44results about How to "Preserve image detail" patented technology

The invention discloses video noise reduction device and a video noise reduction method. The method comprises the following steps of: obtaining a brightness difference histogram of a current image by using a denoising result of a previous frame of image and a gradient magnitude histogram of the current image; carrying out noise level evaluation on the current image according to the brightness difference histogram; calculating the spatial distance of any two pixel points in the current image, so as to obtain the spatial similarity of the any two pixel points; carrying out denoising on the current image according to the spatial similarity; calculating a pixel time domain distance between any pixel point in the current image and the pixel point at the position corresponding to the previous frame of denoised image, and calculating the corresponding time domain similarity; carrying out three-dimensional recursive denoising on the video image according to the obtained time domain similarity, the spatial similarity denoising result and the previous frame of denoising result. By adopting the device and method disclosed by the invention, three-dimensional recursive denoising is carried out by using the pertinence of the pixel in space and time, so that strong complicated noise can be removed; an image detail can be kept; the stability of the denoising effect also can be ensured.

The invention discloses a weighted adaptive super-resolution reconstructing method for an image sequence, which is superior to the conventional method in the aspects of robustness and practicability, and has important application value for obtaining a high-quality image. The method comprises the following steps: (1) acquiring a plurality of continuous frames of low-resolution images obtained by the same sensor, and resampling a low-resolution image sequence to obtain a resampled low-resolution image sequence; and (2) reconstructing a frame of high-resolution image by utilizing the resampled low-resolution image sequence, wherein the method for reconstructing the frame of high-resolution image comprises the following steps: firstly, establishing a high-resolution image degradation model; secondly, converting a solving process of the high-resolution image in the degradation model into an optimizing process of a solution of a reconstruction optimization model of the high-resolution image according to a predetermined degradation model of the high-resolution image and a regularization theory; and finally, optimizing the reconstruction optimization model of the high-resolution image by utilizing a progressive non-convex algorithm so as to obtain an optimal estimation value of the high-resolution image.

The invention discloses a space-time united image sequence multi-scale geometric transformation denoising method. By that an image multi-scale geometric analysis is introduced into the image denoising process, and image sequence space-time relevant information is united, a statistical distribution model of an image multi-scale geometric transformation domain coefficient, the built image statistical model serves as priori knowledge, a Bayesian lest square estimation method is adopted to obtain an optimal estimation result of non-noise-pollution image signals, and the problems that the situation that an image detail coefficient and image noise are difficult to distinguish often occurs in the denoising process of a conventional wavelet domain are solved. Meanwhile, plenty of image samples which are the same to a to-be-processed image in scene are collected and image statistical distribution modeling is conducted on the image samples, the statistical distribution model of the image multi-scale geometric transformation domain coefficient directly reflects statistical distribution features of texture details of edges and surface of the to-be-processed image and provides reliable priori distribution knowledge for Bayesian lest square estimation of the image, and space-time information united image sequence non-noise image signal estimation is achieved. A denoising effect of the image is improved, at the same time, image details are kept to a great extent.

The invention discloses an image processing method and apparatus, belonging to the image display field. The method is based on brightness compensation and contrast-tone adjustment, and successively performs brightness compensation and contrast-tone adjustment on an image to be displayed to obtain an output image after compensation processing, allowing the output image displayed by a display device to still possess enough contrast and tone, and meanwhile guarantee image details when the display device reduces brightness, and thereby allowing the image displayed by the display device to possess enough visibility while reducing the power consumption of the display device.

The invention provides a non-uniform illuminance image enhancement method, and belongs to the technical field of digital image processing. The method aims to solve the problems of an image enhancement method in the prior art; according to the method, image enhancement processing can be carried out on a non-uniform illuminance image. The non-uniform illuminance image enhancement method includes the following steps that a histogram of the non-uniform illuminance image is made, a threshold value is determined according to peak values of high illuminance and low illuminance, the histogram is divided into a high illuminance sub-band and a low illuminance sub-band, a sub-band brightness enhancement method is maintained for the high illuminance sub-band and a corresponding gray level mapping enhancement method is adopted for the low illuminance sub-band. According to the non-uniform illuminance image enhancement method, the situation that the image is excessively enhanced locally is avoided, overall noise of the image is inhibited, and the image which is uniform on the whole is obtained. The calculation amount of an algorithm is little, and the method is stable and reliable and can be used in real-time project systems.

The invention discloses a convolutional neural network (CNN) model generation method. The method comprises the following steps: constructing a CNN model structure and setting a loss function expression; A plurality of training image pairs are acquired, each training image pair comprises an input image and a target image, and the input image and the target image are images obtained by shooting thesame scene through a mobile terminal and a single lens reflex; Each training image pair is registered, so that an input image included in the training image pair is aligned with the target image; Using the registered training image pair as a training sample to train a CNN model; and continuously updating model parameters of the CNN model, calculating a loss function value of the CNN model according to the loss function expression every time the model parameters are updated, and when the loss function value is converged, stopping the training process to obtain the model parameters of the CNN model so as to generate the CNN model. The invention also discloses a corresponding image quality optimization method and computing equipment.

The invention provides a method for image noise reduction processing. The method for the image noise reduction processing comprises the following steps: a directional template is established and weight deviation is calculated according to the directional template; a weight deviation mean value is calculated according to the weight deviation; a determinative factor is calculated according to the weight deviation mean value; a weight filtering mean value is calculated according to the result and output as a strengthened image. According to the technical scheme, noise reduction of the image is achieved in a self-adaptive mode through consideration of directional information and deviation information. Meanwhile, image details can be well kept, the visual effect of the image can be improved in a self-adaptive mode, so that the influence by the noise is reduced to a certain extent and the method for the image noise reduction processing and the system for the image noise reduction processing are used for the noise reduction of video images.

