2049 results about "Image restoration" patented technology

A method of digital image enhancement and sharpening (11, 12) which may be applied to restoration of blurred digital images of arbitrary origin. The method uses a specially formulated constraint (15) to reconstruct the original images. In particular, the constraint of the present method minimizes the area where strong image parameter variations and discontinuities occur. This new constraint is called a minimum gradient support (MGS) constraint. The MGS constraint generates a stable sharp solution of the linear ill-posed image restoration equation with an arbitrary blurring operator.

Digital focusing method and apparatus in an image processing system, for digitally focusing an out-of-focus image, are provided. A defocused image is divided into sub-images of a predetermined size. An edge direction of each of the divided sub-images is estimated. Step responses with respect to the respective edge directions are calculated. A mean step response is obtained by averaging a predetermined number of the step responses. Point Spread Function (PSF) coefficients are obtained using the mean step response. An image blur transfer function is obtained using the PSF coefficients. An image restoration transfer function is obtained using the image blur transfer function. An original in-focused image is obtained by multiplying the image restoration transfer function by the defocused image in a frequency domain. Thus, an image can be restored in real time, and the size and weight of the image processing system can be reduced.

The present invention provides a compressed low-resolution image restoration method based on a combined deep network, belonging to the digital image / video signal processing field. The compressed low-resolution image restoration method based on the combined deep network starts from the aspect of the coprocessing of the compression artifact and downsampling factors to complete the restoration of a degraded image with the random combination of the compression artifact and the low resolution; the network provided by the invention comprises 28 convolution layers to establish a leptosomatic network structure, according to the idea of transfer learning, a model trained in advance employs a fine tuning mode to complete the training convergence of a greatly deep network so as to solve the problems of vanishing gradients and gradient explosion; the compressed low-resolution image restoration method completes the setting of the network model parameters through feature visualization, and the relation of the end-to-end learning degeneration feature and the ideal features omits the preprocessing and postprocessing; and finally, three important fusions are completed, namely the fusion of the feature figures with the same size, the fusion of residual images and the fusion of the high-frequency information and the high-frequency initial estimation figure, and the compressed low-resolution image restoration method can solve the super-resolution restoration problem of the low-resolution image with the compression artifact.

The invention provides an image inpainting method and system based on antagonistic generation neural network, which comprises the following steps: firstly, constructing a self-encoder convolutional neural network (including encoder and encoding discriminator), a decoder (generator) convolutional neural network, a discriminator convolutional neural network, a global discriminator, and a local discriminator; Then, different loss functions are constructed for the five networks, and the whole network is trained by step-by-step training. Finally, when the network training is completed, the defect image is put into the network for repairing, and the result graph generated by the decoder (generator) is the final repairing result graph. The invention has the advantages that: the potential constraint of the image is kept and the image is thinned; an end-to-end image restoration network is implemented. The dependence of the repair network on the missing position mask information of the image iseliminated. The robustness in practical application is improved.

The invention discloses a semantic image restoration method based on a DenseNet generative adversarial network. The method comprises the following steps of preprocessing collected images, and constructing a training and testing data set; constructing a DenseNet generative adversarial network; training a DenseNet generation adversarial network stage; and finally, realizing repair processing of thedefect image by using the trained network. Under the framework of the generative adversarial network, the DenseNet structure is introduced, a new loss function is constructed to optimize the network,the gradient disappearance is reduced, network parameters are reduced, meanwhile, transmission and utilization of features are improved, and the similarity and visual effect of large-area semantic information missing image restoration are improved and improved. Examples show that face image restoration with serious defect information can be achieved, and compared with other existing methods, the restoration results better conform to visual cognition.

