Abstract: A computer-implemented method is provided for processing images. The method can include down-sampling a plurality of first images having a first resolution for obtaining a plurality of second images having a second resolution and training an artificial neural network model to process an input image and output an output image having a higher resolution than the input image.

Abstract: A method for identifying traffic sign is provided. The method obtains an original image of a traffic sign, and performs an image enhancing processing on the original image to form an enhanced image. The method performs a first conversion on the original image and the enhanced image to together form a first matrix of image data, and identifies a position of the traffic sign accordingly. The method further masks the original image and the enhanced image respectively to form a first identified region of the original image and a second identified region of the enhanced image accordingly. The method further performs a second conversion on the first identified region and the second identified region to together form a second matrix of the identified region of the image data, and identifies passage instruction information of the traffic sign accordingly. An electronic device and a non-transitory storage medium are also disclosed.

Abstract: An image restoration method and apparatus are provided. The image restoration method includes determining auxiliary data corresponding to a plurality of filter kernels by filtering target data with the plurality of filter kernels, determining new input data by combining the auxiliary data with at least some input data of layers of a neural network-based image restoration model, generating, based on the new input data, a restored image of the input image by executing the neural network-based image restoration model, wherein the filter kernels are not part of the neural network-based image restoration model.

Abstract: Methods and systems for identifying quantisation parameters for a Deep Neural Network (DNN). The method includes determining an output of a model of the DNN in response to training data, the model of the DNN comprising one or more quantisation blocks configured to transform a set of values input to a layer of the DNN prior to processing the set of values in accordance with the layer, the transformation of the set of values simulating quantisation of the set of values to a fixed point number format defined by one or more quantisation parameters; determining a cost metric of the DNN based on the determined output and a size of the DNN based on the quantisation parameters; back-propagating a derivative of the cost metric to one or more of the quantisation parameters to generate a gradient of the cost metric for each of the one or more quantisation parameters; and adjusting one or more of the quantisation parameters based on the gradients.

Abstract: An encoding apparatus partitions a digital image into multiple regions for subsequent encoding. A first encryption code is associated with a first region, a second encryption code is associated with a second region and the first code, and a third code is associated with the first code, the second code and a third region. An authentication apparatus authenticates the digital image in an inverse process.

Abstract: A method of obtaining X-Model/Plus-Model filter by combining and then mathematical reducing SSF (Sharpening Spatial Filter) and CF (Clamp Filter) and scaling high quality video with this filter in real time on field programmable gate array (FPGA) is provided. With the method, the developed X-Model/Plus-Model filter was applied before the application of interpolation to increase the quality of the scaled video. The said filter is applied in real time with advantages such as low computational complexity, low memory requirement, low power and low resource consumption, and high operating frequency.

Abstract: Aspects of the present invention relate to a computer-implemented training method for training a segmentation model to segment an image. The method includes receiving a plurality of training data sets each including image data representing an image comprising a foreground and a background; and sample image data comprising one or more sample image occurring in the background of the image. The image data is captured by at least one visible electromagnetic radiation imaging device. The method includes processing each training data set using the segmentation model. The processing of each training data includes supplying the sample image data to the segmentation model; and segmenting the image data to generate a candidate segmentation in dependence on the sample image data. An error is determined for the candidate segmentation. The segmentation model is updated in dependence on the determined error.

Abstract: Methods, systems, and apparatuses, including computer programs encoded on computer storage media, for training a denoising neural network and, once the denoising neural network is trained, generating new images using the denoising neural network. In particular, the described techniques include obtaining data specifying a trained initial denoising neural network, obtaining training data, and training, using the training data and on a denoising objective, a target denoising neural network. The target denoising neural network includes the core-subnetwork of the trained initial denoising neural network, but also includes a target encoder and decoder neural networks that are not included in the trained initial denoising neural network. By training a target neural network that includes the core-subnetwork of a trained initial denoising neural network, the system achieves stable training of a large-scale target denoising neural networks that can be used to generate high resolution images.

Abstract: An image processing apparatus applies an image to a first learning network model to optimize the edges of the image, applies the image to a second learning network model to optimize the texture of the image, and applies a first weight to the first image and a second weight to the second image based on information on the edge areas and the texture areas of the image to acquire an output image.

Abstract: The present disclosure relates to the field of image processing technologies, and provides an ultrasound image segmentation method and apparatus, a terminal device, and a storage medium. With the method, simulated ultrasound images are synthesized based on Computed Tomography (CT) images. An image segmentation model is pre-trained using the synthesized simulated ultrasound images. The pre-trained image segmentation model is migrated, by employing a transfer learning method, to real sample ultrasound images for further training to obtain a final image segmentation model. A segmentation processing on an ultrasound image to be segmented is completed by the final image segmentation model. In this way, the ultrasound images synthesized based on the CT images can be used to replace a part of training data, thereby solving a problem of lack of training data when training the image segmentation model.

