98results about How to "Rich image detail" patented technology

The invention discloses a rapid image splicing method based on wide-angle lenses. The method includes the steps that firstly, distortion correction is conducted on images acquired through the wide-angle lenses; secondly, cylindrical surface mapping is conducted on the images on which distortion correction is conducted, and a two-dimensional discrete coordinate mapping table is generated for each wide-angle lens; thirdly, the two-dimensional discrete coordinate mapping table of each wide-angle lens is loaded, and distortion correction and cylindrical surface mapping are sequentially conducted on the distortion images acquired through the wide-angle lenses according to the tables; fourthly, translation parameters between adjacent visual angle images on which cylindrical surface mapping is conducted are acquired, and the adjacent visual angle images are registered; fifthly, the registered adjacent visual angle images are fused, wherein Laplacian pyramids are established for registered adjacent visual angle image sequences respectively, linear fusion is conducted on a high-frequency pyramid image overlaying region and a low-frequency pyramid image overlaying region, and then a fused image of the adjacent visual angle images is obtained. The method has the advantages of being high in splicing speed and real-time performance.

The invention relates to the digital image processing field and particularly relates to a white balance and dark primary color adaptive histogram underwater image enhancement method. The method comprises steps that step 1, color correction for an image is carried out through utilizing a dynamic threshold white balance algorithm; step 2, a dark channel map of the image after color correction is solved through utilizing a dark channel model; step 3, a weight factor of block images is calculated; step 4, underwater image processing is carried out through utilizing a CLAHE method; step 5,the weight factor and gray mapping relationship tables are utilized to acquire a final gray mapping relationship table of a block through fusion calculation; and step 6, a bilinear interpolation algorithm is utilized to calculate a gray mapping value corresponding to each pixel in block images one by one, enhanced block images are acquired, interpolation calculation for pixels connected between the adjacent block images is carried out, and images of underwater images after enhancement are acquired. The method has advantages of simple calculation and strong timeliness.

The invention discloses an image super-resolution reconstruction method based on a self-attention high-order fusion network, and the method is characterized by including the following steps: 1), building a reconstruction model; 2) performing CNN network feature extraction; (3) performing self-attention branch feature extraction in a self-attention module, (4) performing trunk branch feature extraction in the self-attention module, (5) performing feature high-order fusion and (6) performing image reconstruction. The method can effectively solve the problem of extra calculated amount caused by preprocessing, and more texture details can be recovered to reconstruct a high-quality image.

The invention discloses a recursive residual attention network-based image super-resolution reconstruction method. The method is characterized by comprising the following steps of: 1) preprocessing data; 2) establishing a reconstruction model; 3) extracting features of a first residual attention module of a residual attention network branch; 4) extracting features of a first recursive module of arecursive network branch; 5) fusing the features; and 6) reconstructing an image. According to the method, noises caused by preprocessing can be solved, more high-frequency information can be obtainedto enrich the image details, the network parameters can be decreased, new parameters are not increased while the layers are increased; and the precision of super-resolution reconstruction can be improved.

The invention discloses an image restoration method based on a double-edge wave filter and margin deconvolution, comprising the following steps: (1) obtaining a restoration image with rings by adopting a standard RL algorithm to carry out higher iteration according to an image to be restored and a known fuzzy core; (2) adopting a double-edge wave filter to carry out filtering operation of the restoration image with rings to obtain the restoration image without rings; (3) carrying out the deconvolution on the restoration image without rings by the known fuzzy core to obtain a re-fuzzy image; (4) obtaining a restoration image detail by adopting the margin deconvolution by the re-fuzzy image obtained by the step (3) and the image to be restored, and adding the restoration image detail with the restoration image after eliminating ring to obtain a new restoration image containing ring; and (5) repeating the steps (2)-(4) for many times to obtain a final restoration image. The image restoration method can effectively inhibit middle and high frequency rings in the restoration image of the iterative algorithm, remain rich image details and obtain high-quality restoration images.

