37 results about "Color normalization" patented technology

A method for improvement of the consistency of color and brightness across boundaries of multicamera and / or multidisplayed overlapping or near overlapping composite images, reduction of objectionable artifacts at overlapping image seams of composite images, adjustment of color and brightness / intensity on either side of overlapping image seams, application of color correction across a composite image, reduction of color and intensity mismatches of composite images, gradual change of color across image seams of a composite image, including altering color outside of overlap regions is provided so that the seams are less discernible and to avoid sharp color changes across a composite or mosaic image are reduced or avoided. Color difference between two source images which form a composite is estimated by looking at color pixels from each source image region, determining a centroid of a cluster in a color space formed by the pixels of each source image, and determining the difference between cluster centroids for overlapping or nearly overlapping source image regions which are a measure of the vector color difference between such regions. The vector color difference between at least two overlapping source image regions are interpolated or mapped across those images.

A background removal apparatus comprises a color normalization module that normalizes an original image, and an image segmentation module that segments the normalized image into several segmented regions. A background mesh generation module simulates the variation of pixel colors in background and generates an interpolated background mesh. A comparison module compares the normalized image and the interpolated background mesh to form a background mask by extracting the coherent regions between them. A refinement module uses a refined rule to determine a final background mask, and a background removal module through which a pure foreground image is obtained. The apparatus combines the efficiency of both color and spatial clustering, and improves the capabilities of current image segmentation method to perform background removal.

The invention provides an automatic analysis system and method for a full-slice digital pathological image. The method comprises the steps of: obtaining the full-slice digital pathological image; carrying out partitioning processing on the full-slice digital pathological image; dividing the pathological image block into a background blank area and a foreground tissue area, and executing color normalization on the foreground area image block to obtain small blocks which can be directly used for analyzing the foreground tissue area; after image preprocessing, classifying the pathological image blocks to obtain a predictor type classification result, a segmentation result and a target detection result of each pathological image small block; after a full-slice digital pathological image is acquired, partitioning and preprocessing the full-slice digital pathological image, so that image classification, segmentation and target detection tasks can be carried out by using a deep learning model, and then visual display is carried out, thereby assisting a pathologist to observe and analyze the image, and improving the diagnosis efficiency and the diagnosis accuracy.

The invention discloses a method for aerial photography map splicing based on super-pixels and SIFT. The method comprises the steps that (1) super-pixel self-adaptation segmentation is conducted on aerial photography maps; (2) three to ten super-pixel pre-matching pairs are obtained by using the color normalization average value of a plurality of super-pixels obtained in the step (1) as guidance information, SIFT features of the super-pixel pre-matching pairs are extracted, and the super-pixel pre-matching pairs which fail in matching are removed after matching of the SIFT features is conducted; (3) a supervisory mechanism is generated in an on-line mode, and matching point pairs meeting the requirements are obtained; (4) a transformation matrix is calculated through the matching point pairs which are accurately selected in the step (3), images are fused, and image splicing is achieved. The method for aerial photography map splicing based on the super-pixels and SIFT achieves rapid and efficient splicing of the aerial photography maps, and the effectiveness of the method is verified through experiments.

The invention discloses a color image color feature extraction method and a device thereof used in the field of image processing technology. According to a traditional color image segmentation method, a high-quality refined segmentation result is difficult to obtain, and real-time performance can not meet a requirement of a large-scale database. The method of the invention comprises the following steps of: establishing a background model when receiving a video image, and extracting a foreground image of the video image; carrying out image normalization processing on the obtained foreground image, and obtaining a normalized image which is taken as an object of image data segmentation; according to dissimilarity measurement of color characteristics between pixel points, utilizing an image segmentation algorithm to carry out clustering on each pixel point of the image, dividing the image into pieces of blobs, and extracting color and position characteristics corresponding to each blob, wherein a color difference in the blob is small and a color difference between the blobs is large. According to the method and the device, target image recognition can be carried out accurately and efficiently with high real-time performance. If not specified, the image referred in the invention is a color image.

