32results about How to "Increased PSNR value" patented technology

The invention discloses a two-channel convolutional neural network-based single image super-resolution calculation method. The method comprises the steps of (1) performing fuzzy degradation processing on a known high-resolution image to obtain a low-resolution image with the same size; (2) decomposing the obtained low-resolution image after the fuzzy processing in the step (1) into a texture part and a smooth structure part of the low-resolution image, and obtaining a texture part and a smooth structure part of the high-resolution image; (3) combining the low-resolution texture part obtained in the step (2) and an original low-resolution image to obtain a two-channel input, and obtaining an output of the high-resolution texture part; (4) combining the obtained output of the high-resolution texture part in the step (3) and the original low-resolution image to obtain a final image super-resolution reconstruction result, thereby finishing super-resolution reconstruction; and (5) calculating a difference value between the high-resolution texture parts obtained in the steps (4) and (2) to obtain texture part loss, and minimizing a sum of the texture loss and image loss to optimize network structure parameters.

The invention discloses a self-adaptive regularized super-resolution image reconstruction method which can keep marginal definition, which mainly solves the problem that the prior method has edge fog in reconstruction of a degraded image. The method comprises the following steps: an imaging model is constructed; on the basis of an unconstrained objective function constructed by a Lagrangian multiplier method, gradient is increased to approach a bound term; the objective function is expanded; L1 norm is adopted to measure a data approximation term; a self-adaptive bilateral total variation model which can carry out local adaptive control on the smoothing effect is utilized to construct a self-adaptive regular term; a gradient approximation term is added to be as constraint of gradient consistency; edge information is kept; the self-adaptive regular term and a gradient consistency bound term are introduced as constraint conditions; an expanded Lagrangian objective function is constructed and optimized; and an optimized unconstrained objective function is utilized to reconstruct an image, thereby obtaining a high-resolution image of which the edge is kept. The method can keep image edge clear, can inhibit noise and is suitable for restoration treatment on the degraded image.

The invention relates to a method for rate control of video coding, which includes: it allocates the surplus or overruns bit to the follow-up GOP in the controlling method for the GOP layer rate. It sets the quantitative parameters in accordance with the image complexity and pre-bit-rate allocation of I frame to make the use of resources more rationally and effectively in frame-rate controlling method. It controls the quantitative parameters of the current P frames to contain the serious downward trend of P Frame encoding quality due to lack of residual rate when GOP P frame coding rapidly decline.

The invention relates to a coding method and a decoding method of a spatial domain enhancement layer with low code rate based on an interested area. The coding method is as follows: the information of the interested area is loaded from a basic layer; original input images of an enhancement layer are cut according to the information of the interested area; and the interested area obtained after cutting is coded. The invention realizes the function of coding the spatial domain enhancement layer with low code rate based on the interested area, greatly decreases a transmitting bit number after the coding of the spatial domain enhancement layer while the quality of the video information in the spatial domain enhancement layer is ensured not to be changed, and is especially suitable for the transmitting requirement of security protection monitoring.

The invention discloses a super-resolution implementation method based on reconstruction optimization and a deep neural network. The method comprises the following specific steps of 1, establishing arelationship between a high-resolution image and a down-sampled image by means of a down-sampling method, and establishing a target function by means of a least square method; 2, utilizing a conjugategradient descent algorithm to conduct iterative optimization on the target function to obtain a high-resolution image based on a reconstruction optimization algorithm; 3, establishing and completinga three-layer full convolutional neural network through training; 4, inputting the high-resolution image obtained in the step 2 into the three-layer convolutional neural network to further improve theresolution of the image. By means of the method, good results can be achieved in both subjective evaluation and objective image quality evaluation.

The invention discloses a method for improving the video error concealment effect by utilizing a generative network. According to the method disclosed by the invention, on the basis of an error concealment technology based on the H.264 standard, a CNN structural network G is constructed; a 16*16 repaired macro block is subjected to G processing; therefore, the purpose of improving the PSNR is realized; in addition, an input image is directly mapped before an output image in the G network, and added with the network fitting residual, so that the final output is obtained; notably, the method isonly for the H.264 standard, and not suitable for other video coding standards; by means of the method disclosed by the invention, the network convergence is relatively rapid; a training result is obtained rapidly and better; a deep neural network is a high-efficiency tool in image processing; on the basis of an inter-frame concealment technology, a neural network is added; therefore, the psnr value of the repaired macro block can be increased; the quality of a whole picture is relatively high; and thus, the video error concealment effect is also better.

