47results about How to "Rich texture information" patented technology

The invention which relates to an optical instrument and detection imaging technique discloses a large field depth iris image acquisition system and a method based on a plurality of fixed-focus cameras. The system comprises an iris image collecting device and an image recognition device. The method comprises the following steps: respectively imaging the iris by means of the fixed-focus cameras having different imaging object distance ranges and producing a multi-path fixed-focus image results; choosing one or a plurality of multi-path fixed-focus image results through distance measurement or image quality evaluation method; carrying out image processing and identification of chosen qualified iris image data. The invention effectively integrates the imaging ranges of different fixed-focus cameras in the mode of self-adapting switchover, thereby taking advantages of the cameras and at same time removing disadvantages of the cameras, enlarging the integral imaging range of the acquisition system, and improving the feasibility of an iris recognition system. Compared with the prior iris imaging method adopting a zoom camera, the invention has the advantages of low cost, quick and noiseless focusing.

The invention discloses a dark-channel-prior-method-based non-uniform-light-field underwater target detection image enhancement method. The method comprises: an original color image of a non-uniform-light-field underwater detection target is obtained, and a corresponding transmission graph is calculated by using a minimum value filter algorithm; the original color image is converted into an original gray-scale image and the original image is divided into a bright region, a normal region, and a dark region; illumination vectors of all regions are calculated; with a guiding filter method, a guiding filter output image of the original color image is calculated on the condition of the transmission graph; recovered enhanced images of regions, with different brightness values, of the original image are calculated; and quantitative evaluation is carried out on a dodging processing recovered image in terms of a mean value, a variance, a contrast ratio, and an information entropy until the recovered enhanced image meets an index requirement. According to the invention, because of the guiding filter method with illumination vector weighting, smooth balancing processing is carried out on regions with different brightness values, so that the overall coordination of the processed image is enhanced and the texture information is enriched.

The invention provides an image fusion method and a device based on NSST and adaptive dual-channel PCNN. NSST is used as a multi-scale transform tool to decompose the image, NSST removes the sample operation in the decomposition phase, and has translation invariance, at the same time, a localize small size shearing filter can avoid spectral aliasing, so that the effect of image decomposition and reconstruction is better, using an adaptive simplified two-channel PCNN model with three kinds of link strengths, through three different link strengths, the features of the image are integrated, the infrared and visible images are fused effectively, the contrast is high, the texture information and detail information are well preserved, and the image transition is natural, no artifact phenomenon occurs.

The invention belongs to the technical field of image processing and pattern recognition, particularly relates to a three-dimensional face reconstruction method, system and device for a fine structure, and aims to solve the problem of poor three-dimensional face reconstruction precision. The method comprises the steps of obtaining a to-be-reconstructed two-dimensional face image; obtaining a three-dimensional space transformation function and an initial three-dimensional face shape; performing spatial transformation on the initial three-dimensional face shape, and mapping each pixel of an image face region to a UV texture space of a 3DMM model to obtain a UV texture map; obtaining a UV visible image and extracting features to obtain an attention feature map; mapping each point of the initial three-dimensional face shape to a UV texture space to obtain a UV shape graph; multiplying the attention feature map and the UV texture map, and adding the multiplied attention feature map and UV texture map to the UV shape map; and obtaining the update amount of each point of the 3DMM face model, and adding the update amount to each point corresponding to the initial three-dimensional face shape to obtain a three-dimensional reconstruction result. According to the invention, the precision of face model reconstruction is improved.

The invention discloses an adaptive gradient gain underwater image enhancement method based on a target imaging model. The method comprises the following steps of performing global background illumination vector estimation; dividing an integral image into a plurality of blocked images with the same dimension; performing optimized transmissivity estimation; obtaining an estimated value of the to-be-improved transmissivity; calculating a dehazing clear image which corresponds with an original observed image; calculating a gradient image which corresponds with the original observed image; obtaining an enhanced image with added gradient information based on an original observation image; and performing quantitative evaluation on the enhanced image. According to the method of the invention, theself information of the image is detected by means of a single inhomogeneous-brightness low-signal-to-noise-ratio low-contrast underwater target; de-noising enhancing processing, adaptive gain and integrated quantitative evaluation index evaluation are performed on the image for obtaining an adaptive gradient gain enhanced image based on a target imaging model. The adaptive gradient gain underwater image enhancement method can sufficiently utilize the abundant gradient information of the image for realizing image enhancement processing, thereby improving visual quality of the processed imageand realizing abundant texture information.

