Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

32results about How to "Good image detail" patented technology

An ancient font classification method based on a convolutional neural network

The invention discloses an ancient font classification method based on a convolutional neural network. According to the method, firstly, an ancient font category image data set is crawled by using a crawler technology; through data expansion, training set samples tend to be balanced; graying processing is carried out on the balanced training set sample and setting an image size to a target image size; histogram equalization processing is performed on the sample set, isolated noise points are removed in the image through an N8 connected noise reduction algorithm, and finally binarization processing is performed on the image based on a fuzzy set theory and by using a Shannon entropy function, so that detail features of the image are well reserved; based on the objective function of the classification task. The center loss function and the traditional cross entropy loss function are matched for use. The inter-class distance is increased. The intra-class distance is reduced. The distinguishing capability of features is improved to a certain extent, preprocessed images are trained through a pre-defined network model, and the accuracy of a classification result is evaluated through a confusion matrix. According to the method. The preprocessing effect on the degraded ancient font image is remarkable, and a more accurate ancient font classification effect is achieved by optimizing parameter setting and utilizing appropriate training skills to train the convolutional neural network model.
Owner:HANGZHOU DIANZI UNIV

System and method for recognizing ships under visible light on basis of convolutional neural network

The invention relates to a system and method for recognizing ships under visible light on the basis of a convolutional neural network. The system comprises an image preprocessing module, an image augmentation module, a ship object detection model training module, a false alarm removal module, a ship classification model training module, a region zooming module and a ship classification model, wherein the image preprocessing module is used for removing interference information in images; the image augmentation module is used for augmenting a picture training set through carrying out cropping, segmentation and variable illumination on existing images; the ship object detection model training module is used for training a model for ship detection by using an object detection model framework;the false alarm removal module is used for removing recognized parts; the ship classification model training module is used for training an object detection model framework by using a data set; the region zooming module is used for randomly selecting a region for the images so as to enlarge the data set; and the ship classification model is used for classifying detected ships. At present, the system for recognizing ships under visible light on the basis of the convolutional neural network has precision higher than that of traditional ship recognition method in the field of ship recognition.
Owner:UNIV OF ELECTRONIC SCI & TECH OF CHINA

Underwater image enhancement method based on foreground model

The invention discloses an underwater image enhancement method based on a foreground model. The method comprises the following steps that: improving a background light estimation method so as to effectively avoid the influences of underwater image overexposure, artificial light sources and the like; combining with the cognition of people for the underwater image, and utilizing a dark channel priori algorithm to remove background scattering and extract the foreground model; and combining with a white balance algorithm to put forward a color correction method suitable for the underwater image, and utilizing the attenuation characteristic of light in water to correct channel gains according to a relationship between channel attenuation coefficients so as to compensate color distortion caused by attenuation; and utilizing the channel gains to regulate a fogless image, and finally, obtaining the enhanced underwater image. By use of the method, the enhancement effect of an object part is clear and distinct, and a visual effect is better; image blurring is effectively removed, so that the definition of the enhanced underwater image is greatly improved, image details are better, the image enhancement of the background part is not affected on the basis of the color correction of the foreground model, the enhanced underwater image has more natural integral colors, and image brightness is within an acceptable range.
Owner:TIANJIN UNIV

X-ray image multi-scale detail enhancement method in integrated circuit packaging

The invention discloses an X-ray image multi-scale detail enhancement method in integrated circuit packaging, which comprises the steps of 1) acquiring X-ray images oriented to an integrated circuit packaging process; 2) performing Laplacian pyramid decomposition to the X-ray images to obtain X-ray pyramid detail images at all scales; 3) adjusting the brightness and the contrast of the images at the bottom layer of a pyramid by adopting a logarithm enhancement method; 4) adjusting the brightness and the contrast of the detail images at the top layer of the pyramid by adopting a histogram equalization enhancement method; 5) reconstructing the detail images at all scales to obtain detail enhanced images. The method disclosed by the invention has the advantages that the method is simple and fast, the dynamic range compression of images can be realized aiming at the X-ray images with low signal-to-noise ratio and poor defect contrast in the integrated circuit packaging, the useful information of the images is expanded, the image details are enabled to be more outstanding, the contrast of the images is improved and the accuracy is higher.
Owner:SOUTH CHINA UNIV OF TECH

