100 results about "Local histogram" patented technology

Methods for contrast enhancement of digital images are provided. A method of adaptive histogram equalization is provided that determines weighting factors for discriminating between sub-regions of a digital image to be more enhanced or less enhanced. Another method for content adaptive local histogram equalization is provided that uses a mapping function in which the dynamic range is not changed by the transformation. A third method for contrast enhancement is provided that includes dividing a digital image into a plurality of regions of pixels, and for each region in the plurality of regions, determining a threshold gray level for the region, generating a mapping curve for the region based on the threshold gray level, and applying the generated mapping curve to each pixel in the region to enhance contrast.

Techniques for accelerated elastic registration include receiving reference scan data and floating scan data, and a first transformation for mapping coordinates of scan elements from the first scan to coordinates of scan elements in the second scan. A subset of contiguous scan elements is determined. At least one of several enhancements is implemented. In one enhancement cubic spline interpolation is nested by dimensions within a subset. In another enhancement, a local joint histogram of mutual information based on the reference scan data and the floating scan data for the subset is determined and subtracted from an overall joint histogram to determine a remainder joint histogram. Each subset is then transformed, used to compute an updated local histogram, and added to the remainder joint histogram to produce an updated joint histogram. In another enhancement, a measure of similarity other than non-normalized mutual information is derived from the updated joint histogram.

Methods and apparatus are configured to provide data to render (medical) images using direct volume rendering by electronically analyzing a medical volume data set associated with a patient that is automatically electronically divided into a plurality of local histograms having intensity value ranges associated therewith and programmatically generating data used for at least one of tissue detection or tissue classification of tissue having overlapping intensity values.

The invention relates to a face recognition method based on Gabor wavelet transform and local binary pattern (LBP) optimization. Two-dimensional Gabor wavelet transform can associate pixels of adjacent areas so as to reflect the change conditions of image pixel gray values in a local range from different frequency scales and directions. The feature extraction and the classification recognition are carried out on the basis of a face image two-dimensional Gabor wavelet transform coefficient. For a high-dimensional Gabor wavelet transform coefficient, overall histogram features are extracted by adopting the LBP, and then the image is blocked by utilizing priori knowledge to extract the features of each piece of LBP local histogram. The method has better recognition rate, better robustness to illumination and wide using prospect in the fields of biometric recognition and public security monitoring.

The invention provides a gray scale image fitting enhancement method based on local histogram equalization. The gray scale image fitting enhancement method has advantages of improving gray scale image contrast and detail information and eliminating block effect and over-enhancement. The gray scale image fitting enhancement method comprises the steps of performing segmental linear transformation on a gray scale image in an overwide dynamic range, obtaining the gray scale image in an appropriate dynamic range, dividing an image gray scale distribution interval to two segments to multiple segments, adjusting the gradient of a segmenting point and a transformation line of each image gray scale distribution interval, performing expansion or compression on a random gray scale interval; performing subblock part overlapping histogram equalization on a transformation result, obtaining the transformation function of the current subblock through performing weighted summation on a subblock transform function in the neighborhood, performing histogram equalization processing on the current subblock by means of the transformation function; and performing nonlinear fitting on the gray scale map after histogram equalization, and performing histogram distribution correction on the gray scale image after subblock part overlapping histogram equalization.

Methods and apparatus are configured to provide data to generate and / or render (medical) images using medical volumetric data sets by electronically analyzing a medical volume data set associated with a patient that is automatically electronically divided into a plurality of local histograms having intensity value ranges associated therewith and programmatically generating at least one α-histogram of data from the local histograms used for at least one of peak detection, transfer function design or adaptation, tissue detection or tissue classification.