The invention discloses a color image enhancement method based on improved multi-scale Retinex. Firstly an unprocessed color image is read, a conventional multi-scale Retinex algorithm is improved, and the average brightness component of a current pixel is calculated by adopting bilateral filtering so that brightness component estimation based on the bilateral filtering method is obtained; secondly local contrast enhancement is performed according to the size relation of brightness of each pixel point and domain average brightness thereof; and finally the brightness image after enhancement is compared with the brightness component of the original image, RGB information of the recovered and enhanced image is linearly adjusted, and color image enhancement is obtained finally. The color image enhancement method has advantages of image detail recovery, elimination of the "halo" phenomenon, meeting the human eye visual effect and short operation time.

The invention relates to a multi-exposure image fusion method, which solves the technical problems of incomplete detail information retention and incomplete ghosting elimination, and comprises the following steps of: 1, constructing an initial weight map by using exposure brightness and chroma information of a multi-exposure image sequence; 2, firstly, moving object detection is conducted on the low-dynamic image sequence, a moving area is calculated, and then a ghosting eliminating method based on super-pixel segmentation is used for eliminating ghosting; 3, completing ghosting correction onthe initial weight map in the step 1; 4, performing normalization processing on the weight map after ghosting correction in the step 3; and 5, constructing a weighted Gaussian pyramid according to theweight graph in the step 4, constructing a Laplacian pyramid of a low-dynamic image sequence, defining detail gain items, calculating a fusion pyramid, and performing image reconstruction according to the fusion pyramid to obtain a fusion image HDR. The problem is well solved, and the method can be used for image processing.

The invention relates to a turbulence-degraded image blind restoration method based on dark channel and Alternating Direction Method of Multipliers. The method includes the following steps: firstly on the basis of the multiple dimension theory, in each dimension, applying dark channel prior constraint on an image, applying sparse constraint and energy constrain on a point spread function, then using the coordinate descent method and conducting alternating iteration to estimate a fuzzy kernel and the image in current dimension, if the dimensions arrive at the maximum thereof, a final estimated fuzzy kernel is obtained, finally, in combination with a total variation model, using a derivative Alternating Direction Method of Multipliers to make details of the image restored quickly. According to the invention, the method, by using the dark channel prior information of a clear image as a constraint item, can help a cost function to converge to a clear solution in the iteration process, addresses the susceptibility of obtaining a fuzzy solution by using tapered prior information under the Maximum posterior probability in current blind restoration algorithm, such that the method herein can restore more image details, has less ring effect, and effectively increases restoring quality.

The invention provides an image noise reduction method and device, electronic equipment and a storage medium. The method comprises the following steps: determining exposure parameters according to a shooting scene, shooting by adopting the exposure parameters to obtain a shot image, determining a matched noise reduction model according to the equipment attribute of imaging equipment for shooting, and inputting the exposure parameters and the shot image into the noise reduction model to perform noise reduction on the shot image, wherein the noise reduction model is learned to obtain a mapping relation between the exposure parameters and the noise characteristics. According to the method, after the matched noise reduction model is determined according to the equipment attribute of the imaging equipment used for shooting, noise reduction processing is carried out on the shot image shot in the current shooting scene, and more image details are reserved while the image purity is ensured. Besides, the shot image is subjected to noise reduction through the noise reduction model matched with the equipment attribute, the noise characteristic can be better identified, the purpose of noise reduction is achieved, the noise reduction effect of the image is improved, and a clearer image can be obtained.

The invention provides a multi-grade backlight detection method and a device. The multi-grade backlight detection method comprises the steps of S101, blocking a target image and acquiring a brightness value of each image block; S102, carrying out detail analysis on the image blocks and converting the brightness value of each image block into a brightness evaluation value of the image block according to a brightness score mapping table; S103, synthesizing the evaluation values of all image blocks so as to acquire the backlighting degree of the whole image; and S104, guiding exposure selection according to the backlighting degree so as to provide an appropriate image for HDR synthesis. According to the invention, backlighting scenes, overexposure / underexposure areas and the degrees thereof are accurately differentiated, and the backlighting degree of the image is evaluated; exposure selection is guided according to the backlighting degree of the current image, appropriate exposure enables image details to be richer, and an image with high quality can be acquired through HDR synthesis.