The invention discloses a face image restoration method based on a generation antagonism network. The method comprises the following steps: a face data set is preprocessed, and a face image with a specific size is obtained by face recognition of the collected image; In the training phase, the collected face images are used as dataset to train the generating network and discriminant network, aimingat obtaining more realistic images through the generating network. In order to solve the problems of instability of training and mode collapse in the network, the least square loss is used as the loss function of discriminant network. In the repairing phase, a special mask is automatically added to the original image to simulate the real missing area, and the masked face image is input into the optimized depth convolution to generate an antagonistic network. The relevant random parameters are obtained through context loss and two antagonistic losses, and the repairing information is obtainedthrough the generated network. The invention can not only solve the face image repairing with serious defective information, but also generate a face repairing image which is more consistent with visual cognition.

An image stabilizing device has an image restoration processor configured to perform image restoration process for stabilizing hand movement included in an input image by using an image restoration filter created based on hand movement information, and an edge emphasizing processor configured to perform edge emphasizing process for emphasizing an edge of the input image. The image stabilizing device has a stabilization mode determiner configured to determine an image stabilization mode for stabilizing the input image based on a size of the hand movement created by the hand movement information. The image stabilization mode includes a first mode for performing the image restoration process, a second mode for performing the edge emphasizing process, and a third mode for performing none of the image restoration process and the edge emphasizing process.

The present invention discloses a method and a device for identifying a reticulate pattern face image based on a multi-task convolutional neural network. The method comprises the steps of: collecting reticulate pattern face image and corresponding clear face image pairs, then using the multi-task convolutional neural network to respectively design object functions based on regression and classification, training a face image reticulate pattern removing model, and finally inputting the reticulate pattern face image into the trained reticulate pattern removing model to obtain a face image without reticulate pattern, thereby performing subsequent face image identification tasks. According to the method, a multi-task learning frame is adopted, the task for restoring a reticulate pattern image to a clear image is expressed as two object functions which are assistant with each other, and the convolutional neural network is utilized to learn complicated nonlinear transformation referred therein. The method not only effectively improves convergence rate during model training, but also can greatly improve image restoration effect and generalization ability, thereby greatly improving identification accuracy rate of the reticulate pattern face image.

A blur correction camera system includes a blur correction lens driven based upon the vibration detection signal detected by an angular velocity sensor, that corrects an image blur, a point-image function computing unit that computes a point spread function, and an image restoration computing unit that corrects an image blur by executing image restoration through image processing on a captured image by using the point spread function. The image blur that cannot be completely corrected by the blur correction lens is further corrected through image restoration so as to obtain a high quality image.

An image compression divides an input image into blocks having the predetermined number of horizontal and longitudinal pixels, and scans the divided blocks in a main scan direction and a sub scan direction so as to select them in order as a processing block, and selects as a reference block a block in which positional relation with the selected block and a relation of a pixel value satisfies a predetermined condition. Subsequently, the pixel values of the processing block and the reference block are subjected to an exclusive-OR (XOR) so as to generate a differential image, and in case the differential image satisfies the predetermined condition, the processing block is replaced with the differential image. Further, the image including the differential image obtained by executing the block replacement in the processing block is encoded, and the code data obtained by executing this image encoding, the presence or absence of replacement with the differential image of each processing block obtained by executing the block replacement, and positional information of the reference block are combined and outputted. The image decoding is executed in such way that the image including the differential image from the code data is decoded, and this decoded image is decoded into the original image by an exclusive-OR with the processing block which is divided into blocks and the reference block.

An image processing device capable of obtaining a high precision image while aberration asymmetry is corrected.An image processing device includes image obtaining means configured to obtain an input image, image restoration means configured to restore the input image using a generated or selected image restoration filter in accordance with a transfer function of an image pickup system used for forming the input image from an object image. The image restoration filter makes a difference between absolute values of transfer functions of two azimuthal directions at a time when the restoration image is obtained from an object smaller than a difference between absolute values of transfer functions of the two azimuthal directions of the image pickup system.