A dynamic thresholding level set method combines two optimization processes, i.e., a level set segmentation and an optimal threshold calculation in a local histogram, into one process that involves a structure called a “propagating shell.” The propagating shell is a mobile 3-dimensional shell structure with a thickness that encompasses the boundary of an object, the boundary between two objects or the boundary between an object and a background. Because the local optimal threshold tends to shift to a value of a small region in a histogram, the shift can drive the propagating shell to an object boundary by pushing or pulling the propagating shell. The segmentation process is an optimizing process to find a balanced histogram with minimal threshold shift. When the histogram in the propagating shell is balanced, the optimal threshold becomes stable, and the propagating shell reaches a convergence location, i.e., an object boundary. This method can be applied to computer-aided organ and tumor volumetrics.

The invention discloses a video image enhancement method and device. The method comprises the following steps: dividing an image to be enhanced into a plurality of image blocks with the same size, acquiring the image information of the image blocks, acquiring a local histogram height threshold according to the image information, adaptively adjusting the contrast of the image blocks according to the local histogram height threshold, acquiring grey values of central pixel points of all image blocks according to a contrast adjustment result, acquiring the grey values of the central pixel points of non-image blocks in the enhanced image by interpolation, and outputting the enhanced image. The device comprises a division module, a computing module, a contrast adjusting module and an interpolation module. The method and the device disclosed by the invention are used for adopting corresponding local histogram height thresholds to enhance the contrast of different image blocks, uniform and sufficient tiling on the gray level can be realized, and the adaptability to different monitoring scenes is strong.

The invention discloses a human face recognition method based on local feature learning, which comprises the following steps of: (a) partitioning a known classified human face sample into blocks, and calculating data of the human face sample of each block through an LBP (Length Between Perpendiculars) operator and an LTP (Long-Term Potentiation) operator to obtain a local histogram vector a of data of the human face sample of each block; (b) performing chi-square histogram distance calculation on local histogram vectors of the same positions on any two different human face samples of the same person in all human face samples to obtain a positive sample feature library; and (c) performing chi-square histogram distance calculation on local histogram vectors of the same positions on any two human face samples of different persons in all human face samples to obtain a negative sample feature library. The human face recognition method based on the local feature learning, provided by the invention, has the advantages of quick response, high accuracy and good recognition effect.

Methods and apparatus are configured to provide data to render (medical) images using direct volume rendering by electronically analyzing a medical volume data set associated with a patient that is automatically electronically divided into a plurality of local histograms having intensity value ranges associated therewith and programmatically generating data used for at least one of tissue detection or tissue classification of tissue having overlapping intensity values.

The invention relates to the field of image processing technology, in particular to an image enhancement processing method for an industrial or medical X-ray image inspection system. The method comprises the following steps of : S1. inputting a collected X-ray image; S2. enhancing image edges by adopting Laplacian edge enhancement algorithm which can adjust intensity; S3. processing each detail in the enhanced image by adopting quick local histogram equalization (LHE) ; S4. stretching the gray range of the image data which has been processed through the step S3; S5. outputting the enhanced image. Compared with the perspective image enhancement algorithm on prior inspection equipment, the invention has the advantages of better enhancement result, adaptive ability, high automation degree, simple operation and quick execution.

The invention relates to an X-ray imaging stretching fault automatic detection method for a steel cable core conveying belt joint, which belongs to the field of X-ray nondestructive detection. Defects of lower reliability, poorer accuracy, non-visual display, dependence mainly on manual detection and the like exist in the current stretching fault detection of a current steel cable core conveying belt joint. The invention consists of the following three aspects: 1. on the basis of the gray level change characteristic of X-ray images of a steel cable core conveying belt, screening out images containing partial or all joint information from the X-ray images of the whole steel cable core conveying belt by adopting local histogram information; 2. splicing the screened images into the corresponding image containing the complete joint information; and 3. realizing the judgment for the stretching of each joint by adopting an improved condition Y-difference algorithm and a clustering method. The invention can automatically, rapidly and efficiently realize the stretching fault detection of the steel cable core conveying belt joint.