The invention provides an image preprocessing grayscale space division method based on average filtering. The image preprocessing grayscale space division method comprises the following steps that weighted average operation is performed on three components in different weights according to the importance of RGB three components and other indicators; histogram correction is preferably performed by adopting histogram equalization, the histogram of an original image is corrected into uniform distribution by utilizing a grayscale transformation function, and then histogram equalization is performed; average filtering is performed on the image, a global threshold method is adopted, the method of one global threshold T is used only in the binarization process, the grayscale value of each pixel of the image is compared with T, and foreground color is taken if the grayscale value of each pixel of the image is greater than T; or background color is taken; and a center point P with the maximum number of corresponding pixels within L grayscale range acts as an initial class average value. Each pixel is inspected in the ith time of iteration, distance between each pixel and the average value of each grayscale is calculated, and the class, of which the average value is closest to the pixel, is assigned to each piexl. A new clustering center is calculated as for j=1,2,...l, the class average value is updated and all the pixels are inspected one by one.

The invention discloses lamination imaging technology for carrying out lamination reconstruction constraint through pre-lighting imaging. The lamination imaging technology is characterized by, obtaining a formed image of a periphery image of a sample to be tested through pre-lighting as a known condition for lamination recovery; then, obtaining diffraction images of the sample to be tested and the periphery image through lamination scanning, and simultaneously substituting the diffraction images and pre-lighting images into a lamination iterative algorithm based on the pre-lighting imaging designed in the invention for processing; and finally, carrying out reconstruction to obtain a complex amplitude image of the sample to be tested. Constraint is carried out on lamination recovery through the pre-lighting imaging mode, so that recovery quality of the sample to be tested on the aspect of complex amplitude, especially on the aspect of phase, can be improved greatly, and experiment operation is simple; and compared with the conventional lamination imaging technology, the lamination imaging technology in the invention can obtain a better recovery result without extra cost. Meanwhile, the technology has obvious improvement on resistance to noise and aperture position offset.

The invention discloses a graph regularization sparse coding-based magnetic resonance super-undersampled K data imaging method which comprises the following steps: (a) performing graph regularization sparse coding expression on a double-layer Bergman iteration frame to obtain an image sparse model; (b) updating a learning dictionary and a sparse coefficient on the inner-layer iteration of double-layer Bergman iteration by introducing an auxiliary variable and an alternate solving technology; (c) performing image updating on the outer-layer iteration of the double-layer Bergman iteration to obtain an imaging result by utilizing a part of super-undersampled K data as constraints. According to the method, a proximity graph is established to code local structural data and dig geometric data constraints thereof by introducing adaptive dictionary learning into graph regularization sparse coding, so that image data can be sparsely expressed better; in addition, an image with more complex local geometric characteristics can be processed, a local image structure can be effectively captured, more image details can be recovered, and an obtained image result is higher in fidelity.

The invention discloses a light field camera refocusing method based on orthogonal lattice point conversion. Firstly, a light field camera is used for shooting a uniform white wall or a uniform whitelight plate to obtain a white image, an inter-class variance method is used for solving an optimal threshold value of the white image and carrying out binaryzation, and a hexagonal-orthogonal latticepoint conversion method is used for correcting a white image contour area; on the basis, two-dimensional coordinates of a central point of a white image contour are calculated by using image invariantmoment, four-dimensional information of a light field is represented by using a biplanar parametric representation method, and finally, refocusing of the light field image is realized by using a light field imaging principle and an energy function of light. The refocused image is rich in details and high in calculation efficiency, can be used as an image post-processing algorithm to stably operate in the current infrared monitoring equipment, and improves the scene depth information collected by the current equipment.

The invention relates to a color enhancement method suitable for remote sensing images, relates to the remote sensing image processing technology field and solves problems of limitation, poor robustness, poor image enhancement effect and noise enhancement existing in a remote sensing image color enhancement method in the prior art. The method comprises steps that to-be-processed remote sensing images and color images are inputted; color migration processing for the to-be-processed remote sensing images is carried out; deblocking processing of the images after color migration is carried out; fuzzy kernels of the images after deblocking processing are calculated; Wiener filtering is utilized for the deblocked images for deblurring; the images are subjected to Laplacian sharpening after deblurring, and the final result images are obtained. The method is advantaged in that color richness of the remote sensing images can be substantially improved, the image contrast is improved, the edge information is strengthened, and the enhanced images have the good visual effect.

