186 results about "Gradient operators" patented technology

The invention relates to a video auto-focusing system based on an embedded media processor. The system comprises a video collection module, an image acquisition and processing module, a focusing search module and a focusing execution module, wherein the acquisition and processing module is used to perform noise processing to the obtained original video image data and extract Y-component in the data, wherein the induction area selection algorithm based on region segmentation is adopted to extract focus area, the Tenengrad function based on Sobel gradient operator is adopted to calculate definition; the focusing search module is used to determine the moving direction of the camera lens and calculate the required step length (mov_step = step*level) of the lens; the position corresponding to the maximum definition value is used as focus position; and the focusing execution module is used to transmit control signals to a stepping motor used for driving the lens so as to adjust the position of the lens, according to the determined moving direction and step length of the lens in the focusing search module. By using the video auto-focusing system, the focusing accuracy and stability can be effectively increased, and the smoothness is improved.

An image-based measurement technique that directly measures the crepe pattern on a moving sheet employs a suitable arrangement of illumination, optical elements, and an imaging device to obtain digital images of the sheet surface. The image resolution is sufficient to represent tissue crepe folds at the scale of the crepe. A spectral analysis of some or of all the images reveals the crepe folding pitch. Further analysis of the image optionally reveals other crepe structural parameters, such as the distribution of crepe fold orientation angles and distribution of linear fold lengths. Corrective actions can be implemented in response to changes in the crepe structure. The pitch of creping is the primary parameter for describing the crepe pattern; in addition, analyzing the images with a gradient operator can yield information regarding the orientation of the crepe folds and distribution of angles of the crepe furrows on the hood side or equivalently of the crepe seams on the cylinder side.

The invention discloses a single image super-resolution reconstruction method combining depth learning and gradient transformation. The method comprises the steps that a super-resolution method based on depth learning is used to carry out upsampling on an input low-resolution image to acquire an upsampling image; a gradient operator is used to carry out gradient-extracting on the upsampling image; a depth convolutional neural network is used to convert extracted gradient; a cost function is reconstructed by using the input low-resolution image and the converted gradient as constraints; a gradient descent method is used to optimize the reconstructed cost function to acquire a final output high-resolution image. According to the single image super-resolution reconstruction method provided by the invention, the reconstructed image has a fine structure in the subjective visual effect, is free of artificial effect, and has a high objective evaluation parameter value. The invention provides the effective single image super-resolution reconstruction method.

An image-based measurement technique that directly measures the orientation of fibers in a moving web that comprises nonwoven material measures the orientation angles of the individual fibers so that a more robust estimate of the statistical distribution of fibers is obtained. The technique includes the steps of: (a) illuminating an area on at least one side of the web with radiation; (b) obtaining at least one digital image of the illuminated area; and (c) calculating the fiber orientation of the web by processing the at least one digital image with a gradient operator thereby analyzing the distribution of observed fiber orientation angles within the image. The gradient operator is preferably of a non-integer order between ⅓ and ⅔ and particularly between ¼ and ¾. The use of fractional-gradient operators yields more reliable results than when integer order gradients are employed.

The invention relates to a method for acquiring the binocular image maximum parallax based on an optical flow method, which is characterized in that a series of processing is carried out on images acquired by left and right cameras so as to acquire a motion vector of each point on the images, analysis and judgment are carried out the motion vectors to acquire the maximum parallax finally so as to be convenient for carrying out further processing. The optical flow method is insensitive to influences such as weak textures, repeated textures, noise interferences and the like, so that improvement is carried out on a gradient operator and a window weighing function of data items in an optical flow algorithm. A characteristic of uniqueness of the maximum parallax is considered, and an optical flow is calculated for two times with one time being simple and the other time being complicated, thereby effectively reducing the calculation amount, and accelerating the calculation speed. On the basis of a CLG-TV pyramid optical flow method, a global diffusion coefficient is introduced to a smooth item of the optical flow algorithm, and improvement is carried out on the gradient operator and the window weighing function of the data items in the optical flow algorithm, thereby enabling the method to be more applicable to calculation in the circumstance that a binocular image only has horizontal displacement. As for the pyramid calculation method, the optical flow is calculated for two times with one time being simple and the other time being complicated, thereby effectively reducing the calculation amount, and accelerating the calculation speed.