The invention discloses a single-frame resolution ratio reconstruction method based on sparse coding and combined mapping. The method comprises the following steps that: processing an initial high-resolution training set image to obtain an expanded high-resolution feature block sample and an interpolated medium and high resolution feature block sample; training an obtained feature sample, obtaining a dictionary atom as a clustering center, and clustering samples by the center; according to a corresponding relationship among different resolutions, solving the mapping matrix of each cluster; on the basis of the low-resolution image processing way of the training set, processing an input low-resolution test image, and solving the sparse coefficient of the low-resolution test image by the dictionary atom obtained by training; taking the sparse coefficient as a weight, taking each mapping matrix obtained by clustering as a combined element, carrying out matched combination to obtain a mapping relationship required by image reconstruction, and directly multiplying the mapping matrix by an interpolated medium and high resolution feature block to obtain a high-resolution feature block; and carrying out overlap removal and block fusion, and adding original low-frequency information to obtain the reconstructed high-resolution image.

The invention relates to an image fusion system based on a low-light and infrared sensor. The system is divided into a low-light and infrared sensor information acquisition module, an FPGA core processing module, a DSP co-processing module and an output display module; video images collected by the external low-light and infrared sensor are input to a device through an interface; the input video images are subjected to image preprocessing including image denoising, image enhancement and image registration, wherein image denoising and image enhancement are low in delay requirement and are achieved in an FPGA chip, the computation amount required for an image registration algorithm is large, and image registration is achieved in a DSP chip; the pre-processed video images are fused through the FPGA chip, wherein grayscale fusion and pseudo-color fusion are included; an FPGA control logic can select one path of images to be output through the interface, wherein one path of images include asingle low-light video, a single infrared video, a grayscale fused video and a pseudo-color fused video. The image fusion system is high in precision, low in delay and good in expansibility and has important practical significance on nighttime observation and exploration application.

The invention discloses an underwater target detecting image enhancement method with contrast limited adaptive histogram equalization. The method comprises the following steps of calculating a four-directional Sobel edge detector of a gray image that corresponds with an original colorful image, a gradient image and an adaptive gain function; transforming the original color image from an RGB spaceto an HIS space through nonlinear transformation; performing enhancement processing on a brightness vector in the HIS space image by means of a contrast limited adaptive histogram equalization algorithm; transforming the enhanced HIS space image to the RGB space; performing generalized bounded multiplication operation based on an adaptive gain function on an R component, a G component and a B component in the enhanced RGB image, thereby acquiring the enhanced image based on the gradient information of the original image; performing image displaying after enhancement; and performing quantitative evaluation on the enhanced image. The underwater target detecting image enhancement method can sufficiently use the texture of the original image for realizing image enhancement processing, therebyimproving visual quality of the processed image and obtaining abundant gradient information.

The invention discloses a face deception detection method and system based on multi-scale illumination invariance textural features. The method comprises the following steps: framing a video, extracting a face image, and carrying out channel separation to obtain a color channel graph; obtaining an illumination invariant texture feature map through an illumination separation texture reservation module, performing normalization, combining the normalized illumination invariant texture feature map with the color channel map to obtain face features, and performing data enhancement to obtain to-be-trained input features; constructing a multi-scale texture module by using the center difference convolution of multiple receptive fields, and embedding the multi-scale texture module into a lightweight network to construct a lightweight multi-scale texture network; weighting the pixel-level loss and the cross entropy loss into total loss; inputting the input features into a lightweight multi-scaletexture network to learn deception features of texture naturalness; updating network parameters according to the loss function, and storing a network model and the parameters after training is completed; and predicting a classification result according to the stored model. The deception features of texture naturalness are accurately extracted, the generalization performance of the model is effectively improved, and the storage and calculation consumption of deployment is reduced.

An image fusion method and apparatus based on potentially low rank representations and structural tensors, Aiming at the shortcomings of the existing dual-channel PCNN model in solving the problem ofinfrared and visible image fusion, The LSR operator based on LatLRR algorithm is used as the external stimulation of PCNN model, to solve the large difference between the infrared image and the visible image, and the TSV operator based on structural tensor singular value decomposition is used as the link strength, Characterization of the image feature changes, to some extent, can solve the infrared image and visible image spectral differences resulting in lower contrast of the fusion image, while retaining a relatively complete wealth of texture information and detail information, and image transition naturally.