The invention discloses a significant object detection method based on sparse subspace clustering and low-order expression. The method comprises the steps of: 1, carrying out super pixel segmentation and clustering on an input image; 2, extracting the color, texture and edge characteristics of each super pixel in clusters, and constructing cluster characteristic matrixes; 3, ranking all super pixel characteristics according to the magnitude of color contrast, and constructing a dictionary; according to the dictionary, constructing a combined low-order expression model, solving the model and decomposing the characteristic matrixes of the clusters so as to obtain low-order expression coefficients, and calculating significant factors of the clusters; and 5, mapping the significant value of each cluster into the input image according the spatial position, and obtaining a significant map of the input image. According to the invention, the significant objects relatively large in size in the image can be completely and consistently detected, the noise in a background is inhibited, and the robustness of significant object detection of the image with the complex background is improved. The significant object detection method is applicable to image segmentation, object identification, image restoration and self-adaptive image compression.

An image processing method includes the steps of obtaining an image generated by an image pickup system, and performing correction processing for the image by utilizing an image restoration filter generated or selected based on an optical transfer function of the image pickup system. The image restoration filter is a filter configured to reduce a phase degradation component of the image.

The temperature distribution associated with a design of an integrated circuit is calculated by convoluting a surface power usage represented by a power matrix with a heat spreading function. The heat spreading function may be calculated from a simulation of a point source on the integrated circuit using a finite element analysis model of the integrated circuit or other techniques. To account for spatial variations on the chip, the heat spreading function may be made dependent on position using a position scaling function. Steady-state or transient temperature distributions may be computed by using a steady-state or transient heat spreading function. A single heat spreading function may be convolved with various alternative power maps to efficiently calculate temperature distributions for different designs. In an inverse problem, one can calculate the power map from an empirically measured temperature distribution and a heat spreading function using various de-convolution techniques. While the forward problem is analogous to image blurring, the inverse problem is analogous to image restoration.

The image processing method includes a step of acquiring an input image produced by image capturing using an optical system, an image restoration step of performing an image restoration process on the input image to produce a restored image, the image restoration process using an image restoration filter produced based on an optical transfer function of the optical system, and a color shift detection step of detecting color shift included in the restored image, the color shift corresponding to chromatic aberration of magnification of the optical system. The method further includes a restored image color shift reduction step of performing a color shift reduction process on the restored image to reduce the color shift detected in the color shift detection step.

An image acquisition apparatus has an imaging optical system. An imaging section has a plurality of imaging pixels arranged at non-uniform intervals within an imaging surface. The imaging section picks up an object image imaged by the imaging optical system, and converts the object image into image signals. An image restoration processing section has a coefficient memory in which a predetermined coefficient sequence is recorded, and an image restoration processing circuit which carries out computation processing between the coefficient sequence recorded in the coefficient memory and the image signals from the imaging section. The image restoration processing section generates an image in a desired display pixel arrangement.

The invention relates to a video image super-resolution reconstruction method. The method comprises the steps that 1), for a plurality of low-resolution images capable of being obtained through the same scene, continuous sequence frames of needed videos are selected and converted to static JPG or BMP files, and continuous multi-frame sub pixel images in a pixel are selected with the combination of scene parameters; 2) according to the image sequence frames of the multi-frame sub pixel images, a target interest point is selected, and motion estimation of a block-shaped target is conducted; 3) according to a motion estimation result, a non-uniform interpolation reconstruction algorithm is adopted, and a high-resolution image is reconstructed. According to the video image super-resolution reconstruction method, the resolution, on details, of a video image can be improved on the premise that the image quality of an existing imaging device is not improved, the defects in methods for traditional image restoration, enhancement and the like are overcome, and the video image super-resolution reconstruction method has the advantages of being easy and convenient to operate, and high in efficiency. The video image super-resolution reconstruction method can supply optimized image data foundations to the application in the image data field, and therefore the city video information application field can be enlarged.

A system for enhancing images from an electro-optic imaging sensor and for reducing the necessary focal length of a sensor while preserving system acuity. This system uniquely reduces the necessary focal length and enhances images by collecting a video sequence, estimating motion associate with this sequence, assembling video frames into composite images, and applying image restoration to restore the composite image from pixel, lens blur, and alias distortion. The invention synthetically increases the pixel density of the focal plane array. Thus it reduces the necessary size of the projected blur circle or equivalently it reduces the minimum focal length requirements.