This invention discloses a method for false color process to medical grey image signals, which first of all utilizes a Gauss low-pass filter to eliminate noises of a polluted medical image, then utilizes a cross gradient operator to acquire the gradient value of each pixel of the image then to determine a threshold value based on the gradient value, then carries out false color coding in two modes based on the comparison of gradient value and threshold value, if the gradient value is greater than the threshold value, it applies red to false-code, otherwise, it applies a non-linear coding method based on the grey value, and medical images process by this method meets vision character of man-eye.

The invention discloses a different-source image matching method based on refined feature optimization extraction, which mainly solves the problem of low matching precision in the prior art, and adopts the technical scheme that 1, gradient moduli and directions of an optical image and an SAR image are calculated by adopting different gradient operators respectively; 2, completing region-of-interest calibration by utilizing the specific property of the target in the image; 3, according to the result of the step 1, obtaining edge features of the interested areas of the two images, and obtaininga coarse matching affine transformation matrix between the two images through template matching; 4, obtaining a transformed optical image by using the coarse matching affine transformation matrix, andthen extracting refined features; 5, performing preliminary matching on the refined features, and eliminating singular point interference to obtain a fine matching affine transformation matrix; and 6, obtaining a final affine transformation relation matrix between the optical image and the SAR image according to the results of the steps 3 and 5. The method can achieve the precise registration ofa heterologous image, and can be used for the guidance of an aircraft.

The invention discloses a method for detecting an edge of a color image on the basis of a local self-adaption color difference threshold and belongs to the field of image processing and computer vision. The method comprises the following steps: firstly, establishing a weighting function containing a color difference threshold of a background luminance mask and contrast ratio sensitivity function; confirming the color difference threshold of each pixel point according to neighborhood information; if a calculating result of a color difference gradient operator of the pixel point is greater than the threshold, judging the present pixel as an edge point and displaying; and if not, setting the pixel luminance as zero. A test proves that according to the method provided by the invention, the luminance mask effect of human eyes and contrast ratio sensitivity characteristics are considered, the visual sensing characteristics of the human eyes are approached, an image edge sensed by the human eyes is effectively detected, and meanwhile, the problem of edge over-detection caused by threshold judgment in the traditional algorithm is avoided, and the noise resistance is better.

The invention provides a method used for automatically cutting X-typed overlapped and conglutinated chromosomes, comprising the steps as follows: step 1: binarization segmentation processing is carried out to a chromosome image so as to obtain a binarization template; step 2: the X-typed overlapped chromosome image is extracted from the template, the chromosome profile is obtained by a gradient operator and treated with Freeman chain coding so as to obtain a boundary chain code; step 3: the overlapped chromosome is processed by thinning algorithm so as to extract a single-pixel framework and determine the central point of an overlapping area; and step 4: the distance from the profile pixel point to the central point of the overlapping area after the Freeman chain code is calculated; four profile concave points which are not adjacent to each other are selected as cutting points so as to divide the profile into two profile sets; and the segmented chromosome profile sets are used for automatically completing the operation of segmenting the chromosome grey images, thus obtaining the automatic segmenting result of the X-typed overlapped and conglutinated chromosome. The method has quick and effective cutting operation and greatly reduces the working quantity and difficulty of manpower segmentation.

A method and system for identifying and determining banding artifacts in digital video content composed of a sequence of moving video pictures includes creating a mask image corresponding to a picture from said sequence of moving video pictures based on global gradient changes to detect potential areas containing banding artifacts. The values of the mask image are scaled thereby making banding artifact detection possible using gradient operators. The banding artifacts are then identified/detected based on the local gradients.