The current invention describes a method for filtering an input image with a bilateral filter to produce an output image. The bilateral filter includes a spatial filter and a range filter. The method comprising the steps of: constructing an integral histogram from an input image including pixels, and wherein each pixel has an intensity; applying, for each pixel, the spatial filter to the integral histogram to produce a local histogram, each local histogram having a bin for a specified range of intensities of the pixels, each bin associated with a coefficient indicating a number of pixels in the specified range and an index to the coefficient; subtracting, for each bin in each local histogram, an intensity of the pixel from each index of the bin to produce a difference value; applying, for each bin, the range filter to each difference value to produce a response; scaling, each response by the corresponding coefficient to produce a scaled response; summing, for each local histogram, the scaled responses to produce a local response for the local histogram; summing, for each local histogram, the coefficients to produce a sum of the coefficient; and dividing, for each pixel, the local response by the sum of the coefficients to produce a response for the bilateral filter, which forms an output image.

The invention provides a laser underwater imaged image denoising and enhancing method for ocean exploration, and belongs to the field of image processing and pattern recognition. The core of the method is that at first, noise processing is performed on a laser underwater imaged image by applying SVD (singular value decomposition) and reconstruction in an airspace; then, aiming at the problem that the image has a little noise after SVD processing, wavelet decomposition and reconstruction are performed on the image in a wavelet domain to realize the purpose of wavelet denoising and enhancing; at last, FFT enhancing is performed on the image in a frequency domain to improve the image contrast and effectively keep the image resolution. Experiments show that the method can overcome a blocking effect due to equalization of local histogram equalization, and the image noise is effectively inhibited and the image resolution is kept under the premise that the image enhancement is realized.

In an image processing technology, an original image is inputted by accepting an input of an original image to be image-processed; a local area is set, with a pixel of which brightness is to be converted being a center of the local area, from the original image; a local histogram relating to brightness of the local area is calculated; a local cumulative histogram is calculated by accumulating the local histogram; first and second monotone increasing functions are determined, respectively, corresponding to cumulative frequency values in first and second classes of the local cumulative histogram; first and second weighting functions are determined, respectively, corresponding to the first and second monotone increasing functions; a conversion function relating to brightness of the pixel is produced from the first monotone increasing function weighted by the first weighting function and the second monotone increasing function weighted by the second weighting function; and the brightness of the pixel is then converted by the conversion function.

The invention provides an image defogging method based on a dark channel prior and a Markov random field. Aiming at problems that image quality is greatly reduced under a haze environment and a lot of defogging image structure detail information is lost by using an existing algorithm, a single image defogging method combining the dark channel prior (DCP) and the Markov random field (MRF) is provided. The method is characterized by firstly using sub-block portion overlapping local histogram equalization (POSHE) to enhance an original fog image so as to increase a contrast ratio, and through a DCP algorithm, acquiring an optimized transmissivity; using a constraint characteristic of a MRF model to image structure detail information to carry out modeling on the transmissivity so as to further refine the transmissivity; and according to a substantial characteristic of a sky domain, through a partitioning search method, calculating an atmospheric optical value. Compared to a traditional defogging method, by using the method of the invention, an accurate transmissivity image can be acquired, image structure information is effectively maintained, and the defogged image presents abundant details and a real color visual sense effect.

The invention discloses a vehicle target detection method and system based on YOLOv2, which comprises the steps of obtaining sample traffic video data; dividing the sample traffic video data into frame images to serve as sample images; performing noise reduction, shadow elimination, local histogram equalization and local invariant analysis on the sample images to obtain training images; inputtingthe training images into a preset YOLOv2 neural network model for training to obtain a vehicle detection model; obtaining real-time vehicle video data; performing noise reduction, shadow elimination,ghost elimination, local smoothing and local invariant analysis on the real-time vehicle video data to obtain secondarily processed real-time vehicle video data; segmenting the secondarily processed real-time vehicle video data into frame images, and inputting the frame images into the vehicle detection model to obtain a result map. The vehicle detection accuracy and the detection speed can be improved according to the vehicle target detection method and system based on YOLOv2 provided by the invention.

