Disclosed are a method of classifying lesions of chest x-ray radiographs based on data normalization and local patches and an apparatus thereof. The method includes converting an input chest x-ray radiograph into a normalized image, segmenting the converted normalized image into an organ area by using a first neural network based on a pre-learned segmentation model, generating local patches for the segmented organ area, and classifying a lesion in the input chest x-ray radiograph by using a second neural network based on a pre-learned classification model for the generated local patches.

The present invention relates to compensating for brain shift in catheter trajectory planning. First brain shift information is determined from an initial brain image dataset, an initial planning dataset, a patient orientation dataset, and first burr hole dataset. The brain image dataset is updated based on the first brain shift information and a trajectory of a first catheter is updated based on the updated brain image dataset. For a subsequent catheter placement, subsequent brain shift information is determined based on the updated brain image dataset, the patient orientation dataset, and a subsequent burr hole dataset. The brain image dataset is updated again based on the subsequent brain shift information. The re-updated brain image dataset is utilized to update trajectories of the subsequent catheter as well as any preceding catheters.

A method for filtering noise for imaging includes receiving an image frame having position and range data. A filter size divides the frame into filter windows for processing each of the filter windows. For the first pixel, a space to the center pixel and a range difference between this pixel and the center pixel is determined and used for choosing a selected weight from weights in a 2D weight LUT including weighting for space and range difference, a filtered range value is calculated by applying the selected 2D weight to the pixel, and the range, filtered range value and selected 2D weight are summed. The determining, choosing, calculating and summing are repeated for at least the second pixel. A total sum of contributions from the first and second pixel are divided by the sum of selected 2D weights to generate a final filtered range value for the center pixel.

According to one general aspect, an apparatus for extracting an object includes an image receiver configured to receive an image; a coupled saliency-map generator configured to generate a coupled saliency-map which is the sum of the product of a global saliency-map of the image and a predetermined weight value and a local saliency-map: an adaptive tri-map generator configured to generate an adaptive tri-map corresponding to the coupled saliency-map; an alpha matte generator configured to generate an alpha matte based on the adaptive tri-map: and an object detector configured to extract an object according to transparency of the alpha matte to generate an object image.

Techniques and architectures for video stabilization can transform a shaky video to a steady-looking video. A path smoothing process can generate an optimized camera path for video stabilization. With a large smoothing kernel, a path smoothing process can remove both high frequency jitters and low frequency bounces, and at the same time can preserve discontinuous camera motions (such as quick panning or scene transition) to avoid excessive cropping or geometry distortion. A sliding window based implementation includes a path smoothing process that can be used for real-time video stabilization.

A source deconvolutional network is adaptively trained to perform semantic segmentation. Image data is then input to the source deconvolutional network and outputs of the S-Net are measured. The same image data and the measured outputs of the source deconvolutional network are then used to train a target deconvolutional network. The target deconvolutional network is defined by a substantially fewer numerical parameters than the source deconvolutional network.

A method automatically parameterizes error detection of an image inspection system by a computer. The method includes digitizing a reference image in order to determine desired values and subdividing the reference image into homogeneous image regions with few edges, and inhomogeneous image regions with strongly structured image areas and many edges. Lower tolerance values for the homogeneous image regions, and higher tolerance values for the inhomogeneous image regions of the digitized reference image are determined by statistical image analyses. The determined tolerances are assigned to their respective desired values in dependence on a position of the desired values in homogeneous or inhomogeneous image regions. An inspection sensitivity is calculated from desired values and their respective tolerances. The parameters of the image inspection system are set with the aid of the inspection sensitivity configuration of the image inspection system using the parameters.

A method of jitter detection and image restoration for high-resolution TDI CCD satellite images. The method includes: obtaining a parallax image after a sub-pixel matching method relying on a correlation coefficient and least squares algorithm; transferring a jitter offset from an image space to an object space by an integral transformation function (ITF); dealing with a continuous push-broom mode of cameras in a discrete way with a denser sampling strategy according to a continue dynamic shooting model (CDSM), obtaining a specialized CDSM and feeding it back to the ITF; applying the accurate ITF to the obtained parallax image and conducting the jitter curve fitting to achieve the jitter detection; estimating a correspondent partial PSF according to the CDSM and the obtained jitter curve; carrying out an adaptive restoration based on context through optimal window Wiener filtering, by means of multiple input and single output to completes the image restoration.

The present disclosure provides an adaptive stereo matching optimization method, apparatus, and device, and a storage medium. The method includes: acquiring images of at least two perspectives of the same target scene, accordingly obtaining, through calculation, disparity value ranges corresponding to pixels in the target scene; and obtaining optimized depth value ranges by adjusting the disparity value ranges of the pixels in the target scene in real time through an adaptive stereo matching model; adjusting an execution cycle in the adaptive stereo matching model in real time through a DVFS algorithm according to a resource constraint condition of the processing system; and/or training on a plurality of scene image data sets through a convolutional neural network, so that the specific function parameters in the adaptive stereo matching model are correspondingly adjusted in real time according to the acquired different scene images.

Disclosed in the present disclosure are an image registration method and a model training method thereof. The image registration method comprises obtaining a reference image and a floating image to be registered, performing image preprocessing on the reference image and the floating image, performing non-rigid registration on the preprocessed reference image and floating image to obtain a registration result image, and outputting the registration result image. The image preprocessing comprises performing, on the reference image and the floating image, coarse-to-fine rigid registration based on iterative closest point registration and mutual information registration. The non-rigid registration uses a combination of a correlation coefficient and a mean squared error between the reference image and the registration result image as a loss function. Further disclosed in the present disclosure are an apparatus and a system for image registration and a computer-readable medium corresponding to the method. The present disclosure can realize precise and efficient image registration with high applicability between images of different time, different modalities, or different sequences.