Abstract: An electronic device is provided. The electronic device includes a display, a camera module disposed under the display, and a processor electrically connected to the display and the camera module. The processor is configured to acquire a sample frame by using the camera module, identify whether a light source object is included in the sample frame, determine an imaging parameter for acquisition of first multiple frames when the light source object is identified to be included in the sample frame, acquire multiple frames, based on the imaging parameter, composite the multiple frames to generate a composite frame, identify an attribute of the light source object included in the composite frame, and perform frame correction of the composite frame, based on the identified attribute.

Abstract: Systems and techniques are provided for processing image data. According to some aspects, a computing device can determine an optical flow between a current frame having a first resolution and a first previous frame having the first resolution. The computing device can warp a second previous frame having a second resolution based on the determined optical flow to generate a warped previous frame having the second resolution, the second resolution being higher than the first resolution. The computing device can process, using a diffusion machine learning model, a noise frame, the current frame, and the warped previous frame to generate an output frame having the second resolution.

Abstract: A medical-image processing apparatus according to the present invention includes an obtaining unit configured to convert resolution of a medical image of a first resolution subjected to a noise reduction process to obtain a medical image of a second resolution lower than the first resolution and a training unit configured to train a learning model using training data including the medical image of the first resolution subjected to the noise reduction process and the medical image of the second resolution.

Abstract: An image processing method, an image processing device, a training method of a neural network, an image processing method based on a combined neural network model, a constructing method of a combined neural network model, a neural network processor, and a storage medium are provided. The image processing method includes: obtaining, based on an input image, initial feature images of N stages with resolutions from high to low, N is a positive integer and N>2; performing, based on initial feature images of second to N-th stages, cyclic scaling processing on an initial feature image of a first stage, to obtain an intermediate feature image; and performing merging processing on the intermediate feature image to obtain an output image. The cyclic scaling processing includes hierarchically-nested scaling processing of N?1 stages, and scaling processing of each stage includes down-sampling processing, concatenating processing, up-sampling processing, and residual link addition processing.

Abstract: A method includes training a semantic segmentation network to generate semantic segmentation maps having class-wise probability values. The method also includes generating a semantic segmentation map using the trained semantic segmentation network. The method further includes utilizing the semantic segmentation map during training of an image generation network as part of a loss function that includes multiple losses. The semantic segmentation network may be trained to be sensitive to picture quality of an output image generated by the image generation network during the training of the image generation network such that increased degradation of the picture quality of the output image results in decreased prediction confidence by the semantic segmentation network. The semantic segmentation network may be trained to vary the class-wise probability values based on the picture quality.

Abstract: The present invention relates to an ultrasound image enhancement processing system and method thereof. The ultrasound image enhancement processing system includes at least one first ultrasound device and at least one server apparatus. The server apparatus receives a first ultrasound original image file, processes the first ultrasound original image file through a speckle reduction algorithm to generate a first processed image file, and performs a deep learning training on the first ultrasound original image file and the first processed image file to generate a first neural network model. The first neural network learning module is used to output a first speckle reduction enhancement image file. The image content of the first speckle reduction enhancement image file is approximating to the image content of the first processed image file.

Abstract: A method of automatic segmentation of a maxillofacial bone in a CT image using a deep learning, the method includes receiving input CT slices of the CT image including the maxillofacial bone, segmenting the input CT slices into a mandible and a portion of the maxillofacial bone excluding the mandible using a convolutional neural network structure and accumulating 2D segmentation results which are outputs of the convolutional neural network structure to reconstruct a 3D segmentation result. The convolutional neural network structure includes an encoder including a first operation and a second operation different from the first operation in a same layer and a decoder including a third operation and a fourth operation different from the third operation in a same layer.

Abstract: A machine learning algorithm is trained on a number of microscopic images and a measure of outcome of each image. Each image is divided into tiles. The measure of outcome is assigned to each tile of the image. The tiles are then used to train the machine learning algorithm. The trained algorithm may then be used to evaluate images.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for processing an input image using a super-resolution neural network to generate an up-sampled image that is a higher resolution version of the input image. In one aspect, a method comprises: processing the input image using an encoder subnetwork of the super-resolution neural network to generate a feature map; generating an updated feature map, comprising, for each spatial position in the updated feature map: applying a convolutional filter to the feature map to generate a plurality of features corresponding to the spatial position in the updated feature map, wherein the convolutional filter is parametrized by a set of convolutional filter parameters that are generated by processing data representing the spatial position using a hyper neural network; and processing the updated feature map using a projection subnetwork of the super-resolution neural network to generate the up-sampled image.

Abstract: A method of training an image decomposition model, a method of decomposing an image, an electronic device, and a storage medium are provided. The method of training the image decomposition model includes: acquiring a training set; inputting first and second training images into first and second adversarial neural networks respectively, so as to determine a first loss function value; inputting a third training image into the first and second adversarial neural networks respectively, so as to determine a second loss function value; determining a third loss function value according to a comparison result between an acquired fusion image and the third training image, where the fusion image is generated by fusing generated images of the first and second adversarial neural networks, and adjusting a parameter of the image decomposition model according to at least one of the first to third loss function values.