In one embodiment, an apparatus comprises a communication interface and a processor. The communication interface is to communicate with a plurality of devices. The processor is to: receive compressed data from a first device, wherein the compressed data is associated with visual data captured by sensor(s); perform a current stage of processing on the compressed data using a current CNN, wherein the current stage of processing corresponds to one of a plurality of processing stages associated with the visual data, and wherein the current CNN corresponds to one of a plurality of CNNs associated with the plurality of processing stages; obtain an output associated with the current stage of processing; determine, based on the output, whether processing associated with the visual data is complete; if the processing is complete, output a result associated with the visual data; if the processing is incomplete, transmit the compressed data to a second device.

Disclosed herein is a method that comprises obtaining an image of a network section through which flow occurs; where the flow is selected from a group consisting of fluid, electrons, protons, neutrons and holes; subjecting the image to a low pass filter to increase contrast in portions of the network sections; computing a local mean of visible light intensity at each pixel that is present in the image; calculating a visible light intensity difference between each pixel and the local mean of visible light intensity and producing a differentiated image using this calculation; creating a base image of the differentiated image; where the base image comprises a hand segmented gold standard dataset; removing objects below a minimum threshold size from the base image; and retaining the remaining objects if they approximate the line or spine.

The present invention relates to improved systems and methods of image processing and more particularly to improved systems and method of image processing using modified image data to produce enhanced data and images using fewer processing cycles and lower system power.

Embodiments of the present application provide image processing methods and apparatus. A image processing method disclosed herein comprises: acquiring, from an image, two regions which have a textural similarity higher than a first value and have different depths; performing frequency-domain conversion on each of the regions, to obtain a frequency-domain signal of each region; and optimizing the image at least according to the frequency-domain signal of each region, the depth of each region and a focusing distance of the image.

A three-dimensional object detection device includes an image capturing unit, an image conversion unit, a three-dimensional object detection unit and a light source detection unit. The image conversion unit converts a viewpoint of the images obtained by the image capturing unit to create bird's-eye view images. The three-dimensional object detection unit detects a presence of a three-dimensional object within the adjacent lane. The three-dimensional object detection unit determines the presence of the three-dimensional object within the adjacent lane-when the difference waveform information is at a threshold value or higher. The three-dimensional object detection unit set a threshold value lower so that the three-dimensional object is more readily detected in a rearward area than forward area with respect to a line connecting the light source and the image capturing unit.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving a video feed of a scene. Selecting a first portion of the video feed and a second portion of the video feed based on a probability of an object being present in the first portion of the video feed compared to a probability of the object being present in the second portion of the video feed. Processing a first portion of the video feed using a first detection algorithm to detect the object in the first portion of the video feed. Processing a second portion of the video feed using a second detection algorithm to detect the object in the second portion of the video feet, where the first detection algorithm is different from the second detection algorithm.

To conduct defective/non-defective determination on an inspection image with high accuracy, while preventing a feature amount from becoming higher in dimension, and increasing in arithmetic processing time, an inspection image which includes an object to be inspected is acquired; a plurality of hierarchy inspection images by conducting frequency conversion on the inspection image is generated; feature amounts corresponding to types of defects which may be included in the object to be inspected regarding at least one hierarchy inspection image among the plurality of hierarchy inspection images are extracted; and information on the defect of the inspection image based on the extracted feature amount is output.

An illustrative point cloud compression system accesses an input point cloud dataset representative of a point cloud comprising a plurality of points. The point cloud compression system identifies a first attribute dataset and a second attribute dataset within the input point cloud dataset. Based on an application of a transform algorithm to the first and second attribute datasets, respectively, the point cloud compression system generates 1) a first low-frequency component and a first high-frequency component of the first attribute dataset, and 2) a second low-frequency component and a second high-frequency component of the second attribute dataset. The point cloud compression system then generates an output point cloud dataset that prioritizes both the first and second low-frequency components above both the first and second high-frequency components. Corresponding methods and systems are also disclosed.

Described are systems and methods for training a machine-learning model to generate image of biological samples, and systems and methods for generating enhanced images of biological samples. The method for training a machine-learning model to generate images of biological samples may include obtaining a plurality of training images comprising a training image of a first type, and a training image of a second type. The method may also include generating, based on the training image of the first type, a plurality of wavelet coefficients using the machine-learning model; generating, based on the plurality of wavelet coefficients, a synthetic image of the second type; comparing the synthetic image of the second type with the training image of the second type; and updating the machine-learning model based on the comparison.

A fingerprint processing system includes an input unit, a calculation unit and an output unit. The input unit is applied to input an original fingerprint image. The calculation unit is applied to decompose the original fingerprint image to a decomposed image by singular value decomposition (SVD) and the decomposed image is transformed into a plurality of sub-band images by discrete wavelet transformation (DWT) with a template. A plurality of compensation weight coefficients of DWT are calculated to compensate the sub-band images to generate a plurality of compensated sub-band images which are rebuilt by an inverse DWT. After rebuilding the compensated sub-band images, the output unit is applied to output an enhanced fingerprint image.