Local histogram and local histogram-based functions can be determined by generating offset-kernel images based on domain-shifted tonal filter kernels. The offset-kernel images can be reused for multiple image locations and / or local neighborhood sizes, shapes, and weights. A neighborhood filter representing the desired local neighborhood size, shape, and frequency domain characteristics is applied to each of the offset-kernel images. Neighborhood filters may include a temporal dimension for evaluating neighborhoods in space and time. Neighborhood filtered offset-kernel images' values represent samples of local histogram or local histogram-based function corresponding with the domains of their associated domain-shifted tonal filter kernels. Arbitrary functions may be used as tonal filter kernels. A histogram kernel may be used to sample values of local histogram functions. A tonal filter kernel that is a derivative or integral of another tonal filter kernel may be used to sample a derivative or integral, respectively, of a function.

The invention discloses a high dynamic range image tone mapping method and system based on deep learning. The method comprises the following steps of constructing a tone mapping network framework; preprocessing after a high-dynamic-range image is inputted, and calculating a total loss function through a global sensory characteristic loss function and a local histogram characteristic loss function;training the network for the tone mapping network framework according to the total loss function; and when the training result converges, stopping training and obtaining the output of the tone mapping network. Due to the neural network framework in the invention, two cost functions based on a histogram can be optimized, also the tone mapping can be realized end to end, the boundary problem between small areas is solved, and a high-quality low-dynamic-range image is directly obtained.

The invention discloses a pedestrian and vehicle accessory identification and retrieval method based on deep learning. The method comprises the following steps that: firstly, sampling eight directionsof a pixel in an image, carrying out quantitative sampling to obtain texture information, and coding two groups of double-cross subsets on each pixel by a double-cross coder to form a total descriptor; according to the descriptor and local gray level simultaneous distribution density , extracting a local histogram vector, and forming texture features; according to the extracted texture features,training to obtain an initial classifier, and setting learning frequencies and accuracy requirements; adopting an active learning algorithm to optimize the classifier, and stopping until a preset accuracy requirement is achieved; and finally, using a multi-instance multi-tag classifier which finishes being trained for identification to obtain a high-accuracy identification result. The system whichis put forward by the invention has the advantages of being high in adaptivity, high in confidence level and steady in integral performance. When image features are extracted, a double-cross mode coding method is adopted, maximum combination entropy can be realized, an image signal-to-noise ratio is maximized, and image robustness is enhanced.

The invention discloses a dynamic contrast ratio enhancement method based on function curve transformation, and relates to the technical field of image processing. The method comprises the following steps that a frame image in video is obtained; a gray-scale value of each pixel point in the frame image is obtained; global histogram statistics of the current frame image is established, a gray-scale value a1 is selected, a ratio n is set, the relation between the a1 and the n meets the condition that in the global histogram statistics, the ratio of the sum of the pixel points of which the gray-scale values range from 0 to the a1 to the total pixel points is n, or local histogram statistics of the current frame image is established, and a gray scale average a2 is calculated; the obtained gray-scale value of each pixel point is substituted into a formula, and a corrected gray-scale value of the pixel point is obtained; the gray-scale values of all the pixel points are modified to be I, and an enhanced frame image is obtained; a normalized ratio K is set, a ratio k of a gradient difference before and after adjustment of the current frame image is calculated, if abs(k-K) is smaller than or equal to m, processing of the frame image is completed, otherwise, coeff is adjusted, and the steps from third to fifth are repeated or the coeff is used in a formula of a next frame image.

For the specific problem of image defogging, the invention provides a fast defogging algorithm based on local histogram enhancement to improve the accuracy of the defogging algorithm and the efficiency of algorithm operation, so that the defogging algorithm meets a real-time requirement. The defogging algorithm comprises the steps that black level processing is carried out on an image to be processed, and fixed black level is subtracted; the image to be processed is corrected based on gamma correction and sub block partition is carried out on the image to be processed to carry out gray scale statistics; limited gray-scale stretch is carried out; and finally, pixel by pixel mapping is carried out to acquire a stretched clear fog-free image of high comparison. According to the algorithm provided by the invention, a limited local histogram equalization enhancement method is used; pixel by pixel gray-scale mapping is carried out on the image; four adjacent sub blocks are weighted to acquire the mapping relationship between arbitrary points, and finally the fog-free image of high comparison is output; through the simple mapping relationship, image computation is simplified; the algorithm efficiency is improved; and through pixel by pixel processing, the effect of image defogging is greatly improved.