The invention discloses a method for embedding a robust and reversible watermark. The method includes: watermark embedding and watermark extraction. The method has the advantages that the features of a robust watermark and a reversible watermark are fully utilized, the robust watermark and the reversible watermark are combined to provide the robust and reversible watermark which can resist JPEG compression attacks, noisy points of images are filtered through low-pass filtering, robustness is solidified, image features are identified through the edge detecting technology, image partitioning is performed, the to-be-embedded block of the robust image is selected by the image texture pixel, obtained by edge detection, of each block, so that the position of embedding selected block after jpeg compression is consistent with the position of the block selected and extracted by the same manner; the watermark is embedded to increase the robustness of the watermark, and the JPEG compression attacks are resisted.

The invention discloses a method for improving frame rate up conversion effect by using a generated network. The method comprises the following steps of 1, comprising an original video via JM18.6 to acquire the compressed video, then reserving odd-number frames in the compressed video, and reconstructing even-number frames to acquire reconstructed frames of the even-number frames by using a framerate up conversion algorithm based on direction selection movement estimation; 2, building a generated network G, inputting the reconstructed frames IE into a neural network for training, comparing the corresponding frames IP in the original video with network output, backwards propagating an error, and acquire the optimal remediation effect after multiple times of adjustment; 3, repeatedly executing the steps 1 and 2 until an error variation rate is invariable or less than a set threshold; and 4, applying the trained network, inputting the reconstructed frames has been subjected to frame rateup conversion into the network to acquire an image with high quality. According to the method provided by the invention, psnr values and quality of the reconstructed frames are improved, and thus theframe rate up conversion effect is improved.

The invention discloses an image blind deblurring method based on a dark channel prior and a multi direction weighted TV(total variation), which mainly comprises steps of constructing an image blind restoration model and its solution, and finally obtaining a target image and a fuzzy kernel. The method of the invention aims at the problem that the total variation (TV) regularized image restorationhas the limited detail recovery capability and is sensitive to noise. In the method, multi-directional edge detection is utilized, the conventional TV model is improved, and a multi-direction weightedTV model based on edge detection is obtained; at the same time, in order to make the recovery model more universal and improve the detail recovery capability, the dark channel priors is fused into the above multi-direction weighted TV model. Experiments show that the target image restored by the method of the present invention has a perfect visual effect, the local smoothness of the image is maintained, the details such as the edge and texture of the image can be well restored, the ring effect is significantly reduced, and the fuzzy kernel estimation is more accurate.

The invention discloses a non-local mean filtering method based on direction field estimation. The non-local mean filtering method comprises the following steps of receiving a discrete noise fingerprint image, establishing a direction field estimation model of a pixel block in the discrete noise fingerprint image, and performing non-local mean filtering on the discrete noise fingerprint image based on the direction field estimation model to obtain a final denoising fingerprint image. According to the non-local mean filtering method, the problem of poor inhibition performance and poor robustness on noise in the conventional method is solved, contrast among grains in the fingerprint image is enhanced, and characteristic information in the fingerprint image is protected from damage.

The invention discloses a reversible information hiding method based on block prediction and difference value histogram translation. The invention belongs to the field of digital watermarking and information hiding. The reversible information hiding method comprises the following steps: firstly, dividing an image into 4 * 4 image blocks; then, conducting three neighborhood partitioning, wherein every four pixels form one group, selecting one pixel and three pixels on the periphery of the pixel, generating prediction values by using the three pixels on the periphery, comparing the prediction values with the selected pixels so as to generate difference values, enabling the difference values to be subjected to downward rounding, generating four difference values in one image block, and then generating a difference value histogram. According to the reversible information hiding method, the embedding capacity of the watermark is improved, and the algorithm steps are simplified on the premise of keeping the visual effect of the image, and the watermark can be embedded and extracted only by partitioning. According to the reversible information hiding method, reversible watermarking can beachieved, and watermarking invisibility is good, and the experiment is simple and easy to implement, and the PSNR of the image with the embedded watermarking is high, and the image visual effect is good, and the practical value is achieved.

The invention discloses an image restoration method suitable for a neutron imaging system image and based on a Gauss-Poisson hybrid noise model. According to the method, an image with Gauss-Poisson hybrid noise is transformed into a Gaussian white noise image through GAT transformation, a regional limitation BM3D algorithm is used to process the image obtained after transformation, and then unbiased GAT inverse transformation is performed on the obtained image to obtain a final de-noised image. Through the method, a Lloyd algorithm is used to perform regional division on the image, similar blocks of BM3D Stage2 are matched and limited in the same region, and the problems that through a traditional BM3D method, when Gaussian noise variance is large, object edge information is seriously lost in a filtering process, and an object edge in an image after de-noising is fuzzy are solved. Compared with a traditional BM3D algorithm, through the de-noising method, a restored image, with a better visual effect, of an image subjected to strong Gauss-Poisson hybrid noise interference can be obtained, object edge information is reserved more completely, and a higher PSNR value is obtained.

