Abstract: The present disclosure relates to image processing and in particular to modification of an image using a processing such as neural network. The processing is performed to generate a correction image based on an input image. Then, the input image is modified by combining it with the correction image. The processing with the neural network includes at least one stage including image down-sampling and filtering of the down-sampled image; and at least one stage of image up-sampling. An advantage of such approach is increased efficiency of the neural network, which may lead to faster learning and improved performance. The embodiments provide methods and apparatuses for the processing with a trained neural network, as well as methods and apparatuses for training of such neural network for image modification.

Abstract: A method and apparatus are provided for processing with a trained neural network, and for training of such neural network for image modification, which relate to image processing and in particular to modification of an image using the processing such as the neural network. The processing is performed to generate an output image. The output image is generated by processing the input image with the neural network. The processing with the neural network includes at least one stage including image down-sampling and filtering of the down-sampled image and at least one stage of image up-sampling. The image down-sampling is performed by applying a strided convolution. According to the application, efficiency of the neural network is increased, which may lead to faster learning and improved performance.

Abstract: The invention relates to a real time-capable analysis of a sequence of electronic images for estimating the pose of a movable object captured by means of the images. The invention further relates to implementing the invention in software and, in this connection, to a computer-readable medium that stores commands, the execution of which causes the method according to the invention to be carried out. The invention proceeds from a skeleton model, which is described by a small number of nodes in 3D space and permits a good data compression of the image information when the co-ordinates of the nodes describe at any time the position of predetermined parts of the moving object. The skeleton model simultaneously represents previous knowledge of the object, by defining e.g. node pairs and optionally also node triplets in the skeleton model that describe cohesive object parts or optionally object surfaces, which are contained in the measured 2½-D image information, i.e. are visible to the camera.

Abstract: The invention relates to a method for the real-time-capable, computer-assisted analysis of an image sequence of an object consisting of elements that can be moved relative to each other and are interconnected, said sequence containing a variable pose, wherein the individual images of the image sequence are recorded by way of a time-of-flight (TOF) camera such that said images can be processed by a computer, and contain brightness and distance data as functions of the pixel coordinates of the camera for each image of the sequence, comprising the following steps: a. Capturing the pixels of an individual image forming the object, b. calculating a three-dimensional (3D) point cloud in a virtual space, said point cloud representing the surface of the object that is visible to the camera, by a computational projection of object-depicting pixels in such a space, while taking captured distance data to the object into consideration, c.

Abstract: The invention relates to a method for the real-time-capable, computer-assisted analysis of an image sequence of an object consisting of elements that can be moved relative to each other and are interconnected, said sequence containing a variable pose, wherein the individual images of the image sequence are recorded by way of a time-of-flight (TOF) camera such that said images can be processed by a computer, and contain brightness and distance data as functions of the pixel coordinates of the camera for each image of the sequence, comprising the following steps: a. Capturing the pixels of an individual image forming the object, b. calculating a three-dimensional (3D) point cloud in a virtual space, said point cloud representing the surface of the object that is visible to the camera, by a computational projection of object-depicting pixels in such a space, while taking captured distance data to the object into consideration, c.

Abstract: The invention relates to a method for the real-time-capable, computer-assisted analysis of an image sequence of an object consisting of elements that can be moved relative to each other and are interconnected, said sequence containing a variable pose, wherein the individual images of the image sequence are recorded by way of a time-of-flight (TOF) camera such that said images can be processed by a computer, and contain brightness and distance data as functions of the pixel coordinates of the camera for each image of the sequence, comprising the following steps: a. Capturing the pixels of an individual image forming the object, b. calculating a three-dimensional (3D) point cloud in a virtual space, said point cloud representing the surface of the object that is visible to the camera, by a computational projection of object-depicting pixels in such a space, while taking captured distance data to the object into consideration, c.

Abstract: The invention relates to a real time-capable analysis of a sequence of electronic images for estimating the pose of a movable object captured by means of the images. The invention further relates to implementing the invention in software and, in this connection, to a computer-readable medium that stores commands, the execution of which causes the method according to the invention to be carried out. The invention proceeds from a skeleton model, which is described by a small number of nodes in 3D space and permits a good data compression of the image information when the co-ordinates of the nodes describe at any time the position of predetermined parts of the moving object. The skeleton model simultaneously represents previous knowledge of the object, by defining e.g. node pairs and optionally also node triplets in the skeleton model that describe cohesive object parts or optionally object surfaces, which are contained in the measured 2½-D image information, i.e. are visible to the camera.

Abstract: The invention relates to a method for the real-time-capable, computer-assisted analysis of an image sequence of an object consisting of elements that can be moved relative to each other and are interconnected, said sequence containing a variable pose, wherein the individual images of the image sequence are recorded by way of a time-of-flight (TOF) camera such that said images can be processed by a computer, and contain brightness and distance data as functions of the pixel coordinates of the camera for each image of the sequence, comprising the following steps: a. Capturing the pixels of an individual image forming the object, b. calculating a three-dimensional (3D) point cloud in a virtual space, said point cloud representing the surface of the object that is visible to the camera, by a computational projection of object-depicting pixels in such a space, while taking captured distance data to the object into consideration, c.