The invention discloses a remote sensing image fusion method based on sparse representation. The method comprises the following steps of: firstly, establishing a linear regression model between a multispectral image and a brightness component thereof; secondly, performing sparse representation on a panchromatic image and the multispectral image by using high and low resolution dictionaries respectively, and acquiring sparse representation coefficients of the brightness component of the multispectral image according to the linear regression model; thirdly, extracting detail components according to the sparse representation coefficients of the panchromatic image and the brightness component, and implanting the detail components to the sparse representation coefficients of each band of the multispectral image under a general component replacement fusion framework; and finally, performing image restoration to obtain a multispectral image with high spatial resolution. According to the method, the sparse representation technology is introduced into the field of remote sensing image fusion, so that the defect that high spatial resolution and spectral information cannot be simultaneously preserved in the prior art is overcome; and the fusion result of the method is superior to that of the conventional remote sensing image fusion method on the aspects of spectral preservation and spatial resolution improvement.

Imaging apparatus includes a mosaic image sensor (24), which is configured to generate a stream of input pixel values belonging to a plurality of input sub-images, each sub-image responsive to light of a different, respective color that is incident on the mosaic image sensor. An image restoration circuit (26) is coupled to receive and digitally filter the input pixel values in each of the input sub-images so as to generate a corresponding plurality of enhanced output sub-images. An image signal processor (ISP) (28) is coupled to receive and combine the plurality of the output sub-images in order to generate a color video output image.

The invention discloses a laser two-dimension code identification method based on image restoration and belongs to the technical field of image processing. The laser two-dimension code identification method includes steps of preprocessing acquired metal part laser two-dimension code images, acquiring outer contour of two-dimension codes by means of image morphology, detecting outer edges of the two-dimension codes by Soble operator, roughly and accurately positioning the two-dimension codes, partitioning the two-dimension code images, restoring and identifying and the like. By the laser two-dimension code identification method, the more satisfying two-dimension code images can be redrawn, the two-dimension code images can be restored and identified, the problem that an existing identifying method is hard to extract correct two-dimension code data due to the fact that laser two-dimension code images on metal pieces are always fuzzy or defective because of chemical reaction of many metal products during marking and illumination and oil contamination and the like during acquisition is solved, and the method is applicable to popularization and application.

The invention relates to an image restoring method which dose automatically restoring course after user enters into the restored area. The automatically restoring course uses the edge side as the unit of primary iteration packing and then iterates the following four steps until packs all the restored area: first step, computing all the precedent degree of the edge; second step, ascertaining the packing edge side of the current iteration by the precedent degree and the similarity degree criterion; third step, adopting line synthesis method and expends the known line and structure information; fourth step, updating the synthetic picture element ascertain value of the current iteration side.

An image processing method comprises obtaining a color input image obtained by taking an image using an optical system 101, calculating an amount of characteristics relating to a color of a pixel in the input image as a first calculating step (S4), performing an image restoration process for the input image using an image restoration filter selected or made based on aberration information of the optical system (S5), calculating an amount of characteristics relating to a color of a pixel of a restored image obtained by the image restoration process as a second calculating step (S6), and correcting a signal value of a specific pixel in the restored image in accordance with the amount of characteristics of the specific pixel in the restored image calculated by the second calculating step and the amount of characteristics of a specific pixel of the input image calculated by the first calculating step (S8).

A gradient-based image enhancement and restoration method and system which applies an orientation-isotropy adaptive filter to the gradients of high structured regions, and directly suppresses the gradients in the noise or texture regions. A new gradient field is obtained from which image reconstruction can progress using least mean squares. The method generally comprises: inputting image data; calculating image gradients; defining the gradients as having large or small coherence; filtering the large coherence gradients for edge enhancement; suppressing the small coherence gradients for noise reduction; assembling an enhanced gradient field from the filtered large coherence and suppressed small coherence gradients; and optimizing the assembled gradient field into a restored image.