A system (100), method and computer program product for determining plant diseases. The system includes an interface module (110) configured to receive an image (10) of a plant, the image (10) including a visual representation (11)of at least one plant element (1). A color normalization module (120) is configured to apply a color constancy method to the received image (10) to generate a color-normalized image. An extractor module (130) is configured to extract one or more image portions (11e) from the color-normalized image wherein the extracted image portions (11e) correspond to the at leastone plant element (1). A filtering module (140) configured: to identify one or more clusters (C1 to Cn) by one or more visual features within the extracted image portions (11e) wherein each cluster isassociated with a plant element portion showing characteristics of a plant disease; and to filter one or more candidate regions from the identified one or more clusters (C1 to Cn) according to a predefined threshold, by using a Bayes classifier that models visual feature statistics which are always present on a diseased plant image. A plant disease diagnosis module (150) configured to extract, byusing a statistical inference method, from each candidate region (C4, C5, C6, Cn) one or more visual features to determine for each candidate region one or more probabilities indicating a particulardisease; and to compute a confidence score (CS1) for the particular disease by evaluating all determined probabilities of the candidate regions (C4, C5, C6, Cn).

The invention discloses a pathological section color normalization method and system, and the method comprises the steps: generating a picture of a target style through a generator, carrying out the discrimination of the generated picture and a real picture of the target style through a discrimination network, and executing the learning and discrimination adversarial training in a domain. In orderto reduce the difference between the generated picture of the non-target style picture and the target style picture, the generated picture of the non-target style picture and the target style pictureare identified through another identification network, inter-domain learning and identification confrontation training are executed, the difference between the generated picture of the non-target style picture and the target style picture is further reduced, and the performance of the generation network is optimized. According to the method, color normalization is carried out on pathological section data of different color styles, and the technical problems that a depth model trained under a single color style is difficult to have the same or similar performance in data of another color style, and the model is difficult to converge when pathological sections of different color styles are used as data to train the depth model are solved.

The invention belongs to the field of image processing and provides a reference-free tone mapping image quality evaluation method based on color statistics and information entropy. According to the method, the technical proposal adopted by the invention is that an image quality evaluation method without reference tone mapping based on color statistics and information entropy is provided. The method comprises the following steps: first, data preparation; 2, extract color statistical features, mainly comprise that following three steps: (1) decomposing color channel; 2) carry out division normalization operation on that channels in (1) to obtain respective color normalization coefficient; (3) parameter fitting; 3, feature extraction of information entropy; Step 4, quality prediction. Featureextraction of the test set, input to the trained model, predict the quality of the test chart score. The invention is mainly applied to image processing occasions.

The invention discloses a cervical abnormal cell detection method and system, and relates to neural network detection of abnormal cells. The method comprises the following steps: establishing an abnormal cell detection model, and labeling a cervical slide image as a labeled image containing pathological information; segmenting the annotated image into block images; respectively performing multi-scale transformation on the block images to respectively obtain scale images, and in a coordinate system established by taking the upper left corner of each scale image as an original point, marking real position coordinates and pathological marking information corresponding to each marking box; performing color normalization processing on the scale images to obtain normalized images; screening outeffective annotation images of which the annotation box widths and heights are smaller than those of the filter boxes from the normalized images; using a fixed frame to select the image area with themost tagging frames in the effective tagging images as a training image to obtain an abnormal cell detection model; the method has the beneficial effect that the detection accuracy is greatly improvedby carrying out multi-scale transformation on the cervical slide image.

Systems and methods are provided for capturing time-stamped data from whiteboard video signals and producing high-resolution whiteboard images. Local patches around a multitude of pixels in the whiteboard are used in classifying background white pixels and foreground color pixels for each foreground marker color. Clustering is performed in alternative color spaces globally and locally in defining background white and each foreground marker color. Color normalization is performed for each foreground pixel classified as a foreground marker color and for each image sensor color plane separately utilizing the maximum local background white and the darkest pixel intensities in local patches. Strokes are reconstructed based on spline interpolation of inflection points of cross sections along the length of each stroke for a foreground marker color with a predetermined width. Also provided is an intelligent whiteboard collaboration system including a messaging utility whereby participants based on relevant biometrics information are enabled to access time-lapse whiteboard data and communicate with the system and other participants.

The present invention discloses a normalization method for a multi-feature point constraint histogram of remote sensing image color normalization. The method comprises: accounting histograms of input images and reference images separately and performing normalization to obtain ratios of different gray-scale values, and performing filtering by using a gaussian filter to obtain smooth histograms; considering the smooth histograms as curves formed by connecting corresponding gray-scale value ratios of the gray scale in ascending order, and extracting feature points by a Douglas algorithm; based on gray scale range normalization treatment of the histograms, establishing correspondence relations between the feature points according to a minimum distance and a feature point type; establishing a gray scale equation from the input images to the reference images by using histogram normalization under a constraint of the feature points; and performing gray scale resampling on the input images according to the gray scale equation to obtain a result image. According to the normalization method for a multi-feature point constraint histogram of remote sensing image color normalization provided by the present invention, gray scale value compression or expansion situations of different gray scale ranges can be fitted, so that error accumulation and transfer are overcome.