An end-to-end image defogging processing method based on deep learning

The invention relates to an end-to-end image defogging processing method based on deep learning. According to the method, a foggy image is converted into a non-foggy image through a trained deep convolutional neural network, and the deep convolutional neural network comprises a feature extraction module which comprises a plurality of convolution sub-modules and is used for carrying out convolutioncalculation on an input image and extracting a multi-dimensional feature map; the feature pooling module comprises a plurality of pooling layers, and after each pooling layer is correspondingly connected to one convolution sub-module, redundancy removal processing is carried out on the multi-dimensional feature map; the recovery module comprises a plurality of deconvolution sub-modules, is connected to the feature pooling module and then outputs an output image with the same resolution as the input image; and a plurality of inter-layer jump connection layers which are used for realizing inter-layer jump connection between the output of the pooling layer and the input of the corresponding deconvolution sub-module and fusing the multi-scale feature map. Compared with the prior art, the method has the advantages of good defogging effect, simple process and the like.
Owner:聚时科技(上海)有限公司

MSRCR traffic image haze removing method combining HE and guide filter

The invention discloses an MSRCR traffic image haze removing method combining HE and a guide filter. The method comprises the steps of firstly conducting HE and MSRCR enhancement on images separately, wherein MSRCR enhancement replaces a gaussian function estimated illumination component with the guide filter with a smoothing and edge protection function, multi-scale processing is conducted on the guide filter singularly, and then conducting weight fusion on the two enhanced images according to a certain image fusion rule. An experimental result shows that the MSRCR traffic image haze removing method combining HE and the guide filter can obtain an excellent fog removing effect for the haze traffic images, and has the advantages of being high in efficiency, good in color fidelity, obvious in image detail after fog removing and the like.
Owner:CHONGQING NORMAL UNIVERSITY

Retinex-based color image enhancement method

The invention discloses a Retinex-based color image enhancement method, which comprises the following steps: calculating a source image gray value to obtain a first grayscale image; carrying out multi-scale Retinex processing and pixel correction on the first grayscale image to obtain a second grayscale image; calculating a dyeing factor according to the first grayscale image and the second grayscale image; carrying out dyeing processing on a source R channel image, a source G channel image and a source B channel image according to the dyeing factor to obtain a second R channel image, a secondG channel image and a second B channel image; carrying out dyeing processing on the source R channel image to obtain a third R channel image; and carrying out synthesis on the third R channel image,the second G channel image and the second B channel image to obtain an enhanced color image. The technical scheme can improve image brightness and contrast, prevent image distortion, eliminate the problem of halo easily generated in contrast-obvious areas, and can enhance image edge information and highlight image details.
Owner:CHENGDU MEDICAL COLLEGE

Image defogging method based on convolutional neural network

The invention discloses an image defogging method based on a convolutional neural network. The method comprises the following steps: transforming an atmospheric scattering model; building an Encoder-decoder network, and estimating an intermediate transmission graph; processing an image restoration problem; and establishing a Dehazer network to realize a Dehazer function, and outputting a defoggedimage. The Encoder-decoder network does not need to change the network structure and related parameters, the influence of noise and jitter can be reduced, important features related to a target imageare obtained, and an accurate intermediate transmission graph is output. The Dehazer network is simple in structure, convenient to train, capable of sharing multiple parameters, appropriate in calculation overhead and stable in network performance, gradient disappearance and explosion can be effectively prevented, and defogged images can be conveniently and rapidly output. According to the methodprovided by the invention, the defogged image can be efficiently and quickly output, the performance of the established network is relatively stable, the influence of fog or haze can be well eliminated, the defogging quality of the image is effectively improved, and the defogging effect is relatively ideal.
Owner:SHANDONG INST OF BUSINESS & TECH

