2156 results about "Image noise" patented technology

The present invention provides a system and method for recognizing a 3D object in a single camera image and for determining the 3D pose of the object with respect to the camera coordinate system. In one typical application, the 3D pose is used to make a robot pick up the object. A view-based approach is presented that does not show the drawbacks of previous methods because it is robust to image noise, object occlusions, clutter, and contrast changes. Furthermore, the 3D pose is determined with a high accuracy. Finally, the presented method allows the recognition of the 3D object as well as the determination of its 3D pose in a very short computation time, making it also suitable for real-time applications. These improvements are achieved by the methods disclosed herein.

The invention relates to the field of image recognition, and especially relates to a complex character recognition method based on deep learning. Through the analysis of character complexity, a training sample, which contains a to-be-recognized image noise model and a distortion characteristic model, generated by a random sample generator is employed for the training of a deep neural network. The training sample comprises complex noise and distortion, and can meet the demands of the recognition of various types of complex characters. A few of manually annotated first training sample sets and a large amount of randomly generated second training sample sets are mixed and then inputted to the deep neural network, thereby solving a problem that a large number of manually annotated training samples are needed for character recognition through the deep neural network. Moreover, the most advanced deep neural network is employed for automatic learning under the condition that the noise and distortion of a to-be-recognized image are retained, thereby avoiding information loss caused by noise reduction in a conventional OCR method, and improving the recognition accuracy.

There is described image-processing methods and image processing apparatus, which enable sharpness, scaling coefficients, and image quality to be adjusted with minimum image noise and minimum image quality deterioration and at high processing speed. One of the image-processing methods for creating processed image data by applying a spatial-filtering processing to source image data, comprises the steps of: setting a predetermined upper-limit value for a variation amount of the source image data, before performing an image-conversion processing through which the source image data are converted to the processed image data by applying the spatial-filtering processing; and performing the image-conversion processing for the source image data within a range of the variation amount limited by the predetermined upper-limit value. In the above, a plurality of spatial-filtering processing(s), characteristics of which are different each other, are performed either simultaneously in parallel or sequentially one by one in the image-conversion processing.

The invention provides a three-dimensional terrain model real-time smooth drawing method with combination of a GPU technology, and belongs to the technical field of image processing. The objective of the invention is to provide the three-dimensional terrain model real-time smooth drawing method with combination of the GPU technology so that cache reuse in multiple times of drawing can be realized based on the current popular programmable GPU technology with a global digital elevation model acting as a data source, and load of computation space is effectively reduced. The method comprises the steps of construction of a multi-resolution pyramid model, elimination of image noise points, filtering of images, partitioning of planar projection of the earth according to equal latitude and longitude, and construction of different hierarchical levels of pyramid layers according to a mode from the top to the bottom. Acceleration and enhancement of terrain rendering are realized based on the programmable GPU technology, i.e. all phases of a graphical drawing pipeline are controlled by using shader languages, two and textures are respectively generated by vertex information and index information of elevation data to be stored in video memory for scheduling of whole terrain drawing; and vertex interpolation and migration are performed in the geometric phase by utilizing a curved surface subdivision and fractal technology so that procedural details are generated and the phenomenon of edges and corners of the terrain mesh when resolution is insufficient can be compensated.

The invention relates to an attention mechanism U-shaped densely connected retinal (a novel retinal blood vessel segmentation method fusing DenseNet and Attention U-net network) blood vessel segmentation method, comprising the steps of retinal blood vessel image preprocessing and constructing a retinal blood vessel segmentation model. The method of the invention can effectively solve the problemsthat adjacent blood vessels are easy to connect, the micro blood vessels are too wide, the small blood vessels are easy to break, the blood vessel intersection is insufficient in segmentation, the image noise is too sensitive, the gray value of the object and the background is crossed, the optic disc and the lesion focus are missegmented, and the like. The method of the invention integrates a plurality of network models under the condition of low complexity, the excellent segmentation results are obtained on DRIVE dataset, the accuracy and sensitivity are 96.95% and 85.94%, respectively, whichare about 0.59% and 7.92% higher than those published in the latest literature.