The invention provides a video super-resolution method based on non-local regularization. The method comprises the realization steps: (1) a video X is input; (2) bicubic interpolation amplification is carried out on a frame low-resolution image in the video X to obtain the amplified image, and high-pass filtering is carried out on a high-resolution image closest to the amplified image to obtain low and high frequency components; (3) image blocks are extracted from the amplified image and the low frequency component, the non-local regularization is carried out on the amplified image blocks, and k-means clustering is carried out on the low frequency image blocks; (4) the new amplified image blocks and a cluster center are compared, the most similar clusters are found out, and the similar low frequency image blocks are found out in the most similar clusters; (5) the corresponding high frequency image blocks are found out according to the similar low frequency image blocks, and non-local weighting is carried out on the high frequency image blocks to obtain a reconstructed high-resolution image; (6) the steps from (2) to (5) are repeated for each frame low-resolution image of the video to obtain a high-resolution video. The method is used for enhancing or restoring the video.

The invention relates to a method of improving out-of-gamut color overlap mapping. The method includes: step 10, converting RGB number values of a color point P in a source gamut into L*a*b* values ofa Lab color space; step 20, determining a hue plane where the color point P is located, and further determining values of a hue angle H, color saturation C and brightness L; step 30, respectively representing intersection points of a straight line, which is defined by the color point P and a color point Lm on a brightness axis of the hue plane, and a boundary of a target color gamut and a boundary of the source color gamut as a color point PC and a color point Ps, and combining the color point Lm and a parameter alpha to determine a reference color point Pi; step 40, judging whether the colorpoint P is inside or outside a line segment |LmPi|, if the color point P is inside the line segment |LmPi|, mapping the color point P to the target color gamut, then still generating the color pointP, and entering a step 50, and if the color point P is outside the line segment |LmPi|, mapping the color point P to the target color gamut, and then generating a color point P'; and step 50, converting L*a*b* values of the color point P, after being mapped to the target color gamut, into RGB number values. The method of improving out-of-gamut color overlap mapping of the invention enables variouscolors to no longer be mapped to the same color point.

Disclosed is a gray scale space division method of image preprocessing based on mean value filtering. The method comprises the following steps: according to importance of RGB three components and other indexes, performing weight average operation on the three components by means of different weights; histogram correction is firstly carried out by use of histogram equalization, a histogram of an original image is corrected to be uniformly distributed by use of a gray scale conversion function, then the histogram correction is carried out; carrying out the mean value filtering on the image by use of a global threshold method in which only one global threshold T is used in a binary process, the gray scale value of each pixel of the image is compared to T, if the value is greater than T, the value is taken as a foreground color, and otherwise, the value is taken as a background value; taking center points P with maximum quantities of corresponding pixels in L gray scale grade scopes as initial mean values; and during i-th iteration, investigating each pixel, calculating the spacing between each pixel and the mean value of each gray scale grade, endowing each pixel with a mean value to a closest cluster, for j=1, 2, ...1, calculating a new cluster center, updating a cluster mean value, and investigating all the pixels one by one.

The embodiment of the invention provides an image processing method and device of a terminal and terminal equipment. The terminal comprises a black-and-white camera and a color camera, and the method comprises the steps that after a photographing function of the terminal is operated, in the preview process, a color RAW image is collected through the color camera, and a black-and-white RAW image is collected through the black-and-white camera; after a shooting instruction is obtained, the color camera and the black and white camera are controlled to be restarted, the color camera is instructed to output images in a pixel merging mode, and the black-and-white camera is instructed to output images in a full-size mode; RAW domain processing is performed on the color RAW image acquired by the color camera to obtain a color YUV image; the black-and-white RAW image collected in the preview process is processed through a second Bayer domain processing algorithm link to obtain a black-and-white YUV image; the color YUV image and the black-and-white YUV image are fused to obtain a target color image, so that more image details are reserved, and the color rendition degree of the target color image is improved.