The invention relates to the technical field of relay protection pressing plates, in particular to a method for recognizing the in-out service state of the relay protection pressing plate based on color template matching. The method adopts a template matching technology, uses the steps of image photographing of the relay protection pressing plate, image preprocessing, template acquisition, image recognition and the like, avoids the defects that different types of protection screens need to design different algorithms, reduces the difficulty of use, has high practicality, solves the influence of the photographing angle and the distance on the recognition result, can be applied to a handheld terminal or a mobile phone, greatly facilitates the use for operating personnel, adopts a RGB color normalization method, eliminates the effects of light and improves the reliability of recognition.

The invention discloses a color normalization method for HE staining pathological images. The method comprises the following steps: a step of subjecting pathological slice images to channel decomposition operation and obtaining pathological images, a step of calculating an optimal deconvolution matrix via establishment of an optimization model, and a step of using the optimal deconvolution matrixto normalizing the pathological images; the color of the pathological images under different staining conditions tends to be uniform after normalization, and fast calculation speed is realized.

The invention discloses a pathological image automatic classification method based on a dyeing intensity matrix. The method comprises directly extracting a dyeing intensity matrix irrelevant to a dyeing agent ratio, a dyeing platform, a scanning platform and some human factors in a pathological image as classified feature information. A normalized dyed image does not need to be recovered, so that all impurity-free information related to diagnosis is reserved, and meanwhile, the phenomenon that the diagnosis effect of an existing pathological image computer-aided diagnosis method based on a traditional color normalization method is changed along with the change of a selected standard pathological section is avoided, and errors caused by the fact that the dyed image needs to be recovered are avoided. According to the invention, the diagnosis precision is higher, the diagnosis effect is more stable, meanwhile, diagnosis of pathological images can be achieved in a shorter time, implementation is easy, and practicability is higher.

The invention discloses an image color extraction processing structure, which is used in the technical field of image processing, and solves the problems that the traditional color image segmentation device has difficulty acquiring high-quality and elaborative segmentation results and the real-time performance cannot meet the requirements of a large-scale database. The image color extraction processing structure comprises the steps of: establishing a background model when receiving a video image, and extracting a foreground image of the video image; subjecting the acquired foreground image to color normalization processing, and regarding a normalized image as an object of image data segmentation; and clustering pixels of the image by adopting an image segmentation algorithm according to dissimilarity measure of color features among the pixel points, segmenting the image into pieces of blob sub-regions, and extracting color and position features corresponding to each sub-region. The image color extraction processing structure can perform target image recognition precisely and effectively, and has high real-time performance. If not specified, the images referred in the invention are color images.

Techniques are described in which a device is configured to determine an overlap region between a first image and a second image, determine a first histogram based on color data included in the firstimage that corresponds to the overlap region, and determine a second histogram based on color data included in the second image that corresponds to the overlap region. The processor is further configured to determine, based on the first and second histograms, a mapping function that substantially maps the second histogram to the first histogram and apply the mapping function to the second image togenerate a normalized second image with respect to the first image.

The invention discloses a method for image feature extraction in that field of image process technology. The traditional color image segmentation method is difficult to obtain high-quality and fine segmentation results, and the real-time property cannot meet the requirements of large-scale database. The method comprises the following steps: an image feature extraction method is characterized in that: background modeling is carried out on the image, so as to extract the foreground image of the image; color normalization of foreground map; The normalized image is segmented into several blob sub-regions by using segmentation algorithm; color and position features corresponding to each blob sub-region are extracted. The invention can accurately and effectively carry out target image recognition, and has high real-time performance. As not specifically describe, that images refer to in the present invention are all color images.