Image processing method and device

The invention provides an image processing method and device. The method comprises the steps that an original brightness image of an original image is decomposed into N orders by wavelet transform; for each decomposed image, the corresponding frequency of each pixel in the decomposed image is adjusted according to a frequency adjustment threshold corresponding to the decomposed image, wherein the frequency higher than the frequency adjustment threshold is adjusted lower, and the frequency lower than the frequency adjustment threshold is adjusted higher; according to each decomposed image obtained after frequency adjustment, an adjusted brightness image corresponding to the original image is reconstructed; according to the original image, the original brightness image and the adjusted brightness image, a processed image corresponding to the original image is determined. The image processing method and device can reproduce image details and enhance details in areas with less prominent details, and can enhance the details of the dark region image of the original image and reproduce the details of the highlighted region image.
Owner:ZHEJIANG DAHUA TECH

Defective pixel detection and correction device and method based on texture recognition

The invention provides a defective pixel detection and correction device based on texture recognition. The device comprises an image sensor data input unit for inputting image data of an image sensor;a chromatic aberration compensation unit connected to the image sensor data input unit, and used for counting the chromatic aberration of each pixel of the image data in an m*n data window, selectinga median value of the chromatic aberration to obtain a chromatic aberration value of the current pixel, subtracting the chromatic aberration value, and performing chromatic aberration compensation; adefective pixel detection unit for calculating a total variation of N pixels around the current pixel, including pixels of the same channel or different channels, determining the texture intensity ofthe current pixel according to the calculation result, and detecting defective pixels; and a defective pixel correction unit for collecting the defective pixels according to the detection result of the defective pixel detection unit. The invention further provides a defective pixel detection and correction method based on texture recognition.
Owner:思特威(上海)电子科技股份有限公司

Residual instance regression super-resolution reconstruction method based on multistage dictionary learning

The invention discloses a residual instance regression super-resolution reconstruction method based on multistage dictionary learning, and the method comprises the following steps: generating a training set through high-resolution images, and establishing block pairs of low-resolution and high-resolution image blocks; extracting feature vectors of low-resolution image blocks, and learning a dictionary with strong representation ability by using K-SVD as an anchor point; performing the least square regression of low-resolution and high-resolution blocks in the block pairs through the dictionaryobtained via learning, and obtaining a linear mapping relation; estimating the high-resolution features, calculating a reconstruction error, and carrying out the mapping of the estimated high-resolution features and the reconstruction error while the further dictionary learning of the estimated high-resolution features; obtaining a group of residual regression devices after the L layer; carryingout the reconstruction through an inputted image and the obtained residual regression devices, and enabling the obtained high resolution features to be used for the reconstruction of a next layer; adding all estimated high-resolution image blocks and forming a high-resolution image through synthesis. The method is stronger in super-resolution capability, and can be used for the amplification of alow-resolution natural image.
Owner:XI'AN POLYTECHNIC UNIVERSITY

Water-based pigment ink used for a high-speed ink-jet printer and a preparing method thereof

InactiveCN108948864AExcellent ejection stabilityPrinting is smooth and fastInksWater basedSolvent
The invention relates to water-based pigment ink used for a high-speed ink-jet printer and a preparing method thereof. The water-based pigment ink includes 5-15 parts of pigment, 0.3-1 part of a defoamer, 1-3 parts of a dispersant, 39.6-67.6 parts of purified water and 26.1-38.4 parts of a waterborne solvent. The method includes (1) preparing the purified water; (2) adding the purified water, thepigment, the dispersant and the waterborne solvent according to a mass ratio into a dispersing machine, and dispersing the materials to obtain a mixture; (3) grinding the mixture until the particle size is less than 500 nm; and (4) adding the mixture into a stirrer, adding a pH adjusting agent, the defoamer, a humectant, an aseptic and a surfactant according to a ratio, and fully stirring the mixture to obtain the ink.
Owner:ZHONGJI OIL COLTHING BEIJING TECH CO LTD