Elastography can produce quality strain images in vitro and in vivo. Standard elastography uses a coherent cross-correlation technique to estimate tissue displacement and tissue strain using a subsequent gradient operator. While coherent estimation methods generally have the advantage of being highly accurate and precise, even relatively small undesired motions are likely to cause enough signal decorrelation to produce significant degradation of the elastogram. For elastography to become more universally practical in such applications as hand-held, intravascular and abdominal imaging, the limitations associated with coherent strain estimation methods that require tissue and system stability, must be overcome. In this paper, we propose the use of a spectral shift method that uses a centroid shift estimate to measure local strain directly. Furthermore, we also show theoretically that a spectral bandwidth method can also provide a direct strain estimation. We demonstrate that strain estimation using the spectral shift technique is moderately less precise but far more robust than the cross-correlation method. A theoretical analysis as well as simulations and experimental results are used to illustrate the properties associated with this method.

The invention discloses a lane departure warning method based on Hough transform. The method comprises the steps of carrying out region of interest segmentation on an image; graying region of interestimages; carrying out edge detection on the grayed images through utilization of an improved unidirectional gradient operator, thereby obtaining edge images; carrying out binarization processing on the images through utilization of an otsu algorithm; obtaining a candidate straight line set through Hough transform, dividing the candidate straight line set into a left lane straight line set and a right lane straight line set according to slopes, screening the two straight line sets through combination of vanishing point constraint and a random sample consistency algorithm, thereby obtaining optimum left and right lane line parameters; establishing two Kalman filters to track left and right optimum lane lines; and estimating relative lateral offset and a change trend through utilization of the left and right lane line slopes, identifying lane departure and carrying out warning. According to the method, the lane departure problem in driving can be effectively identified, and the robustnessand timeliness of the a lane departure warning system are improved.

Methods and apparatus for identifying objects in an image include processing the image with a gradient operator to produce a gradient magnitude and direction for each pixel. A number of different gradient directions in a portion of the processed image is determined. The portion of the processed image is identified as an object if the number of different gradient directions exceeds a threshold number of gradient directions. The determination of the gradient directions can be simplified by employing a lookup table.

The invention relates to the technical field of image processing, in particular to a remote sensing image registration algorithm. The registration algorithm comprises the following steps of: (1) reading primary remote sensing image data from a storage medium; (2) pre-processing each primary image; (3) performing binarization processing to realize image segmentation; (4) corroding and expanding the binary image; (5) extracting a monotropic edge from the eroded and expanded binary image by the Roberts cross gradient operator; (6) calculating the matching degree of the edge image and finding the optimal matching point; (7) conversing images; and (8) outputting registration information and registration images. The remote sensing image registration algorithm can be applied to multispectral data fusion and remote sensing image mosaicking.

Disclosed in the invention is a pupil positioning method. The method comprises: an image including human eyes is obtained and is used as an input image; according to a preset gradient operator, edge information of the input image is extracted; a hot spot is screened out from the edge information based on a preset area ration and a preset aspect ratio; a candidate pupil light spot in the input image is determined according to gray level distribution of the hot spot and the peripheral area; orientation information of the pupil light spot is determined; according to the orientation information, a radius and a circle center coordinate of a candidate pupil corresponding to the candidate pupil light spot are calculated; a first area is selected based on the radius and circle center coordinate of the candidate pupil and whether pixel value distribution in the first area meets a gray level distribution mode is determined; and if so, the candidate pupil is determined as a real pupil. In addition, the invention also provides corresponding pupil positioning equipment.

The invention provides a defogging method for sea fog image. The sea fog image defogging method comprises the steps of obtaining a gradient operator graph corresponding to each to-be-processed image;Segmenting the gradient operator graph to obtain a target image; Performing morphological expansion and corrosion operation on the target image to obtain a sky region; Obtaining a sky region sea fog image and a non-sky region sea fog image; Dividing a target area based on quadtree search to obtain a global atmospheric light value; Calculating a medium transmission image according to the sky regionsea fog image and the non-sky region sea fog image; Obtaining a sky area defogging image corresponding to the sky area sea fog image and a non-sky area defogging image corresponding to the non-sky area sea fog image; And obtaining a final defogged image according to the sky region defogged image and the non-sky region defogged image. By applying the embodiment of the invention, defogging is carried out on the foggy image, so that the unmanned ship can obtain a clear image, obstacle avoidance target identification can be carried out through the clear image, and the accuracy of obstacle identification is improved.