The invention discloses a noise elimination method for an impact noise image. The method specifically comprises the following steps: establishing a mark matrix F of a noise polluted image I; dividing the image I and the mark matrix F into M*N grids according to impact noise pollution density rho; extracting an image block T<m,n> and a mark block L<m,n>, which are composed of pixels in the (m,n)th grid in the image I and the matrix F respectively; establishing a pollution pixel set E and a non-pollution pixel set P in the image block T<m,n> through traversing elements in the mark block L<m,n>; obtaining a linear predication system parameter Psi; according to the linear predication system parameter Psi and an Euclidean distance matrix De, calculating to obtain a pollution pixel value shown in the specification; carrying out matrix transposition operation on the pollution pixel value shown in the specification to obtain a noise elimination pixel value E; obtaining a noise-eliminated image block T<m,n>; writing the image block T<m,n> back to the image and replacing pixels of the image in the (m,n)th grid; and traversing all the grids in the image I.

The present invention relates to a FCM image segmentation method and apparatus, and solves the technical problem of failure of effective completion of image segmentation by a general fuzzy clustering method. The initialization process includes: forming a set S1 of weighted local grayscale items of all neighborhood points relative to sample points; and the iteration process comprises: in current iteration, forming a set S2 of weighted local distance items of the current iteration of all the neighborhood points relative to the sample points through grayscale value differences between the neighborhood points of the sample points and a clustering center Viter-1 of previous iteration, the membership Uiter-1 of the previous iteration and a spatial Euclidean distance, and forming the membership Uiter of the current iteration through the set S1 of the weighted local grayscale items, the set S2 of the weighted local distance items of the current iteration, and the clustering center Viter-1 of the previous iteration; and forming the clustering center Viter of the current iteration according to the membership Uiter of the current iteration.

The invention relates to an infrared image and visible light image fusion method, in particular to an infrared and visible light image fusion method combining improved NSCT transformation and deep learning. The invention aims to solve the technical problem that an existing infrared image and visible light image fusion method is difficult to carry out rapid and effective image fusion, and provides an infrared and visible light image fusion method combining improved NSCT transformation and deep learning. According to the method, infrared and visible light image fusion is carried out in combination with improved NSCT transformation and deep learning to generate a fusion result image conforming to a human eye vision system. An improved non-subsampled contourlet transform NSCT is adopted to perform adaptive decomposition on a to-be-fused image, and deep learning is utilized to determine a fusion weight for a sub-band image of a corresponding scale. By adopting the method of the invention, the infrared image and the visible light image are fused, image details and spectral information can be enriched, the resolution is improved, and people can generate more complete scene perception.

The invention aims at providing a Context model and dual-tree complex wavelet transform-based denoising method for an underwater sonar image, comprising the step of: performing dual-tree complex wavelet decomposition on the underwater sonar image, wherein the low-frequency approximate component obtained by conducting the four-layer dual-tree complex wavelet decomposition on the image keeps invariable; performing denoising processing on the high-frequency component of the image; and performing dual-tree complex wavelet inverse transform on the processed complex wavelet coefficients so as to obtain the final image subjected to denoising. According to the denoising method, the similarity of energies among the dual-tree complex wavelet coefficients is measured by adopting the Context model, the coefficients with the similar energies are classified, different thresholds are determined aiming at each class of the coefficients, and denoising is realized by combining with a soft-threshold function. Threshold section is optimized; and more image details are reserved while the noise is removed, so that the phenomenon that the coefficients are excessively killed is suppressed.

The invention discloses a method and a system for measuring a large part. The method comprises the following steps: determining a conversion relationship between a pixel coordinate system and a worldcoordinate system of each camera in a camera array; aiming at each camera in the camera array, the camera array is divided into a plurality of cameras; acquiring a part image corresponding to the camera; and determining a first image corresponding to each camera according to the part image, splicing the first images corresponding to all cameras according to the conversion relationship between thepixel coordinate system and the world coordinate system of each camera to obtain a second image, and finally determining the size of the part according to the second image. According to the method formeasuring the large part, image splicing of the camera array and an image super-resolution technology are used; the high resolution of the image is ensured while the view field range of the image isexpanded, more image details are provided, the measurement accuracy of the large part is improved, and the method can be applied to a production line to automatically measure the size of the part andhas a high automation degree.

The invention discloses an image fusion method based on fast BEMD and deep learning, and belongs to the technical field of image processing. The method comprises the following steps: carrying out multi-scale decomposition on a to-be-fused image by using a fast BEMD to obtain a two-dimensional empirical mode decomposition component (BEMC) with the frequency from high to low, and fusing the components respectively, and finally obtaining a fusion result graph through BEMD reconstruction. An image fusion rule based on deep learning is designed by utilizing the characteristic that deep learning canextract image features. Experiments show that the fusion result graph based on the fusion method has the optimal visual effect and meets the visual perception of human eyes.