The invention discloses an underwater target detection image enhancement method adopting self-adaptive multi-scale dark channel prior. The method comprises steps as follows: acquiring an original color image of an underwater target; calculating a light vector of the original image; calculating a dark channel image of the original image; calculating transmission graphs corresponding to different-scale windows of the dark channel image; calculating steerable filter output images corresponding to the transmission graphs in different-scale windows of the original color image with a steerable filter method; primarily selecting several most appropriate windows in the minimum filter scale according to the condition that the error between each steerable filter output image and the corresponding transmission graph is minimum; performing weighted average on the screened-out steerable filter output images; calculating a restored and enhanced image of the original image according to the dark channel prior theory; performing quantitative evaluation on the restored image J in the mean value, the variance, the contrast ratio, the information entropy and other aspects. With the adoption of a steerable filter function applied in the method, texture and smoothness balancing processing on the image can be realized, so that the processed image has improved visual quality and rich texture information.

The invention provides an image fusion method and device based on potential low rank representation and NSST. A source image is discomposed into multi-scale and multi-direction by an NSST including asmall-size shearing wave filter. Because NSST eliminates the sampling operation in the decomposition phase, it has translation invariance. At the same time, in the phase of directional localization,localized small-scale shearing filter can avoid spectrum aliasing phenomenon, and make the image decomposition and reconstruction effect better. For the low frequency component, LatLRR can extract thesalient features from the data robustly, and is more robust to noise, so it can identify the salient objects and regions in the image precisely. For the high frequency component, because the averagegradient represents the change of image gray value, it can reflect the details of image edge and texture, so the average gradient operator can better express the image features and achieve better fusion effect.

The invention provides an image fusion method and a device of total variational and structural tensor. The total variational TV model obtains a target gradient field of the fusion image by using the structural tensor of the image, and utilizes Soft. The Max function softens the weight of the gradient field, the fusion gradient field is put into the regular term of TV model, ensuring that the gradient field of the fusion image is close to the target gradient field, The splitting Bregman algorithm is used to solve the new TV model, The final fusion image can be obtained. By adopting this scheme,the features of the image can be expressed better, the texture information of low-light level and the salient features of infrared image can be combined perfectly, and the better fusion effect can beobtained. The abundant texture information and detail information can be completely retained, and the image transition is natural.

The invention discloses a non-illumination face image reconstruction method and a non-illumination face image reconstruction system, which belong to the field of image reconstruction. The method comprises the following steps of: acquiring a face reflectivity image from a to-be-recognized original face image; establishing an illumination subspace according to a face reference image under different illumination conditions and projecting the to-be-recognized original face image and the face reflectivity image to the illumination subspace to obtain corresponding reference images respectively; establishing corresponding texture subspaces according to the corresponding reference images and projecting the to-be-recognized original face image and the face reflectivity image to the corresponding texture subspaces to obtain a projection coefficient of a base image in an original texture subspace and a base image in the reflectivity texture subspace respectively; and reconstructing a non-illumination face image according to the projection coefficient of the base image in the original texture subspace and the base image in the reflectivity texture subspace. The method and the system are mainly used for non-illumination face image reconstruction.

The invention provides a detection method for a frozen lake in a remote sensing image. The method comprises a texture extraction step, an energy analysis step, and a result output step. At the texture extraction step, texture feature extraction is carried out on a to-be-detected remote sensing image and a fractal dimension graph is obtained. At the energy analysis step, an energy characteristic of a local image block of the fractal dimension graph is analyzed by using a regional contrast ratio and a local entropy feature. At the result output step, a detection result is outputted based on the analyzed energy characteristic. In addition, the invention also provides a detection system for a frozen lake in a remote sensing image. According to the detection method and system, the specific texture characteristics of a frozen lake are analyzed and an interference source is extracted from a background, so that the effect in a low-noise-ratio / signal-to-clutter-ratio scene is good. Moreover, the method can be operated quickly and is easy to program; and the method and system can be applied and operated in various systems in a cross-platform manner.

The invention discloses an image sharpening method based on an SSE2 (Streaming SIMD extensions 2) instruction set. The method comprises the following steps of separating a component A, a component G, a component R and a component B from loaded image data, and obtaining a channel differential component with the component G as a reference component; carrying out high-pass filtering and margin gain amplification processing, and adopting a threshold shrinkage way and a value adjustment way to control a flat region of an image and a high-frequency component in an enhanced component G; adding the high-frequency component in the enhanced component G to an original component G to obtain a sharpened component G; summing the sharpened component G and the channel differential component to obtain a sharpened component R and a sharpened component B; and carrying out staggered arrangement on the component A, the sharpened component R, the sharpened component G and the sharpened component B through register shifting, writing into a buffer zone and then outputting. According to the method, when the image is sharpened, color space conversion is not needed, so that the high-efficient color image sharpening processing is achieved.