The invention relates to a pencil sketch simulating method based on an edge tangent stream. The method includes the following steps that firstly, an original image is preprocessed, a bilateral filter is used for performing smoothing on image noise, meanwhile, edge information of an image is reserved, then, a color image is changed into a grey scale image, and therefore a white noise image is generated; secondly, a structure vector field is constructed, a 3*3Sobel gradient operator is used for performing convolution on pixels of the grey scale image in the horizontal direction and the perpendicular direction to generate a gradient field, processing of the edge tangent stream is performed on the basis of the gradient field, so that stroke direction conforms to object features, for an area where the luminance variation is not obvious, threshold value extraction is set, the stroke direction is set to be 45+/-5 degrees, and the vortex stream is removed; thirdly, line integral convolution processing is performed on the white noise image and the structure vector field, and a pencil texture layer is generated; fourthly, a background texture layer is constructed, the stoke direction is rotated anticlockwise by 90 degrees on the basis of the structure vector field, a line integral convolution algorithm is used for generating the background texture layer, the background texture layer is overlapped with the pencil texture layer finally, and the light and shade relation is adjusted to generate a pencil sketch.

The present invention relates to differential phase-contrast imaging of an object (108). When reconstructing image information from differential phase-contrast image data, streak- like artefacts (502) may occur. The artefacts (502) may substantially reduce legibility of reconstructed image data. Accordingly, it may be beneficial for removing or at least suppressing said artefacts (502). Thus, a method (400) for regularized phase retrieval in phase-contrast imaging is provided comprising receiving (402) differential phase-contrast image data of an object (108); generating (404) reconstructed image data of an object (108) and presenting (406) reconstructed image data of the object (108). The differential phase-contrast image and the reconstructed image data comprise a two-dimensional data structure having a first dimension and a second dimension. Generating reconstructed image data comprises integration of image data in one of the first dimension and the second dimension of the data structure. A gradient operator is determined in the other one of the first dimension and second dimension of the data structure and the data structure is employed for reconstructing image data resulting in a reduction of artefacts (500) within the reconstructed image data.

The invention provides a high-robustness real-time lane detection algorithm based on ROI. In order to achieve curve detection, an image is divided into an upper portion and a lower portion, namely a close shot area and a long shot area, and solutions are obtained through Hough and a hyperbolic curve pair model respectively. The whole image is mainly filtered through a transverse gradient operator so that the calculation speed can be increased and the purpose of real-time detection can be achieved; an area of gradient direction angles of a lane boundary image is counted through a sliding ROI window strategy, boundary noise of the abnormal gradient direction angles is eliminated, and therefore the accuracy of lane detection is guaranteed; in the hyperbolic curve pair model adopted in the long shot area, parameters in a close shot model are mainly used as initial parameters, a parameter K is finally determined through a search strategy, and the curve portion is detected. The high-robustness real-time lane detection algorithm is good in robustness in the complex road condition and the complex environment and lane counseling information can be provided in real time.

The invention provides a composite second-order fractional order signal processing-based edge detection method which belongs to the conventional problem in the mode recognition field. The method is a new algorithm for solving an edge by performing gradient operation on all target pixel points in an image by using fractional order signal processing, and comprises the following steps of: generating a gray level matrix for any image; performing the gradient operation on each pixel point in the matrix by using a detection operator to obtain the gradient amplitude of each pixel point respectively; determining an edge position through an obtained zero-crossing image; and if the gradient amplitude of the target pixel point is greater than a threshold value, determining that the target pixel point is an edge point. By the method, smoothing filter preprocessing of the image is saved and the gradient operation is performed by using a fractional order signal processing-based novel composite derivation algorithm; fractional integral in the algorithm suppresses the interference introduced by a fractional differentiation process; and the method has the characteristics of good signal-to-noise ratio, accurate edge positioning and excellent suppression of false edges. The algorithm can be applied to the fields such as automatic target recognition and the like.