The present invention relates to a glass defect type identification method. The method comprises: 1) providing a glass defect type identification system, wherein the identification system comprises a glass holding platform, an image collecting device, a defect image detection device and an embedded processing device; the glass holding platform is used for placing a to-be-tested glass monoblock; the image collecting device is used for collecting glass images of the to-be-tested glass monoblock; the defect image detection device is used for determining a defect image among the glass images; and the embedded processing device is connected with the defect image detection device, and determines a glass defect type based on the defect image; and 2) using the identification system to perform identification. According to the method provided by the present invention, different defect detection parameters can be determined according to different glass types, so that the accuracy of glass defect type identification is improved.

The invention discloses an image denoising method for adaptive equidistant template iteration mean filtering, aimed at addressing the problems of current adaptive methods, such as poor denoising effect and poor recovery quality. The method includes the following steps: (1) using an extremum method to determine noise spots; (2) conducting sliding rotation with the middle points on 4 edges of a filtering window as start points and 4 angular points as terminal points, taking 8 symmetrical points of the 4 edges of the filtering window to construct an equidistant template (the 4 middle points and 4 angular points on the filtering window are 8 value template special cases), and forming a first equivalence template, a second equivalence template......, and conducting recursion clipping mean filtering with the equivalence templates; (3) checking whether a noise point is completed after the completion of each filtering from 3X3, and if the noise point is not completed, enlarging the filtering window and stops until 7X7, and forming adaptive filtering; (4) if the noise point is not finished, adopting iteration filtering. According to the invention, the method effectively processes noise and at the same time can better protect image details, has a higher rate of using information and has a rapid denoising speed.

The invention relates to an artificial intelligence technology, and discloses a multispectral analysis-based night vision image quality improvement method, which comprises the following steps of: partitioning images in an original night vision image set, and carrying out adaptive training by utilizing the partitioned night vision images to obtain a local sparse structure dictionary; performing sparse denoising on the images in the original night vision image set by using the local sparse structure dictionary to obtain a night vision denoised image set, performing spectral filtering processing and color enhancement processing on the images in the night vision denoised image set to obtain an enhanced image set, and performing color recovery processing on the images in the enhanced image set to obtain an enhanced image set; and obtaining a standard night vision image set, and carrying out sub-histogram equalization processing on the images in the standard night vision image set to obtain an enhanced night vision image set. The invention further provides a night vision image quality improving device and equipment based on multispectral analysis. The method can improve the night vision image quality.

The invention discloses an SAR image segmentation method based on decomposition evolution multi-objective optimization and FCM. The method mainly solves the problem that in the prior art of image segmentation, image segmentation precision is not high, the evaluation index is single, and the segmentation effect is not ideal. The method comprises the steps that the Gabor feature and gray level symbiotic feature of each pixel of an image are extracted, and a superpixel is obtained through rough segmentation of a watershed, superpixel features are used as data to be clustered, a clustering center is used as individual species, the species are optimized through the decomposition evolution multi-objective method, the species obtained after evolution are used as the clustering center to initialize the FCM algorithm, a new clustering center is obtained and used as new species for participating in the next evolution of the decomposition evolution multi-objective algorithm. According to the SAR image segmentation method, the better clustering center is obtained through cross adoption of the decomposition evolution multi-objective algorithm and the FCM algorithm, the defect that the FCM initial value is sensitive and falls into a local optimal solution easily is overcome, and the better image segmentation result can be obtained.

The invention discloses a gradient-guided TV-Retinex single-frame image defogging method, belonging to the field of computer vision. The gradient-guided TV-Retinex single-frame image defogging methodcomprises the steps of firstly, realizing brightness offset correction by adopting a half-quadratic punishment calculation method and utilizing a gradient similarity according to a mechanism of imagedegradation caused by fog, and then combining a variational frame (TV-)Retinex and adopting a Split Bregman algorithm to perform image refinement and enhanced defogging. The gradient-guided TV-Retinexsingle-frame image defogging method has the advantages of being capable of restoring a clear image from an original image directly without pre-storing a fog-free image for reference, correcting imagebrightness, improving image detail definition, restoring a clear fog-free image, completing the defogging processing quickly, being adapt to real-time application, and improving the reliability and efficiency of an outdoor intelligent machine vision system.

The invention belongs to the technical field of modern art design, and discloses a system and a method for analyzing the representation of Chinese traditional elements in modern art design. The systemfor analyzing the representation of the Chinese traditional elements in the modern art design comprises a design drawing acquisition module, a main control module, an image feature extraction module,a matching module, a classification module, a retrieval module, an analysis module and a display module. Pictures are classified and stored through the classification module according to different numbers of the pictures, so that the classification efficiency of the design drawing is greatly improved; Meanwhile, a Gram matrix is used as an operator for describing image style characteristics through a retrieval module; style characteristics of all training samples of a given design image set are extracted, an image index is constructed, style characteristics of a to-be-detected image are extracted, an index number of the image with the highest similarity with the to-be-detected image is searched, and a similar image set is returned according to the index number, so that retrieval of the abstract image is realized through the method, and the method has the advantages of simplicity and high retrieval efficiency.