The invention discloses a self-adaptive gain underwater image enhancement method based on an HSI space optical imaging model. The self-adaptive gain underwater image enhancement method comprises the steps of estimating a global background illumination vector; estimating a background illumination covering layer vector; calculating a de-noising recovery image corresponding to the original observation image; converting the original color underwater image into a grayscale image; converting the gray level image into a gray level image capable of keeping the edge of the image, and extracting a gradient image corresponding to the image; fusing the de-noised restored image and the gray level image which maintains the edge characteristics; converting the fused color image into a color space, and separating brightness information from color information of the image; calculating the brightness information to obtain an enhanced image based on the rich gradient information of the original image; converting the enhanced image into a color space; and performing quantitative evaluation on the enhanced image. The four-direction Sobel edge detector used in the invention can fully utilize rich gradient information of the image to realize image enhancement processing, so that the visual quality of the processed image is improved, and the texture information is rich.

The invention discloses a gradient domain adaptive gain underwater image enhancement method based on a YIQ space optical imaging model. The gradient domain adaptive gain underwater image enhancement method comprises the following steps: carrying out global background illumination vector estimation; obtaining the optimized transmissivity; The blocky effect of the transmissivity is eliminated; calculating a de-noising recovery image corresponding to the original observation image; converting the original color underwater image into a grayscale image; converting the grayscale image into a grayscale image capable of keeping image edge characteristics, and extracting a gradient image corresponding to the image; fusing the de-noised restored image and the gray level image which maintains the edge characteristics; converting the fused color image into a color space, and separating the brightness information from the color information; calculating the brightness information to obtain an enhanced image; converting the enhanced image into a color space. The four-direction Sobel edge detector used in the invention can fully utilize rich gradient information of the image to realize image enhancement processing, so that the visual quality of the processed image is improved, and the texture information is rich.

The invention discloses a near-infrared facula projection multi-focal-length micro-lens array remote sensing light field imaging system, which is characterized by comprising a near-infrared facula projection device (100) and a light field imaging assembly (200). The near-infrared light spot projection device (100) is used for scattering near-infrared light spots on an observation target and increasing texture information of a target image; and the light field imaging assembly (200) is used for imaging the target scene light with the texture information. The system can improve the novel remotesensing light field imaging of a target depth-of-field detection range, and especially can achieve the surface reconstruction of a weak texture target.

The invention discloses a method for measuring the terrain of a coastal shallow water area by using a low-altitude unmanned aerial vehicle. The method comprises the following steps: 1, obtaining tide level data of a measured area; using a low-altitude unmanned aerial vehicle for obtaining a measuring area image, and obtaining an underwater digital earth surface model and a route waypoint diagram before correction after processing; 2, generating a route tide level model; 3, obtaining a water depth data set before correction; 4, obtaining actual water depth data h water depth at a corresponding position during aerial survey of the unmanned aerial vehicle; 5, substituting multiple pairs of h water depth and h before correction into a simplified water depth refraction correction formula h water depth = k * h before correction, and obtaining a refraction correction proportion constant k; 6, multiplying the water depth data set before correction by a refraction correction proportion constant k to obtain a water depth data set after correction; and 7, performing one-to-one correspondence subtraction on the route tide level model and the corrected water depth data set to obtain shallow water area underwater topographic data. According to the invention, rapid and high-precision measurement of the coastal shallow water terrain can be realized.

The invention discloses a three-dimensional reconstruction system and method for an electronic microscopic scene. The method comprises the steps: collecting an image sequence of corresponding frames at a certain step through an electronic microscope, carrying out the segmentation processing of the images, carrying out the compiling of the images into a dictionary, solving a sparse coefficient, carrying out the fusion, and carrying out the recovery of the images to generate a clear image. In the specific embodiments, the focusing motion adjustment is completed through the combination of a stepping motor and a screw. The approximate object distance difference of all frames is calculated through a control signal and a positioning height, thereby achieving the three-dimensional reconstruction,and solving a problem of insufficient three-dimensional information caused by a small field depth in a microscopic scene.

The invention belongs to the technical field of image three-dimensional information reconfiguration and relates to a non-contact scanning method for three-dimensional colour point clouds, which uses a structured light projector and two CCD cameras to acquire images and comprises the following steps: acquiring a single-face image of a measured object; performing Gray space-time coding on the colour image of each width strip structured light, which is projected to the surface of the measured object, by using the corresponding information of the stripe light; calculating a three-dimensional coordinate of the measured object according to namesake phase positions of each pair of image pixels and opposite positions of two CCD sensors; calculating colour information of a three-dimensional coordinate point of the measured object by a three-dimensional space linear interpolation method; extracting textural features of the colour image; forming the single-face colour point clouds according to the calculated three-dimensional coordinate and the colour information thereof; forming the multi-face point clouds; and splicing different colour point clouds. The non-contact scanning method for the three-dimensional colour point clouds of the invention has the advantages of simple operation and high precision.