The present invention provides an efficient information hiding method of introducing auxiliary pixels. The hiding capacity is irrelevant with the types of the original data and may be the image data or other data, but the hiding quantity is always fixed, so that the efficient information hiding method has the large hiding capacity and the information safety, and is easy to realize by the hardware and software. The method realizes the joint processing of the secret information and the secret keys by introducing the auxiliary pixels and optimizing the data, and a reception terminal can completely recover the secret information and recover the original data with the high quality. The method can hide the information in each pixel of the original image data, enables the hiding efficiency to be improved, breaks through the limit of the conventional method of hiding in a pixel pair (two pixels), enables any image to reach or exceed the 1/2 hiding capacity, when the 1/2 hiding capacity is realized, enables the PSNR of a carrier image to exceed 41 dB, and is better than the typical hiding methods, for example the least significant bit (LSB) and fully exploiting modification direction (FEMD) hiding methods.

The invention discloses a partitioning compressed sensing method based on adaptive sampling and smooth projection, and relates to a partitioning compressed sensing method. The objective of the invention is to solve the problems of low quality of reconstructed images and high algorithm complexity in the prior art. The method comprises the following steps: 1, setting an original image, and calculating an adaptive sampling rate of an original image subblock; 2, performing Wiener filtering on the obtained original image sub-block with the known adaptive sampling rate to obtain a reconstructed image subjected to the Wiener filtering process; projecting the jth image sub-block in the reconstructed image to obtain an image sub-block; performing direction transformation to obtain transformed imagesub-blocks, and performing smooth projection to obtain smooth projection image sub-blocks; 3, carrying out image reconstruction according to the smooth projection image sub-blocks, judging whether the obtained reconstructed image meets the requirement for the peak signal to noise ratio of the reconstructed image or not, and if the obtained reconstructed image meets the requirement, obtaining thereconstructed image; otherwise, repeatedly executing the second step and the third step until a reconstructed image meeting requirements is obtained. The method is applied to the field of image reconstruction.

The invention discloses a self-adaptive reversible information hiding method based on integer wavelet transform, and belongs to the field of digital watermarking and information hiding. The self-adaptive reversible information hiding method comprises the following steps: firstly, dividing a carrier image into two areas, namely a watermark embedding area and an additional information embedding area; performing integer wavelet transform on the watermark embedding area, and adaptively setting a threshold by utilizing the watermark size and the mean square error of a wavelet coefficient; and determining whether a certain image block is embedded or not according to the approximate component, and then embedding the watermark information into the corresponding diagonal component, wherein the additional information embedding area is used for embedding additional information including a threshold value and the length and width of the watermark. According to the self-adaptive reversible information hiding method, self-adaptive watermark embedding can be realized, and complete blind extraction can be realized; and reversible information hiding can be achieved, and the visual effect is good, and the distortion degree is small, and the PSNR of the image embedded with the watermark is high, and the invisibility of the watermark is good, and the practical value is achieved.

The invention discloses a turbulent image de-noising method, which comprises the following steps: carrying out single-layer two-dimensional discrete wavelet transform on a noised turbulent image; extracting a high-frequency coefficient and carrying out Fast discrete Curvelet transform on the noised turbulent image; and according to Bayesian criterion estimation threshold T, improving an adaptive selection method of the threshold, obtaining an optimum threshold and obtaining a de-noised turbulent image. The turbulent image de-noising method can protect image detail information very well and suppress boundary artifacts, and thus visual effect is improved obviously; and meanwhile, the method obtains higher peak signal-to-noise ratio and smaller mean square error, and removes the noise in the turbulence-degraded images.

According to the invention, the defects in the prior art are overcome; the invention provides a temple mural disease automatic calibration and virtual restoration method. Due to the automatic diseasecalibration function, the workload of mural protection work is greatly reduced; the method has an important auxiliary effect on actual wall painting repairing work, meanwhile, a foundation is laid forvirtual repairing of the wall painting, secondary damage to the wall painting due to repeated repairing in the actual wall painting repairing process is reduced through virtual repairing of the wallpainting, and meanwhile reference is provided for the actual repairing work of the wall painting. According to the invention, the method comprises the following steps based on the obtained mural data,analyzing, calibrating and virtually repairing diseases existing in the mural; selecting or uploading an image of the fall-off disease from the cluster and submitting the image for calibration; the system automatically calling a TS-RG algorithm to perform automatic calibration, and a user capable of checking or downloading a calibration result image online for proofreading after calibration is completed; and after the user requirement being met, the user uploading the calibrated image to submit a repair task on a repair sub-module, and the system calling an ASB-LS algorithm to complete repair.