Methods and electronic devices for performing color-based reaction testing of biological materials. The method includes capturing and interpreting digital images of an unexposed and later exposed paddle at various delay times within an automatically calibrated environment. The test paddle includes a unique identification mechanism (UID), a Reference Color Bar (RCB) providing samples of standardized colors for image color calibration, compensation and corrections, and several test-specific sequences of Chemical Test Pads (CTP). The method further includes locating the paddle in the image, extracting the UID and validating the paddle, extracting the RCB and locating the plurality of CTP in each image. The method further reduces image noise in the CTP and calibrates the image automatically according to lighting measurements performed on the RCB. To determine test results, the method further determines several distances between the CTP and its possible trajectory in the color space described by the Manufacturer Interpretation Color Chart.

An image encoding method according to the present invention calculates first difference information of an encoding target block and second difference information of an adjacent block, calculates a difference between the first difference information and the second difference information, and uses an intra mode only for blocks where a residual image noise occurs by selecting the intra mode when the difference is larger than a predetermined threshold, and selecting an inter mode when the difference is smaller than the predetermined threshold.

The invention discloses a character detection method and a device. The character detection method comprises steps of receiving image to be detected, generating a character area probability graph of the whole text of the image to be detected which is generated by a semantic prediction model, wherein the character area probability graph uses different the pixel value area to distinguish the Chinese character area of the image to be detected and the non-Chinese character area of the image to be detected and performing segmentation operation on the probability graph of the Chinese character area in order to determine the Chinese character area. The Character detection method disclosed by the invention can detect the characters of various languages, directions, colors, fonts and sizes while effectively inhibiting the interference of the complex background. Besides, the Chinese character detection method and the device have characteristics of strong robustness and effectively resisting the interferences of the image noise, image fuzziness, the complex image background and the non-uniform light irradiation.

The present invention provides a Satellite-bone camera space image distortion correction method based on the stellar map matching, the stellar point on the image is used as referecen point, achieving the distortion correction, including following steps: (1)image segmentation, extracting the stellar facula in the image; (2) calculating areal coordinates of each facula by the centroid method, as a real coordinate; (3) processing pre-correction according to the deformation parameter obtained by the ground standardization of the camera; (4) processing stellar map matching by single frame or multiple frames Hausdorff distance stellar map matching method, calculating academic coordinate corresponding to the practical coordinate; (5) calculating deformation parameter of the image by the corresponding relation between the practical coordinate and the academic coordinate; (6) obtaining corrected image by the three steps fitting of a polynomial and interpolation method. In addition, the invention aiming at correctoon impact of the camera plain shaft point system error, provides a adaptive compensation method. The invention has small operation, evident image correcting effect, and effectively preventing the impact of the correction caused by the image noise and the input system error.

The invention provides an image segmentation method based on a watershed algorithm and a morphological marker. The method comprises the steps that median filtering is carried out on a gray level image to obtain a filtered image; an OTSU method is carried out on the filtered image to obtain a binary image; the binary image is processed through a morphological algorithm based on reconstruction to obtain a characteristic marked image; the characteristic marked image is transformed through the watershed algorithm to obtain a segmented image. According to the image segmentation method, the OTSU method and median filtering are utilized for filtering out impurities and noisy points in the image, the image is adopted as the primary mark source of the watershed algorithm, and the interference of noise is effectively eliminated; a morphological operation method is adopted, the information of an effective area cannot be lost, meanwhile, certain fuzzy areas or connected areas can be separated, and the integrity and consistency of image segmentation are guaranteed. Connected domain calculation is combined, the invalid target and information of non-noisy points can be removed, the marker of the watershed algorithm is precisely located, and the over-segmentation phenomenon is eliminated.