An image restoration method includes generating a plurality of damaged images from an image to be restored, and determining corresponding intact images in positions where the damaged images are, the damaged images being images fused by the damaged image and the intact images; using the generated plurality of damaged images and the corresponding intact images in the positions where the damaged images are as training samples, importing a generator of a generative adversarial network, performing feature fusion on an encoder and a decoder in the generator, and learning features of the image to berestored through the generative adversarial network; and restoring the image to be restored according to the trained generative adversarial network. The image to be restored is restored according to the trained generative adversarial network, the problem that a template for an area to be restored needs to be used when the image is processed can be solved, and through feature fusion of the encoderand decoder in the generator in the generative adversarial network, a large-area damaged module can be effectively restored.

The invention discloses a natural image super-resolution and non-uniform motion blur removing method based on a depth feature fusion network, which firstly realizes the restoration of a low-resolutionnon-uniform motion blur image based on a depth neural network. The network adopts two branch modules to extract features for image super-resolution and non-uniform blur removal respectively, and achieves adaptive fusion of the outputs of the two feature extraction branches through a feature fusion module that can be trained. Ultimately, the upsampling reconstruction module realizes the task of removing non-uniform motion blur and super-resolution. This method utilizes the self-generated training data set to train the network offline, thus realizing the restoration of low-resolution non-uniform motion blurred images of any input size. This method has low training difficulty, good effect and high computational efficiency, is very suitable for image restoration and enhancement of mobile devices and monitoring equipment.

A system that extracts text from an image includes a capture device that captures the image having a low resolution. An image segmentation subsystem partitions the image into image segments. An image restoration subsystem generates a resolution-expanded image from the image segments and negates degradation effects of the low-resolution image by transforming the image segments from a first domain to a second domain and deconvolving the transformed image segments in the second domain to determine parameters of the low-resolution image. A text recognition subsystem transforms the restored image data into computer readable text data based on the determined parameters.

The invention relates to a sparse representation-based blind restoration method of a broad image, aiming at solving the technical problem that an image restored by the existing broad image blind restoration method is bad in effect. The method has the technical scheme that the broad image is sparsely represented by a fuzzy dictionary and a brilliant image is rebuilt by a brilliant dictionary to restore the image due to the characteristic that the sparse coefficient of the broad image under the fuzzy redundancy dictionary is coincident with that of the brilliant image under the brilliant redundancy dictionary. The illcondition during deconvolution can be avoided, the ringing effect at a high edge during image restoration can be reduced, and the image which is more brilliant can be obtained.

The invention discloses an image restoration method based on image segmentation, and a system therefor; the method comprises: firstly, manually selecting and marking the area to be restored in image by a user; then, carrying out image domain decomposition by mean shift algorithm, and dividing the image into a number of areas; finally, carrying out repeated iterative operation on the area to be restored until all pixels in the area to be restored is filled to be full. The method optimizes the calculation of priority in image restoration algorithm, thus effectively preventing the over expansionof the restored image from a high-texture area to a low-texture area; furthermore, matched block searching standard based on the image domain decomposition can be formulated on that basis, so that anerroneous block can be avoided being introduced; compared with the original image restoration method based on the sample, the effect of the method is more in accordance with the visual expectation ofhuman beings; furthermore, at present, the method is successfully applied to large size area restoration of various images with complex texture and structural characteristics as well as the aspects such as wiping off characters, removing target objects and the like.

The invention discloses a face image restoration method based on a multi-discriminator generative adversarial network, and the method comprises the following steps: (1) carrying out the preprocessingof images in a disclosed face image database, and inputting the images into a generator, so as to obtain a generated image; (2) inputting the real image and the generated image into a plurality of discriminators to obtain a feedback value; (3) taking the feedback values of the plurality of discriminators as confrontation loss, and performing confrontation training on the generative adversarial network by combining the perception loss and the reconstruction loss; and (4) inputting the missing face image into the trained generator to obtain a repaired face image. Aiming at the problem of repairing the shielded or damaged face image, a generative adversarial network structure with multiple discriminators is adopted, so that the problem of low authenticity of the repaired image is solved, andthe repaired image is more natural and more real.