Leverton Holding, LLC

Carl Zeiss Microscopy Gmbh

Carl Zeiss Meditec Ag

Michael Kieweg

Florian Kuhlmann

Christian Schäfer

Saurabh Shekhar Verma

Ingo Kleppe

Ralf Netz

Yauheni Novikau

Christoph Nieten

Thomas Mohr

Torsten Sievers

A method for splitting text line images includes receiving a text line image and identifying that the text line image comprises a plurality of zones, wherein each zone includes text whose font differs from the text of adjacent zones. The method further includes selecting a splitting position between multiple zones and splitting the text line image at the splitting position into a plurality of image segments, wherein each image segment contains at least one zone of the text line image and performing optical character recognition on each image segment to recognize a text segment of the image segment. In certain implementations, the method further includes generating one or more confidence measurements and selecting a splitting position that corresponds to a large gradient in the confidence measurement.

Text line image splitting with different font sizes

A method for estimating text heights of text line images includes estimating a text height with a sequence recognizer. The method further includes normalizing a vertical dimension and/or position of text within a text line image based on the text height. The method may also further include calculating a feature of the text line image. In some examples, the sequence recognizer estimates the text height with a machine learning model.

Text line normalization systems and methods

A method for identifying errors associated with named entity recognition includes recognizing a candidate named entity within a text and extracting a chunk from the text containing the candidate named entity. The method further includes creating a feature vector associated with the chunk and analyzing the feature vector for an indication of an error associated with the candidate named entity. The method also includes correcting the error associated with the candidate named entity.

Post-filtering of named entities with machine learning

Methods and systems for detecting tables within documents are provided. The methods and systems may include receiving a text of the document that includes a plurality of words depicted in the document image. Feature sets may be calculated for the words and may contain one or more features of a corresponding word of the text. Candidate table words may then be identified based on the features vectors, and may then be used to identify a table location within the document image. In some cases, the candidate table words may be identified using a machine learning model.

Methods and systems for automated table detection within documents

A method by which fundus photographs that are corrected for scattered light can be realised by using a digital fundus camera. The fundus is illuminated and a photograph of the fundus is realized. The photographed area is greater than the illuminated are. The scattered light intensity is determined and is used for correction. Measurement points for determining the scattered light intensity are defined in the non-illuminated area. Values of the scattered light intensity determined at the measurement points are averaged. Average-scattered light intensity thus determined is subtracted from the intensity values of the photograph to correct the photograph.

Method for realizing ocular fundus photographs that are corrected for scattered light

A method for operating a scanning microscope and for determining point spread functions, with which sample images are recorded with the scanning microscope. The method includes scanning a sample with at least one illuminating light beam; recording at least one sample image with a detector device during a scan by the illuminating light beam; and comprising the point spread function, with which a sample image is recorded, from the at least one sample image. A detector device having receiving elements is used, where the distance between the receiving elements is smaller than a diffraction disk that generates a sample point on the detector device. Detector signals, generated by means of the receiving elements, are read out for each of the different positions of the illuminating light beam on the sample, as a result of which the scanning of the sample allows the detector signals, which are read out, to generate a plurality of sample images. In this case the point spread functions with respect to the different detector signals are defined in each instance by means of an illumination point spread function and a detection point spread function. With respect to all of the detector signals, a matching illumination point spread function is assumed that is shifted in accordance with the scanning motion for different detector signals. In addition, with respect to all of the detector signals, a matching detection point spread function is assumed that takes account of a spatial offset between the detector elements. The plurality of sample images are used to compute the illumination point spread function and the detection point spread function, and these functions are used to compute the point spread functions with respect to the different detector signals. In addition, the invention also relates to a corresponding scanning microscope.

Scanning microscope and method for determining the point spread function (PSF) of a scanning microscope

A method for the microscopy of samples using optical microscopy and particle beam microscopy provides that the samples are divided into a partial quantity and a residual quantity and the samples of the partial quantity are prepared to contain registration marks. The samples of the partial quantity are imaged using optical microscopy and particle beam microscopy, with the result that a pair of optical microscopy images and particle beam microscopy images is obtained for each sample of the partial quantity. The pairs are position-registered relative to each other using the registration marks. The images of the position-registered pairs are modified by removing the registration marks. A registration algorithm is trained which evaluates the image contents and issues a quality measure for a position registration of each pair. The objects of the residual quantity are imaged. These pairs are position-registered by the trained registration algorithm to maximize the quality measure.

Growing community of inventors

Berlin, Germany

Michael Kieweg