A system and method for creating a combined or mixed-energy image using both low- and high-energy CT data sets acquired using a dual-energy CT system. The low- and high-energy datasets are mixed using desired weighting factors to mimic a “single-energy” image. The low-energy dataset provides data with improved contrast enhancement, but with increased noise level. The high-energy dataset provides data with lower contrast enhancement, but with better noise properties. By combining the low- and high-energy datasets in accordance with the present method, the resulting mixed-energy images utilize the information of full dose of radiation used in the dual-energy scan. A plurality of weighting metrics can be selected, including patient size, dose partitioning, or image quality, to determine the desired weighting factors based on the weighting metrics. By selecting the proper weight factors, image noise can be reduced and / or the contrast to noise ratio can be increased in the mixed-energy image.

The method comprises: shooting multi pictures with different exposure levels at same scene; according to the response curve of camera, portioning the each picture into several blocks; selecting the blocks having the most information to make block combination; using a fusion function to remove the edge effects between blocks to finally get an enhanced picture with wide dynamical range.

An image reading apparatus for reading indicia on the front and back surfaces of a document using two image sensors includes a document detection sensor to detect the document, a first image sensor for reading an image on one surface of the document, a second image sensor for reading an image on the other surface of the document, and a controller. The controller starts to read the image data from the first and second image sensors when prescribed amounts of time have passed after the document detection sensor detects the leading edge of the document, and stops the reading of the image data from the image sensors. The structure provides a lightweight and compact apparatus while reducing image noise caused by the mutual interferences of the light sources in reading the image data.

The invention discloses a driver fatigue detecting system based on a smart mobile phone. A video colleting module (1) collects images of the face and the eyes of a driver. An image preprocessing module (2) removes image noise through the gray scale adjustment of the image pixel points. A face detecting and locating module (3) achieves face detection and locating based on image pixel nonlinear color transformation. A face following module (4) improves the speed of detection of the face area of the driver through the correlation between every two adjacent images. An eye detecting module (5) conducts binarization on the images based on the improved horizontal Sobel edge detection method. An eye feature parameter extracting module (6) extracts the pupil opening degree feature parameters. A fatigue judging module (7) judges whether the driver is fatigue when driving a vehicle based on the PERCLOS method. According to the driver fatigue detecting system based on the smart mobile phone, the face detection is rapid, efficient and accurate, the requirement for the resolution is low, a camera of an ordinary mobile phone can achieve image collection and detection, popularization and application are convenient, cost is low, and the accuracy is high.

The invention relates to a U-shaped retinal vessel segmentation method adaptive to scale information. The U-shaped retinal vessel segmentation method comprises the following steps: preprocessing a retinal vessel image; And constructing a retinal vessel segmentation model 2. The method can effectively solve the problems that adjacent blood vessels are easy to connect, microvessels are too wide, tiny blood vessels are easy to break, segmentation at the intersection of the blood vessels is insufficient, image noise is too sensitive, a target and a background gray value are crossed, and a visual disc and a focus are segmented by mistake. According to the method, multiple network models are fused under the condition of low complexity, an excellent segmentation result is obtained on a DRIVE dataset, and the accuracy rate and the sensitivity of the method are 97.48% and 85.78% respectively. And the ROC curve value reaches 98.72% and reaches the current medical practical application level.

The invention discloses a method and a device for detecting tissue boundaries by use of ultrasonic images. The method comprises a search step and a detection step, wherein the search step is used for searching an ultrasonic image of a detection target for an initial center point of a target region and determining an initial region boundary curve according to the initial center point; the detection step takes the initial center point as a pole, starts searching for the position of a target region boundary along a polar line passing the pole from an initial region boundary, and calculates the energy functional of gray-scale distribution in different positions, wherein a position corresponding to the minimum value of the energy functional of gray-scale distribution is the boundary of the target region. As the invention first detects the center point and then rapidly detects the boundaries along the polar line, the invention is simpler and rapider in detection. Region segmentation functional adopted in the invention can effectively avoid the influence of ultrasonic image noise and effectively detects weak boundaries. In addition, grey scale correction along the polar line can eliminate the influence of non-uniform fields of image gray-scale distribution.

