319 results about "Subjective perception" patented technology

A method and system for the estimation of the effect of packet loss on a viewer's subjective perception of a packet video stream. The method involves estimating a quality metric for every frame in the video stream, accumulating such metrics over time, and transforming the result into a Mean Opinion Score (“MOS”) which represents the viewer's subjective perception of the video stream. The method takes into account various factors such as frame rate, frame types and sizes, decoding algorithms, propagated error rates, and the effect of video content on viewer perception. The method can be used for both encrypted and unencrypted video streams.

The invention discloses a stereo image quality objective evaluation method based on visual perception. Firstly, a stereo image is divided into a strong edge block, a weak edge block , a flat block and a texture block, and characteristic information reflecting image quality and depth perception of different zone blocks is extracted through computation so as to obtain stereo image characteristic vectors; and then the characteristic vectors of distorted stereo images of the same distortion type in a distorted stereo image set are trained through support vector regression, and each distorted stereo image of the same distortion type is tested by a support vector regression training model to obtain the objective image quality evaluation forecast value of each distorted stereo image. The method has the advantages that the obtained characteristic vector information reflecting the image quality and the depth perception has stronger stability and can better reflect the quality change condition of the stereo images, and the relevance of an objective evaluation result and subjective perception is improved.

The invention discloses a multi-media playing device dynamically varying an outer shell color along with rhythm. The device comprises an outer shell, a light-emitting unit arranged on the outer shell, a control unit used for controlling the light-emitting unit and an audio decoding unit which decodes audio signals played by the multi-media playing device in real time, quantizing the obtained decoded signals and converting the signals into control signals, and the control unit controls the light-emitting mode of the light-emitting unit according to the control signals. A music dynamic effect is expressed on lamp light outside the device, and the multi-media playing device has the advantages that visual effects of all kinds of audios are expanded to the outer shell of the device to be achieved, so that the visualization effect is enhanced; the lamp light outside the device is changed to be in a dynamic light-emitting mode from a static light-emitting mode, and visual perception which is brought to a user by the colorful lamp light is sufficiently enhanced; the shade and color of the lamp light outside the device are combined with the played audios, so that the expressive force of the device is enhanced, and stronger subjective perception is brought to the user.

The invention discloses a non-reference image quality objective evaluation method based on deep learning. Multi-resolution pyramid and Gaussian difference decomposition is performed on distortion images to be evaluated and then natural statistical characteristics can be extracted by performing simple local normalization on sub-band images without extracting characteristics from a transform domain so that complexity is greatly reduced. Degree of distortion of the images is measured by degree of loss of the natural statistical characteristics with no requirement for reference images or distortion types. The change condition of visual quality of the images under the influence of various image processing and compression methods can be objectively reflected by the method, and evaluation performance of the method is not influenced by the image content or the distortion types and is consistent with subjective perception of human eyes. An existing L moment estimation method is adopted to estimate the distribution parameters of the envelope curve of a gray level histogram, and the distribution parameters obtained through estimation are more accurate and have higher generalization capability.

The invention discloses a method for objectively evaluating image quality. The method comprises the following steps of: dividing a reference image and a distorted image into an edge region, a texture region and a planar region in a combined way according to gradient amplitude information of each pixel point in the reference image and the distorted image; performing phase quantization coding on the reference image and the distorted image according to a gradient phase and a gradient amplitude of each pixel point in each region; calculating hamming distances for the phase quantization coding of different regions of the reference image and the distorted image; and giving different weights to the hamming distances of the different regions according to a human visual system so as to acquire a final object score of image quality evaluation. In the method, since the influence of variation of the distorted image relative to the gradient amplitude and the gradient phase of the reference image on the quality of the distorted image is comprehensively considered, an objective evaluation result obtained by the method has high consistency with a human eye subjective perception result; in addition, the method has low calculation complexity and high running speed.