Image processing method

The invention provides an image processing method. According to the image processing method provided by the invention, brightness clustering partitioning is carried out on an original image; generating a plurality of virtual exposure images with different exposure degrees by utilizing the original image; each virtual exposure image is divided into a plurality of virtual blocks according to the brightness clustering partition of the original image; According to the embodiment of the invention, the virtual blocks to be fused are selected from the virtual blocks of each virtual exposure image andthen are spliced to obtain the processed image, so that the image has better image details, the quality of the image is improved, and the image processing speed can be increased.
Owner:SHENZHEN CHINA STAR OPTOELECTRONICS SEMICON DISPLAY TECH CO LTD

Garment style migration method and system based on deep learning

The invention discloses a garment style migration method and system based on deep learning, and relates to the technical field of garment design, and the system comprises a data acquisition and processing module, a feature analysis and extraction module, a style model generation and application module, an image migration and generation module, and an image output module. The data acquisition and processing module is used for selecting style images and preprocessing the images; the feature extraction module is mainly used for extracting color features, texture features and contour features; the style model generation and application module is used for generating a style model; the image migration and generation module comprises a personalized style migration function and a rapid style migration function, and is used for carrying out semantic segmentation on a source clothing image, and style migration can be directly carried out through the above modules, or rapid style migration is realized by calling a style model in a style model library; and the image output module is used for improving and displaying the resolution of the generated image. The style migration time is shortened, and the overall effect of the generated garment image is improved.
Owner:DALIAN POLYTECHNIC UNIVERSITY

Under-sampling lung gas MRI reconstruction method based on multi-task complex value deep learning

The invention discloses an under-sampling lung gas MRI reconstruction method based on multi-task complex value deep learning, and the method comprises the steps of predicting complete k-space data through employing a k-space reconstruction network, obtaining a preliminary reconstruction image through employing an image domain reconstruction network, finally further enhancing the details of the image through employing a multi-task detail enhancement network combining segmentation and reconstruction, and obtaining a finally reconstructed lung hyperpolarized gas MRI image. According to the method, the plurality of convolutional layers are adopted, and phase information in the k space is effectively utilized. Compared with a traditional reconstruction method, the imaging speed is greatly increased while the reconstruction quality is improved. Compared with a network with a single training reconstruction task, the method has the advantages that the two tasks of reconstruction and segmentation are trained at the same time, the two tasks share a feature extraction layer, the segmentation task pays more attention to details and edge parts of the image, more high-frequency features can be extracted, better image details can be reconstructed, and the reconstruction quality is improved.
Owner:INNOVATION ACAD FOR PRECISION MEASUREMENT SCI & TECH CAS

Color specified enhancement plug-in for Maya and color enhancement method thereof

The invention discloses a color specified enhancement plug-in for Maya and a color enhancement method thereof, and particularly relates to the technical field of color enhancement. The plug-in is connected with a color enhancement controller; and the plug-in comprises a color enhancement database, a processor and a determiner. A processor is adopted to process color enhancement operation object data, an original image is converted into a CIELab color space, the CIELab color space is a uniform color space consistent with human visual perception, sub-block processing is carried out, then image decomposition is carried out on sub-blocks, sub-block images are processed in a specified mode, the variation amplitude of the numerical value of the L* component in the histogram specification process is limited by segmenting the L* channel histogram and limiting the mapping range, so that the color difference of the image after color enhancement processing is limited to be small, the color information is kept as far as possible, the details of the image are highlighted, the useful information is enhanced, and the image is easier to identify and process.
Owner:武汉云漫文化传媒有限公司

Coal rock boundary image enhancement method based on chaos sparrow search algorithm