The invention relates to the field of image recognition, and especially relates to a random sample generation method for the recognition of a complex character. In the recognition of the complex character, a large number of samples, which are generated by a random sample generator and comprise a to-be-recognized image noise model and a distortion characteristic model are employed through the analysis of the reason of the character complexity on the basis of a standard character similar to a to-be-recognized character. A training sample automatically generated by the random sample generator comprises various types of complex noises and distortions, and can meet the recognition demands of various types of complex characters. The above random sample is enabled to serve as a training sample in a depth neural network, thereby solving a problem that a training neural network needs a large amount of manual marking for character recognition, enabling the automatic recognition of complex characters and images to be more simple and easier, and remarkably saving the related labor cost.

The present invention discloses an improved full-convolutional neural network-based power transmission line insulator state recognition method. The method includes the following steps that: S1, the picture of a power transmission line insulator is collected through an unmanned aerial vehicle; S2, classification regression and position regression are performed on the image through a target detection network Faster R-CNN so as to intercept a separate insulator picture; S3, semantic segmentation is performed on the insulator picture through using a full-convolutional neural network; S4, fine segmentation is performed through a full-connection condition random field; S5, noise points in the image are filtered by using a morphological operation method; and S6, the insulator is classified through a deep learning classification network, and the status of the insulator is determined. According to the method of the invention, training and parameter adjustment and optimization are performed on labeled insulator pictures; the status of the power transmission line insulator can be effectively identified; the subjective influence of manual setting of thresholds and the randomness of manual extraction of features in traditional insulator status recognition can be avoided; the efficiency of line inspection can be significantly improved; and the difficulty of the line inspection can be decreased.

The invention discloses a vehicle tracking method based on machine learning and optical flow. A vehicle model is obtained through one-off off-line training and used for detecting vehicle block mass Blobs in a video flow in real time, bidirectional pyramid optical flow tracking is performed on calculation characteristic point sets of all the vehicle mass block Blobs, and results of optical flow tracking in the forward direction and the backward direction are analyzed and filtered, so that multiple targets are stably and accurately tracked to form vehicle tracks. According to the complete vehicle tracking solution, the vehicle tracking method based on machine learning and optical flow can be widely applied to the fields of intelligent traffic, electronic polices, video monitoring, unmanned driving and others; by the utilization of the tracking method, a user can solve the classic problems in an existing tracking algorithm well, the multiple targets, such as long-period vehicle staying, size scale changing, shadowing, local shielding and touching, can be stably and accurately tracked; particularly, the vehicle tracking method has the good effects under the conditions of severe weather, a low illumination level and a high noisy point.

A method for reducing noise in images includes utilizing an original image to generate a segmentation mask and a filtered image, and utilizing the filtered image, the original image, the segmentation mask and separate blending parameters for structural and nonstructural portions of the filtered image to generate a blended image. The method further includes contrast matching the blended image with the original image to produce an output image.

A digital camera captures high quality long exposure images by capturing and summing several images of the same scene. The effective ISO of the camera is reduced by scaling the summed image, thus reducing image noise and improving long exposure quality.

An edge-preserving diffusion filter maintains the sharp edges in images while smoothing out image noise. An edge-preserving diffusion filter applies an edge-preserving smoothing filter to an image to form a filtered image. The modified image is then blurred by a blurring filter to form a blurred image. The modified image and the blurred image are blended together to form an output image based on an error metric associated with each pixel. The edge-preserving diffusion filter may be utilized to perform a multilevel decomposition of the image. The edge-preserving diffusion filter may be applied to an unfiltered image to produce a base image. The difference between the unfiltered image and the base image defines a detail image. The detail image may be used as the input for recursively generating additional levels of detail. The multilevel decomposition may utilize filter kernels associated with different contrast levels for each iteration.

