set

described border of the frame and / or an alignment of the same records.

The method of claim 1, wherein providing the parametric frame edge model comprises providing the plurality of datasets (40) or the plurality of further datasets, wherein a parameter of the parametric frame edge model comprises a dataset of the plurality of datasets or the plurality of further datasets indicates.

The method of any of claims 1-9, wherein the tracer records are 2D records.

The method of claim 10, wherein the parametric frame edge model is an SD model.

The method of any of claims 1-10, wherein the parametric

Frame edge model is a 2D model.

A computer-implemented method of providing a frame edge model (50), comprising:

Providing a plurality of data sets (40), each having a course of a

Describe the frame edge of a respective spectacle frame, and

Providing a parametric frame edge model (50) based on the plurality of

Records

characterized in that providing the parametric

Frame boundary model (50) calculating probability distributions for parameters of a given frame edge model on the basis of the plurality of

Includes data sets (40). 14. The method of claim 13, wherein providing the parametric

A frame border model involves transforming the multitude of records into one Create a plurality of further data records, wherein each record of the plurality of further data records based on a record of the plurality of data records.

15. The method of claim 14, wherein the transforming comprises a coordinate transformation and / or a sampling of curves defined by the data sets in the same

Angular distances and / or filtering of the data sets and / or an approximation of the data sets by parametric curves comprises.

16. The method according to claim 15, wherein the coordinate transformation comprises a transformation to mutually corresponding reference points of a frame edge described by the respective data record and / or an equal alignment of the data records.

17. The method of claim 13, wherein calculating

Probability distributions include creating a histogram.

18. The method of claim 13, wherein calculating

Probability distributions include calculating probability distributions of parameters of a parameter-free model.

19. A computer-implemented method of providing a frame edge model (50), comprising:

Providing a plurality of data sets (40) each describing a course of a frame edge of a respective eyeglass frame, and

Providing a parametric frame edge model (50) based on the plurality of

Records

characterized,

in that providing the parametric frame edge model comprises providing the plurality of data sets (40) or a plurality of further data records derived from the plurality of data records, one parameter of the parametric one

Perimeter frame model specifies a record of the plurality of records or the plurality of other records.

20. The method of claim 19, wherein providing the parametric frame edge model comprises transforming the plurality of data sets to form the Create a plurality of further data records, wherein each record of the plurality of further data records based on a record of the plurality of data records.

The method of claim 20, wherein the transforming is a coordinate transformation and / or sampling of curves defined by the data sets in the same

The method of claim 21, wherein the coordinate transformation comprises a transformation to corresponding reference points of a frame edge described by the respective data record and / or an equal alignment of the data records.

A computer-implemented method of providing a frame edge model (50), comprising:

Providing a plurality of data sets (40), each having a course of a

Describe the frame edge of a respective spectacle frame, and

Providing a parametric frame edge model (50) based on the plurality of

Records

characterized,

in that providing the parametric frame edge model comprises transforming the plurality of data sets to generate a plurality of further data sets, each data set of the plurality of further data sets being based on a data set of the plurality of data records, wherein the transforming comprises

The method of claim 23, wherein providing the parametric

A frame boundary model comprises calculating the parametric frame edge model from the plurality of further data sets (40).

The method of claim 24, wherein the computing comprises machine learning on the plurality of further data sets (40). The method of claim 24 or 25, wherein calculating a

Main component analysis and / or independence analysis of the plurality of further data sets (40).

The method of any one of claims 23-26, wherein providing the parametric frame edge model (50) comprises calculating probability distributions for parameters of a given frame edge model based on the plurality of further data sets (40).

Method according to one of claims 23-27, wherein the coordinate transformation comprises a transformation to mutually corresponding reference points of a frame edge described by the respective data record and / or an equal alignment of the frame

Includes records.

The method of any one of claims 23-28, wherein providing the parametric frame edge model comprises providing the plurality of further data sets, wherein a parametric frame edge model parameter specifies a record of the plurality of further data sets.

The method of any one of claims 12-29, wherein the records are 2D records.

31. The method of claim 30, wherein the parametric frame edge model is an SD model. 32. The method according to any one of claims 13-30, wherein the parametric

Frame edge model is a 2D model.

33. A computer-implemented image analysis method comprising:

Providing at least one image of at least part of a head with a spectacle frame,

Providing a parametric frame edge model created by the method of any one of claims 1 to 32, and

Analyze the at least one image using parametric

Frame edge model.

34. The method of claim 33, wherein the analyzing comprises determining camera parameters of a camera used to capture the at least one image. 35. The method according to any one of claims 33 or 34, wherein the analyzing a

Image enhancement of the at least one image includes.

36. The method of any of claims 33-35, wherein analyzing the image

Identify a frame border in the at least one image using the parametric frame border model includes.

37. The method of claim 36, wherein the analyzing comprises adjusting parametric frame edge model parameters based on the identified frame edge.

38. The method of claim 37, wherein identifying the frame border a

Optimization process with a Strafterm depending on a deviation of a socket edge to be optimized from the parametric frame edge model and / or identifying the frame edge based on probability distributions of parameters of the parametric frame edge model.

39. The method of claim 33, wherein the analyzing comprises determining a representation of an edge of a spectacle lens with the aid of the parametric frame edge model.

A computer program having a program code which, when executed on a processor, executes the method of any of claims 1-39.

41. Calculating device, comprising:

a processor,

a data input for receiving a plurality of data records, each describing a profile of a frame edge of a respective spectacle frame, and a memory, wherein in the memory, the computer program according to claim 40 is stored.

42. A method for centering a spectacle lens in a spectacle frame, characterized in that

in a step (i) to the spectacle lens centering parameters are determined, wherein the determining the centering parameters comprises determining the representation of the edge of the spectacle lens with the method according to claim 39 and

in a step (ii), the spectacle lens is centered in the spectacle frame with the centering parameters determined in step (i).

43. A method for Einschieifen a spectacle lens in a spectacle frame, characterized

marked that

in a step (i), centering parameters are determined for the spectacle lens, wherein determining the centering parameters comprises determining the representation of the edge of the spectacle lens with the method of claim 39; and

in a step (ii), the spectacle lens based on that determined in step (i)

Centering parameters for an arrangement in the spectacle frame is ground.

44. A method for producing a spectacle lens, characterized by the

Method step einschieifen the spectacle lens in a spectacle frame with a

Method according to claim 43.