A coal rock boundary image enhancement method based on a chaos sparrow search algorithm introduces a chaos mapping mechanism to perform chaos disturbance on sparrow populations in the sparrow search algorithm, so the initialized populations are uniformly distributed in a random mode, and the convergence speed of the algorithm is accelerated; chaos disturbance is carried out on individuals with low adaptive values in the search process, so that the algorithm jumps out of local optimum more easily, and the stability and precision of the algorithm are enhanced; finally, a chaotic sparrow search algorithm is combined with a normalized incomplete Beta function to adaptively select parameters to obtain an optimal image enhancement parameter, so that an optimal gray curve is found, and adaptive enhancement of the contrast of the coal rock boundary gray image is realized. The method has the advantages that low-illumination and low-contrast image features of an underground coal mine are improved, the problems of excessive enhancement of a local bright area of the image generated after enhancement, poor enhancement effect of details of a dark part and the like can be avoided, and the visual effect and the quality of the enhanced image are remarkably improved.
Owner:TAIYUAN UNIV OF TECH

image processing method

The invention provides an image processing method. In the image processing method of the present invention, the original image is subjected to brightness clustering partitioning, and the original image is used to generate multiple virtual exposure images with different exposures. Each virtual exposure image is divided into multiple virtual blocks according to the brightness clustering partitioning of the original image. , select the virtual blocks to be fused from multiple virtual blocks in each virtual exposure image, and then splice the multiple virtual blocks to be fused to obtain a processed image, which can make the image have better image details and improve the quality of the image. quality and can increase the speed of image processing.
Owner:SHENZHEN CHINA STAR OPTOELECTRONICS SEMICON DISPLAY TECH CO LTD

Image rain removal model training method, image rain removal method and device

The embodiment of the invention provides an image rain removal model training method, an image rain removal method and equipment, in order to eliminate long rain stripes, a recursive convolution structure is used, feature vectors of an input image with rain stripes are extracted through a plurality of convolution kernels, and vector fusion is carried out, so that the long rain stripes are eliminated. And the fusion feature vector and the feature vector of the input image are added and are output through the LSTM network module, so that long-distance context information can be better linked, and the receptive field can be expanded. And comparing the output feature vector with the feature vector of the corresponding rain-free image, if the comparison result is not a preset result, adjusting the parameter of the recursive convolution module according to the comparison result, and re-executing the operation, that is, through cyclic convolution, enabling the model to learn the internal relation between the cross-stage features, better recovering the details of the image, and improving the image quality. Moreover, the size of the model is effectively reduced, and a good rain removal effect can be achieved on portable embedded equipment and mobile equipment with insufficient memory and computing power.
Owner:CHENGDU TD TECH LTD +1

An end-to-end image defogging method based on deep learning

The present invention relates to an end-to-end image defogging processing method based on deep learning, which converts a foggy image into a fog-free image through a trained deep convolutional neural network, wherein the deep convolutional neural network includes : Feature extraction module, including multiple convolution sub-modules, performs convolution calculation on input image, extracts multi-dimensional feature map; feature pooling module, includes multiple pooling layers, and each pooling layer is correspondingly connected to one of the volumes After the product sub-module, the multi-dimensional feature map is de-redundantly processed; the recovery module includes a plurality of deconvolution sub-modules, connected after the feature pooling module, and outputs an output image with the same resolution as the input image ; There are multiple inter-layer skip connection layers, which realize the inter-layer skip connection between the output of the pooling layer and the input of the corresponding deconvolution sub-module, and fuse the multi-scale feature map. Compared with the prior art, the invention has the advantages of good defogging effect, simple process and the like.
Owner:聚时科技(上海)有限公司

A Method of Underwater Image Enhancement Based on Foreground Model

The invention discloses an underwater image enhancement method based on a foreground model. The method comprises the following steps that: improving a background light estimation method so as to effectively avoid the influences of underwater image overexposure, artificial light sources and the like; combining with the cognition of people for the underwater image, and utilizing a dark channel priori algorithm to remove background scattering and extract the foreground model; and combining with a white balance algorithm to put forward a color correction method suitable for the underwater image, and utilizing the attenuation characteristic of light in water to correct channel gains according to a relationship between channel attenuation coefficients so as to compensate color distortion caused by attenuation; and utilizing the channel gains to regulate a fogless image, and finally, obtaining the enhanced underwater image. By use of the method, the enhancement effect of an object part is clear and distinct, and a visual effect is better; image blurring is effectively removed, so that the definition of the enhanced underwater image is greatly improved, image details are better, the image enhancement of the background part is not affected on the basis of the color correction of the foreground model, the enhanced underwater image has more natural integral colors, and image brightness is within an acceptable range.
Owner:TIANJIN UNIV