The invention discloses an attitude-guided virtual human body image generation method and system based on structural similarity and electronic equipment. The method comprises the following steps: step1, acquiring a source human body image and a target human body image; acquiring a target posture image according to the target human body image; step 2, inputting the source human body image and thetarget posture image in the step 1 into a pre-constructed convolutional neural network with an encoder-decoder structure to obtain a virtual target human body image; and step 3, constructing a loss function based on the virtual target human body image acquired in the step 2 and the target human body image acquired in the step 1, and performing iterative optimization on the pre-constructed convolutional neural network with an encoder-decoder structure. After a preset number of iterations is reached, the optimized convolutional neural network with an encoder-decoder structure is obtained, and the convolutional neural network is used for realizing virtual generation of a real scene human body image of a target posture. According to the invention, the real scene human body image of a more realtarget posture can be generated.

The invention discloses a method and a device for acquiring a calibration initial value of a vehicle-mounted LiDAR measuring system, which comprises the following steps: acquiring a two-dimensional image of a to-be-calibrated vehicle-mounted LiDAR measuring system by using a close-range photogrammetry mode; registering and reconstructing the two-dimensional image to generate a corresponding three-dimensional point cloud model; and fitting the three-dimensional point cloud model to obtain position data among instruments in the calibrated vehicle-mounted LiDAR measuring system. The method comprises the following steps of acquiring the two-dimensional image of the measuring system by using the close-range photogrammetry technology, fitting the two-dimensional image into the point cloud modelby using three-dimensional reconstruction, and then calculating to obtain the spatial attribute of the measuring system. The method for acquiring the calibration initial value of the vehicle-mounted LiDAR measuring system is simple and easy to implement, not only improves the data precision and enriches the texture information of the data, but also reduces the operation cost, and can provide goodtechnical support for the calibration of the vehicle-mounted LiDAR measuring system.

The invention discloses an improved SAR image and multi-spectral optical image fusion method. The method comprises the following steps: step 1, subjecting an original SAR image and a multi-spectral optical image to cubic re-sampling to obtain an SAR image and a multi-spectral optical image with the same image size; step 2, decomposing the re-sampled multi-spectral optical image into n wavebands, calculating the haze value of each wave band, and removing the haze value from the multi-spectral optical image after the re-sampling operation; step 3, carrying out preliminary fusion on the SAR imageobtained in the step 1 and the multispectral optical image obtained after the haze removal in the step 2, and then adding the haze value removed in the step 2 into the preliminarily fused image. Theinfluence of the haze on the image fusion quality is reduced to the maximum, so that the shape of a ground feature spectrum is kept to the largest extent in an obtained synthetic image. The intensityof the ground feature spectrum is replaced with the brightness of a corresponding SAR pixel, so that the abundant terrain and texture information can be included as much as possible.

The invention is suitable for the technical field of image processing, and provides a building extraction method. The method comprises the following steps: obtaining target image data (including first data and high-resolution image data); preprocessing the target image data; performing image processing on the preprocessed target image data, and obtaining an initial candidate region and an adaptive segmentation result; carrying out fusion processing on the initial candidate region and the self-adaptive segmentation result to obtain a building candidate region; finally, optimizing the building candidate region according to the initial candidate region to obtain a building extraction result. The invention is advantageous in that, by fusing the first data and the high-resolution image data to obtain a recognition result of a building, and in combination with the characteristics that the first data provides elevation information relative to the ground and is not easily influenced by environmental factors and the high-resolution image data provides rich spectral features and texture information, the robustness of the method, the precision of recognition results, and the stability of high-precision recognition results are improved.

The invention provides a camera position real-time repositioning method of a handheld three-dimensional scanning system. The camera position real-time repositioning method comprises the following execution processes: collecting m frames of RGB images I in advance before an algorithm is executed to obtain an image set A, A = {I1, I2,..., Im}, and point clouds P corresponding to the I one to one, and obtaining a point cloud set B, B = {P1, P2,..., Pm} and P = {p1, p2,..., pN}, wherein p = (x, y, z); before three-dimensional features are extracted, carrying out decentration and principal component analysis of point clouds, so that all the point clouds are located at the origin of world coordinates, and the principal directions are kept consistent, and differences among poses of all the pointclouds are eliminated as much as possible.