The invention relates to the technical field of image processing, and provides an image super-resolution method and device, terminal equipment and a computer storage medium. The method comprises the following steps: extracting image features from a to-be-processed image, inputting the image features into a texture feature extraction module to extract texture features of the image, processing the texture features by adopting a plurality of convolution layers to obtain structural features of the image, fusing the texture features and the structural features of the image, and finally recovering the fused texture features and structural features into an up-sampled image; and obtaining a corresponding super-resolution image. Compared with a traditional image super-resolution method, the image super-resolution method has the advantages that richer image details can be obtained by fusing the texture features and the structural features of the image, and therefore the image super-resolution effect is improved.

The embodiment of the invention provides an image processing method and device of a terminal and terminal equipment. The terminal comprises a black-and-white camera and a color camera, and the method comprises: after a photographing function of the terminal is operated, in the preview process, collecting a color RAW image through the color camera, and collecting a black-and-white RAW image through the black-and-white camera; obtaining a shooting instruction, controlling a color camera to shoot a current shooting scene, and obtaining a color RAW image of the shooting scene; performing RAW domain processing on the color RAW image to obtain a processed color RAW image; processing the processed color RAW image to obtain a color YUV image; processing the black and white RAW image collected in the preview process through a second Bayer domain processing algorithm link to obtain a black and white YUV image; fusing the color YUV image and the black and white YUV image to obtain the target color image, so that more image details are reserved, and the color rendition degree of the target color image is improved.

The invention discloses a non-rigid multimode medical image registration model establishment method and application thereof, and belongs to the field of medical image registration, and the method comprises the steps: establishing a generative adversarial network GAN_dr, a generator G_dr being used for generating a structural characterization graph for deformation recovery, and a discriminator D_dr being used for judging whether the structural characterization graph generated by the G_dr has effectively recovered deformation; calculating structural representation graphs of the reference image, the floating image and the actual registration image in each sample in the medical data set, and training the GAN_dr by using a calculation result; establishing a generative adversarial network GAN_ie, wherein a generator Gie takes the structural representation graph as input and is used for estimating a registration image, and a discriminator D_ie is used for judging whether the estimated registration image is consistent with an actual registration image or not; using the trained G_dr to generate a structural representation diagram of deformation recovery corresponding to each sample in the medical data set, and training the GAN_ie; and connecting the trained G_ie to the G_dr to obtain a registration model. According to the invention, rapid and accurate matching of medical images can be realized.

The disclosure provides an image processing method and device for a TOF depth image, equipment and a storage medium, and relates to the technical field of camera shooting. The image processing methodfor the depth images shot by the time-of-flight TOF assembly comprises the following steps: respectively obtaining multiple frames of TOF depth images shot based on different configuration values of configuration parameters of a shot scene; respectively extracting effective information in each frame of TOF depth image; and fusing the multiple frames of TOF depth images based on the effective information extracted from each frame of TOF depth image to obtain a high dynamic range HDR depth image. According to the method, the problems of low TOF depth image ranging precision and the like are solved, and the obtained HDR depth image contains richer image details and a higher dynamic range.

The invention discloses a near-distance imaging method and a mobile terminal. The method includes the steps: judging whether a photographed object is in an object distance range capable of clearly imaging by the mobile terminal or not to determine whether a black and white fixed focus camera is started or not; starting the black and white fixed focus camera, and executing shooting operation if thephotographed object is in the object distance range. Under application scenes of near-distance topic shooting, the mobile terminal can present clear topic shooting images in the shortest time according to rapid focusing characteristics of the black and white fixed focus camera, the mobile terminal further present more image details according to the characteristics of large sensing areas and highpixel size of a black and white sensor in the black and white fixed focus camera, so that the quality of black and white images shot by the black and white fixed focus camera is higher than that of acolor image shot by a variable focal camera under the condition of the same sensor size, topic shooting images are conveniently converted into test question contents, and the differentiation degree ofthe topic shooting images is improved. According to the method, near-distance imaging quality and efficiency can be improved.