The invention relates to an image preprocessing method for night lane line detection. By means of image analysis of a night lane line, it is concluded that lane line information required by lane line detection is concentrated in the area with a high grey level, and a road surface and other useless information are basically concentrated in the area with a low grey level. In order to obtain a better contrast enhancement effect, the gray threshold cutting is conducted on the lane marking part and the road surface background part firstly, local histogram equalization is conducted on the lane marking part and the road surface background part respectively on the basis, so that the contrast ratio is enhanced.

The invention relates to a remote sensing image segmentation method based on nonnegative low-rank sparse correlated drawing, belonging to the field of remote sensing image processing. The invention aims to solve the problem of low accuracy of segmentation of remote sensing images due to high homogeneity of texture information in high resolution remote sensing images. The image segmentation method includes that the quantification processing of remote sensing images is carried out and the input image to be processed is quantized by the K-means clustering classification method according to the image gray level range; the local histogram characteristics of the image texture information are extracted; the local histogram characteristic matrix l1 / 2 norm constraint is subjected to low-rank decomposition; the low-rank sparse correlated drawing is constructed; the characteristic matrix is segmented based on the nonnegative matrix parameterization method of correlated drawing constraint, and for the weight matrix decomposed by the non-negative matrix parameterization method, the category maximum weight corresponding to the characteristic vector of each pixel is found in the weight matrix by a weighting-off convolution method to determine the classification category of the pixels is determined, and the image segmentation is realized. The remote sensing image segmentation method is used for remote sensing image segmentation.

The present invention relates to a detail-based partially selective mapping infrared image enhancement method which mainly includes four steps: image detail acquisition, detail-based fast segmentation, scene judgment and scene-based mapping, and local histogram interpolation splicing, which can effective compress the useless gray range, making the contrast of the entire image enhanced and effectively improving the contrast of the target of interest while maintaining the background of the natural consistency.

The invention relates to a palm vein recognition system. According to the palm vein recognition system of the invention, An ROI (region-of-interest) image is selected; the difference image of the ROI image is calculated; a local histogram equalization image and a curvature image are calculated based on the difference image; a preprocessed image is obtained; when feature extraction is carried out, a certain number of fixed sampling points are selected from the preprocessed image, the absolute values of Gabor wavelet coefficients in a plurality of directions and at plurality of high frequencies and low frequencies are calculated for each sampling point; high-frequency vector normalization is carried out for each sampling point, so that a high-frequency feature can be obtained; the difference of a low-frequency vector and a 4-neighborhood low-frequency vector is calculated for each sampling point, so that a low-frequency feature can be obtained; when comparison is carried out, the scalar products of the high-frequency feature vectors of the corresponding sampling points of two images are calculated; and the scalar products are added together, so that high frequency similarity can be obtained; corresponding low-frequency feature vectors of the two images are obtained; the number of low-frequency feature vectors of which the corresponding components are identical is divided by the total dimensionality number of the vectors, so that low frequency similarity can be obtained; and the high frequency similarity and the low frequency similarity are added according to a preset weight, so that total similarity can be obtained.

The invention discloses a decision tree-oriented transverse federation learning method. The method comprises the steps that all participants search a quantile sketch of each feature in a data featureset based on a dichotomy; the participants construct a local histogram for each feature by using locally held data features according to the quantile sketch; noise meeting differential privacy is added to all the local histograms, and the noise is sent to a coordinator after being processed through a safety aggregation method; the coordinator merges the local histogram of each feature into a global histogram, and trains a root node of a first decision tree according to the histogram; the coordinator sends the node information to other participants; and all participants update the local histogram and repeat the above process for training to obtain a trained decision tree. The transverse federated learning method has the advantages of being easy and convenient to use, efficient in training and the like, data privacy can be protected, and quantitative support is provided for the data protection level.