A Convolutional Neural Network Based Ancient Font Classification Method

The invention discloses an ancient font classification method based on a convolutional neural network. According to the method, firstly, an ancient font category image data set is crawled by using a crawler technology; through data expansion, training set samples tend to be balanced; graying processing is carried out on the balanced training set sample and setting an image size to a target image size; histogram equalization processing is performed on the sample set, isolated noise points are removed in the image through an N8 connected noise reduction algorithm, and finally binarization processing is performed on the image based on a fuzzy set theory and by using a Shannon entropy function, so that detail features of the image are well reserved; based on the objective function of the classification task. The center loss function and the traditional cross entropy loss function are matched for use. The inter-class distance is increased. The intra-class distance is reduced. The distinguishing capability of features is improved to a certain extent, preprocessed images are trained through a pre-defined network model, and the accuracy of a classification result is evaluated through a confusion matrix. According to the method. The preprocessing effect on the degraded ancient font image is remarkable, and a more accurate ancient font classification effect is achieved by optimizing parameter setting and utilizing appropriate training skills to train the convolutional neural network model.
Owner:HANGZHOU DIANZI UNIV

Real-time enhanced processing system for foggy continuous video image

The invention discloses a block processing and fuzzy algorithm-based real-time enhanced processing system for a foggy continuous video image. The image is divided into a base frame and a follow-up frame; block processing is performed on the base frame; a transit point and a membership function are matched for each subblock; then, the image is transformed into a fuzzy domain from a spatial domain; nonlinear transformation enhancement is performed on the fuzzy domain to obtain a color channel mapping table for a red (R) channel, a green (G) channel and a blue (B) channel of each subblock before and after the image is enhanced; finally, the image is inversely transformed back to the spatial domain; and the follow-up frame is processed according to the base frame to directly enhance the image or update the base frame. The block processing of the processing system disclosed by the invention is more real-time than global processing. The distortion of the image is reduced while the detail of the image is protruded; a visual effect of the continuous video image is improved under the foggy condition; the picture contrast of the image is enhanced, and the quality of the image can be ensured; and the requirements of required real-time clear image information and a monitoring effect can be completely met.
Owner:SICHUAN UNIV

A color image enhancement method based on retinex

The invention discloses a color image enhancement method based on Retinex, which includes: calculating the grayscale value of the source image to obtain a first grayscale image; performing multi-scale Retinex processing and pixel correction on the first grayscale image to obtain a second grayscale image. ; Calculate the dyeing factor based on the first grayscale image and the second grayscale image; dye the images of the source R channel, source G channel and source B channel respectively according to the dyeing factors to obtain the second R channel, the second G channel and the second grayscale image. two B channel images; dye the source R channel image to obtain a third R channel image; synthesize the images of the third R channel, the second G channel and the second B channel to obtain an enhanced color image. The technical solution provided by the present invention can improve image brightness and contrast, avoid image distortion, eliminate the problem of halo easily occurring in areas with obvious contrast, and can enhance image edge information and highlight image details.
Owner:CHENGDU MEDICAL COLLEGE

Defect pixel detection and correction device and method based on texture recognition

The present invention provides a defect pixel detection and correction device based on texture recognition, the device includes an image sensor data input unit for inputting image data of the image sensor; a color difference compensation unit connected to the image sensor data input unit, for all Each pixel of the image data counts the color difference in the m×n data window, takes the median value of the color difference, obtains the color difference value of the current pixel point, subtracts the color difference value, and performs color difference compensation; the defective pixel detection unit , calculating the total variation of N pixels around the current pixel, including pixels of the same channel or not of the same channel, determining the texture intensity of the current pixel according to the calculation results, and detecting defective pixels; the defective pixel correction unit uses The defective pixel is corrected according to the detection result of the defective pixel detection unit. The invention also provides a method for detecting and correcting defective pixels based on texture recognition.
Owner:思特威(上海)电子科技股份有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products