The invention discloses a dynamic three-dimensional measurement method and system for a single color fringe pattern. The method comprises the following steps: S1, projecting three uniform red, green and blue intensity patterns with consistent intensity; shooting a projected uniform intensity pattern, and collecting red, green and blue uniform intensity patterns; S2, calculating a coupling strength coefficient between color channels; S3, projecting a color fringe pattern to the to-be-measured object and shooting the color fringe pattern; S4, decoupling gray fringe patterns of three channels in the color fringe pattern; S5, decomposing the decoupled color fringe pattern to obtain three background-free fringe patterns and three background patterns; S6, solving a modulation degree ratio by using the background image to obtain a color-normalized background-free fringe pattern; S7, demodulating the background-free fringe pattern after color normalization, and obtaining a demodulation phase; and S8, unwrapping the demodulation phase, solving an unwrapped phase, and reconstructing a three-dimensional shape. According to the method, three gray fringe patterns are coded on one color fringe pattern, and a gray phase shift fringe pattern meeting the demodulation requirement of a phase shift method is obtained through projection and decoupling.

The invention discloses a method for aerial photography map splicing based on super-pixels and SIFT. The method comprises the steps that (1) super-pixel self-adaptation segmentation is conducted on aerial photography maps; (2) three to ten super-pixel pre-matching pairs are obtained by using the color normalization average value of a plurality of super-pixels obtained in the step (1) as guidance information, SIFT features of the super-pixel pre-matching pairs are extracted, and the super-pixel pre-matching pairs which fail in matching are removed after matching of the SIFT features is conducted; (3) a supervisory mechanism is generated in an on-line mode, and matching point pairs meeting the requirements are obtained; (4) a transformation matrix is calculated through the matching point pairs which are accurately selected in the step (3), images are fused, and image splicing is achieved. The method for aerial photography map splicing based on the super-pixels and SIFT achieves rapid and efficient splicing of the aerial photography maps, and the effectiveness of the method is verified through experiments.

The invention discloses a method for automatically classifying pathological images based on a staining intensity matrix. The method directly extracts the staining intensity matrix irrelevant to the dyeing agent ratio, staining platform, scanning platform and some human factors in the pathological image as the characteristic information of the classification. There is no need to restore the normalized stained image, and while retaining all the impurity-free information related to the diagnosis, it avoids the diagnostic effect of the existing computer-aided diagnosis method for pathological images based on the traditional color normalization method. The phenomenon that the standard pathological slice changes, and avoids the error introduced by the need to restore the stained image, the diagnosis accuracy is higher and the diagnosis effect is more stable, and the diagnosis of the pathological image can be realized in a shorter time , easy to implement and more practical.

The invention discloses a fast color fringe pattern phase demodulation algorithm and system, and the method comprises the steps: firstly carrying out the pre-calibration of a measurement system through a calibration flat plate, and solving a coupling strength coefficient; secondly, using the system to respectively project a uniform intensity graph and a color fringe graph to an object to be measured, after the two graphs are respectively decoupled, using the uniform intensity graph to solve a background item and a modulation degree, and performing background removal and color normalization on the color fringe graph to obtain three background-free and color-normalized fringe patterns; adopting a phase shift method to solve demodulation phases of the three fringe patterns; and finally, unwrapping the obtained demodulation phase to obtain an unwrapped phase. The phase shift method demodulation of the single color fringe pattern is realized after background item calibration, the robustness of the demodulation of the traditional gray phase shift method is reserved, the fast phase demodulation is realized, and the requirement of dynamic three-dimensional shape measurement is met.

Systems and methods are provided for capturing time-stamped data from whiteboard video signals and producing high-resolution whiteboard images. Local patches around a multitude of pixels in the whiteboard are used in classifying background white pixels and foreground color pixels for each foreground marker color. Clustering is performed in alternative color spaces globally and locally in defining background white and each foreground marker color. Color normalization is performed for each foreground pixel classified as a foreground marker color and for each image sensor color plane separately utilizing the maximum local background white and the darkest pixel intensities in local patches. Strokes are reconstructed based on spline interpolation of inflection points of cross sections along the length of each stroke for a foreground marker color with a predetermined width. Also provided is an intelligent whiteboard collaboration system including a messaging utility whereby participants based on relevant biometrics information are enabled to access time-lapse whiteboard data and communicate with the system and other participants.

A vehicle brand picking-up method bases on Wavelet transform and Radon transform commutation. Switch the really color picture to gray-level picture. Calculate the vertical grads of originality picture and bases the result of Wavelet transform of its plane projection curve to wide orientation about the vehicle brand. Take plane Radon transform to the vehicle brand after wide orientation and locate and delete up and down frames of vehicle brand. Take vertical Radon transformation to the vehicle brand deleted the up frame and down frame and locate and delete left frame and right frame of vehicle brand. Consolidate the colors to get one vehicle brand without frame and having identifiable number. It picks up vehicle brand effectively, quickly and precise in common condition and set veracious identify foundation to the word of vehicle brand.