Abstract: The invention relates to a method for optimizing the automatic fluorescence pattern recognition in immunodiagnosis. In this method, in addition to or together with the fluorescence dye, one or more other indicator dyes for the identification of relevant structures are incubated before an image is taken with a camera.

Abstract: The invention relates to a method for the real-time-capable, computer-assisted analysis of an image sequence of an object consisting of elements that can be moved relative to each other and are interconnected, said sequence containing a variable pose, wherein the individual images of the image sequence are recorded by way of a time-of-flight (TOF) camera such that said images can be processed by a computer, and contain brightness and distance data as functions of the pixel coordinates of the camera for each image of the sequence, comprising the following steps: a. Capturing the pixels of an individual image forming the object, b. calculating a three-dimensional (3D) point cloud in a virtual space, said point cloud representing the surface of the object that is visible to the camera, by a computational projection of object-depicting pixels in such a space, while taking captured distance data to the object into consideration, c.

Abstract: The invention relates to a method for optimizing the automatic fluorescence pattern recognition in immunodiagnosis. In this method, in addition to or together with the fluorescence dye, one or more other indicator dyes for the identification of relevant structures are incubated before an image is taken with a camera.

Abstract: The width of bands to be laminated on a mill train is adjusted by vertical upsetting rollers, resulting, however, in a narrowing at the band ends due to the asymetric material flow there. In order to solve the problem, the upsetting rollers are so designed as to move at the passage of the band ends in keeping with a curve defined according to specified parameters. The parameters are based on neuro-computer made predictions related to the milling process.

Abstract: In the control of a material-processing process in a regulated system, a preliminary adjustment of the system takes place at the beginning of each process cycle as a function of a precalculated process parameter. A material characteristic which is relevant for the process and which in turn is dependent on state variables (such as the composition of the material and its temperature), is included in an advance calculation of the process parameter. The relationship between the state variables and the material property is modelled in a neural network which forms a prediction value for the material property on its output side. As a function of the deviation between the prediction value and an actual value for the material property which is determined based on measuring the process parameter during the process cycle, an adaptive change of the network parameters takes place in the sense of reducing this deviation.

Abstract: In rolling a metal strip in a roughing line and a finishing line, the rolling process in the roughing line is adjusted as a function of a predicted value for the change in width of the metal strip in the finishing line such that the metal strip has a given specified finished strip width on leaving the finishing line. In order to permit a reliable prediction of the change in width despite the lack of accurate information regarding the dependence of the change in width on influencing parameters that affect the process, this dependence is simulated in a neural network whose network parameters are adapted after each passage of a metal strip through the finishing line as a function of the influencing parameters measured or calculated during the passage and the measured actual change in width.

Abstract: In known processes for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run based on at least one process parameter. The at least one process parameter is precomputed with a model of the process, containing at least one model parameter and input values supplied to the model. During the process, the input values and the process parameter are measured and used to adaptively improve the precomputed process parameter after the process run. A neural network is used to determine the model parameters whose dependence on the input values is unknown or insufficiently known. Network parameters of the neural network are modified after each process run to adapt the model to the actual process events.

Abstract: In known methods for conducting a process in an automatically controlled system, the system is preset at the beginning of each process run according to at least one process parameter. The at least one process parameter is precomputed with a model of the process which is supplied with input values. During the process, the input values and the process parameters are measured and are used after the process run to adaptively improve the precomputed value of the process parameters. To simplify and improve the precomputed value of a model having a plurality of partial models, computed results of the partial models are supplied to a neural network. The neural network produces the process parameters to be precomputed as a network response. The network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events.

Abstract: Known methods for conducting a process in an automatically controlled system preset the system at the beginning of each process run based on at least one process parameter. The process parameter is precomputed with a model of the process which is supplied with input values. During the process the input values and the process parameter are measured and are used to adaptively improve the precomputed process parameter after the process run. The present invention simplifies and improves the precomputed value of the process parameter by supplying at least part of the input values to a neural network. The network response of the neural network forms a correction value for the approximate value delivered by the model for the process parameter to be precomputed. The network parameters of the neural network are modified after each process run to adapt the precomputed value to the actual process events.

Abstract: For the control of a process in a controlled system, a presetting of the system takes place at the start of each process sequence as a function of a precalculated process parameter which exhibits a system-induced dependence on faulty input variables. In this case, the description of the dependence takes place by a model of the process which is adapted during the course of the process. To prevent dependence on the creation of models, which as a rule are imprecise, the input variables are fed before the start of the process to a neural network having variable network parameters for the precalculation of the process parameter based on measurements of the input variables and of the process parameter. These variables are recalculated during the course of the process and utilized for the adaptation of the network parameters.