The invention discloses a pointer instrument reading identification method. The method comprises the steps of extracting a dial plate, removing interference, and correcting the dial plate by using affine transformation; performing circle detection and straight line detection on the dial plate to obtain predicted circle center and pointer angle; based on features of a scale connected domain on thedial plate, screening out the scale connected domain, performing linear fitting on the scale connected domain, calculating an intersection point of every two fitting straight lines, and performing screening verification on a to-be-selected circle center by utilizing the predicted circle center and pointer angle to obtain the circle center of the dial plate, wherein a distance from the centroid ofthe scale connected domain to the circle center on the dial plate is the radius of the dial plate; and according to information such as the circle center of the dial plate, the radius of the dial plate, the straight lines of a pointer, extracting an arc scale bar of the dial plate, expanding the arc scale bar in a rectangular shape, then according to horizontal coordinate distribution of scale onthe rectangular scale bar, performing minimum and maximum scale detection, noisy point removal and scale insertion, and finally in combination with the pointer angle, calculating reading of the dial plate. The method is high in universality and high in timeliness; and the reading has relatively high accuracy.

The invention relates to an SAR image denoising method based on a multi-scale hole residual attention network. The method comprises the following steps of by extracting features of different scales ofthe image through a multi-scale convolution group, broadening a convolution kernel receptive field by utilizing cavity convolution; exuecting more context information of the image; and transmitting the feature information of the shallow layer to a deep convolutional layer by using jump connection to keep image details, adding an attention mechanism to intensively extract noise-related features, and automatically learning the SAR image speckle noise distribution form in combination with a residual error learning strategy to achieve the purpose of removing speckle noise. Experimental results show that compared with a traditional SAR image noise removal method, the method has the advantages that the speckle noise removal effect is good, the number of artificial traces is small, detail information of the image is kept, and the calculation speed is higher through the GPU.

A method and apparatus are provided for reconstructing a tomographic image from fan-beam or cone-beam data. The method includes the steps of collecting fan-beam or cone-beam data over an image space, converting the fan-beam or cone-beam data to parallel-beam data with respect to a rotation angle within the image space, performing a shift variant filtration of the parallel-beam data within the image space and converting the processed data to images through backprojection or other means.

The invention discloses a reflective panoramic imaging system and method. The reflective panoramic imaging method comprises the following steps: directly mounting an optical system on an image sensor or an existing camera optical system, and respectively locating image planes of two groups of fisheye lenses on the image sensor through a reflection element in a non-overlapped manner; selecting a camera to enter a shooting mode; dividing images formed by light on the image sensor after passing by the first fisheye lens and the second fisheye lens; respectively processing the divided images, comprising image correction, color adjustment, noise elimination, exposure and contrast adjustment; extracting, screening and matching feature points of the images; carrying out image registration, splicing and fusion; selecting the shooting mode, and compressing and storing a panoramic image and a panoramic video; and selecting a preview mode, rendering the spliced panoramic image, and displaying the panoramic image on a device connected to the camera. The reflective panoramic imaging method disclosed by the invention can be used for shooting a complete panoramic image, guaranteeing the synchronism of video steam and obtaining the panoramic video more easily.

A noise eliminating device may include a noise eliminating mechanism that eliminates an isolated noise in an image being photographed. The noise eliminating mechanism may include a noise pixel detection mechanism and a pixel correction mechanism. The noise pixel detection mechanism may detect a noise pixel by scanning the image. The pixel correction mechanism may correct a level of a detected noise pixel based on a level of a pixel located in a predetermined area from the noise pixel. A noise eliminating method for eliminating an isolated noise in an image being photographed may include detecting a noise pixel by scanning the image and correcting a level of the detected noise pixel based on a level of a pixel located in a predetermined area away from the noise pixel.