Systems and methods are provided for capturing time-stamped data from whiteboard video signals and producing high-resolution whiteboard images. Local patches around a multitude of pixels in the whiteboard are used in classifying background white pixels and foreground color pixels for each foreground marker color. Clustering is performed in alternative color spaces globally and locally in defining background white and each foreground marker color. Color normalization is performed for each foreground pixel classified as a foreground marker color and for each image sensor color plane separately utilizing the maximum local background white and the darkest pixel intensities in local patches. Strokes are reconstructed based on spline interpolation of inflection points of cross sections along the length of each stroke for a foreground marker color with a predetermined width. Also provided is an intelligent whiteboard collaboration system including a messaging utility whereby participants based on relevant biometrics information are enabled to access time-lapse whiteboard data and communicate with the system and other participants.

The invention discloses a face recognition method based on self-adaption and color normalization. The face recognition method comprises the following steps: S1, creating a face image data set; S2, performing adaptive scale selection processing on the images in the face data set based on a face detection network to obtain new face images; S3, carrying out face positioning on the output new face image, carrying out color normalization processing on the eye periphery area of the positioned face, and obtaining a processed face image; S4, inputting the processed face image into a face recognition network, calculating the Euclidean distance between the vector of the processed face image and a pre-stored face image, judging whether the Euclidean distance is smaller than a first preset threshold value or not, if so, obtaining a training sample, and inputting the training sample into a face recognition network for training to obtain a final face recognition network model; and S5, inputting an image to be recognized into the face recognition network model for face recognition to obtain a final recognition result.

The present invention discloses a normalization method for a multi-feature point constraint histogram of remote sensing image color normalization. The method comprises: accounting histograms of input images and reference images separately and performing normalization to obtain ratios of different gray-scale values, and performing filtering by using a gaussian filter to obtain smooth histograms; considering the smooth histograms as curves formed by connecting corresponding gray-scale value ratios of the gray scale in ascending order, and extracting feature points by a Douglas algorithm; based on gray scale range normalization treatment of the histograms, establishing correspondence relations between the feature points according to a minimum distance and a feature point type; establishing a gray scale equation from the input images to the reference images by using histogram normalization under a constraint of the feature points; and performing gray scale resampling on the input images according to the gray scale equation to obtain a result image. According to the normalization method for a multi-feature point constraint histogram of remote sensing image color normalization provided by the present invention, gray scale value compression or expansion situations of different gray scale ranges can be fitted, so that error accumulation and transfer are overcome.

The invention discloses a foreground detection method based on adaptive background updating and selective background updating, which belongs to the technical field of image processing. The present invention first converts the image frame of the video stream to be detected into a grayscale image and an HLS image; if the current image frame is the first frame of the video stream, its grayscale image is used as the background model; for the image frame other than the first frame, If the number of image frames does not exceed the threshold, if so, then continuously update the background model based on the grayscale image of the current frame until the number of image frames exceeds the threshold; when it exceeds the threshold, enter the foreground image detection processing step: based on the background model and the current The grayscale image of the frame is used to obtain the first foreground image, and then the frame difference method is used to obtain the second foreground image, and the union of the two is obtained to obtain the third foreground image, and the third foreground image is processed through the color normalized correlation coefficient, so that Get pixel-level foreground objects. The invention has good detection effect on moving and still objects, and has better robustness.

A simple image color normalization method comprises the following steps: selecting a reference picture and other pictures as comparison pictures to carry out color normalization; carrying out sift feature point matching on a comparison picture needing normalization and the reference picture, and finding out all coarse matching feature points; for all the rough matching feature points, taking one feature point, finding out the first two feature points closest to the Euclidean distance in the reference picture image, and in the two key feature points, if the ratio of the nearest distance dividedby the next close distance is smaller than a certain threshold value, accepting the pair of matching points; sorting all the coarse matching feature points to obtain R, G and B values of all the matching points on the two pictures, and setting the R, G and B values as A (r, g, b) and B (r, g, b); and calculating a correction parameter and correcting the picture needing to be normalized by using the correction parameter. According to the method, picture colors can be normalized quickly and accurately, and data acquisition of pictures is facilitated.