The invention discloses a method for evaluating a stereo image vision comfort level based on machine learning. According to the method, the important vision area mask of stereo images is extracted through the saliency map of a right viewpoint image and a right parallax image, a characteristic vector used for reflecting parallax range characteristics and parallax gradient characteristics and a characteristic vector used for reflecting space frequency characteristics are extracted through the important vision area mask, the characteristic vectors of the stereo images are obtained, training is carried out on all the stereo images in a stereo image set through support vector regression, and finally each stereo image in the stereo image set is tested through a support vector regression training model obtained in training to obtain the vision comfort level evaluation predicted value of each stereo image. The method has the advantages that the obtained characteristic vector information of the stereo images has strong stability and can reflect the vision comfort level changing conditions of the stereo images well, and therefore the correlation of objective evaluation conditions and subjective perception is effectively improved.

The invention discloses a no-reference image quality evaluation method based on a fully convolutional neural network. A full reference image quality evaluation method is adopted to obtain the objective and real quality map of each distorted image in a training set as supervision to train the normalized image of all distorted images in the training set, and an optimal fully convolutional neural network regression training model is obtained; the normalized image of the distorted image to be evaluated is input into the optimal fully convolutional neural network regression training model; the objective quality evaluation prediction quality map of the distorted image to be evaluated is predicted; and the saliency map of the distorted image to be evaluated is used for carrying out weighted pooling on the objective quality evaluation prediction quality map so as to obtain an objective quality evaluation prediction value. Since various features, including full reference features, saliency features and the like, of the distorted image are combined and the features can accurately describe the distorted image, dependency between the objective evaluation result and subjective perception is effectively improved.

The invention, which relates to the digital image processing field, provides a visual-saliency-based random image region quality evaluation strategy, so that an evaluation result meets the subjective perception of the human being. According to the technical scheme, the method comprises: (1), Gaussian filtering is carried out on a reference image IR and a distorted image ID respectively and the filtered image RGBs are converted into Lab color space units; (2), for different image sizes r*c, N windows with random sizes are generated randomly; (3), visual saliency calculation is carried out the reference image and the distorted image respectively by using a visual-saliency extraction algorithm and a visual saliency similarity matrix in each random window is obtained; (4), weighting and integration are carried out on the obtained similarity matrixes of all random windows. The method is mainly applied to an image processing occasion.

The invention discloses a stereo image objective quality evaluation method. The method includes the steps that Riesz transformation is carried out on a left viewpoint image and a right viewpoint image of an undistorted stereo image and a left viewpoint image and a right viewpoint image of a distorted stereo image; a left and right viewpoint feature fusion image of the undistorted stereo image and a left and right viewpoint feature fusion image of the distorted stereo image are obtained by using the obtained respective amplitude images and phase images and parallax information; according to the two left and right viewpoint feature fusion images, an image quality objective evaluation value of the distorted stereo image is obtained, and as the characteristic of left and right viewpoint feature fusion is fully used, the correlation of objective evaluation results and subjective perceptions is effectively improved.

The invention discloses a no-reference image quality evaluation method based on gradient information. Through deep digging of perception characteristics of human vision for image structure, gradient filtering is performed on a distorted image, so as to obtain an amplitude image and a phase image of the gradient information; a local binary pattern operation is performed on the amplitude image and the phase image so as to obtain a local binary pattern characteristic image of the amplitude image and a local binary pattern characteristic image of the phase image; then condition probability characteristics of all pixel points of different pixel values in the amplitude image and the phase image are gained; and finally, according to the condition probability characteristics, an objective quality evaluation prediction value of the distorted image to be evaluated is predicted by use of support vector regression. The method has the advantages that the impact of gradient structure change on visual quality is fully considered, so that the obtained objective quality evaluation prediction value can accurately reflect the subjective perception quality of human vision, and the correlation between an objective evaluation result and subjective perception can be improved effectively.

The invention discloses an eye fundus image no-reference quality evaluation method. The method comprises two processes including a training stage and a test stage. The influence of brightness, naturalness and structure layout on eye fundus image quality is considered. A dark channel proportion feature, a bright channel proportion feature, a non-uniform brightness feature, a naturalness quality evaluation score and a structure layout index are extracted to form eigenvectors; and then the eigenvectors of all eye fundus images in a training image set are trained by utilizing support vector regression, and a quality prediction model is built. In the test stage, the eigenvectors of the eye fundus images used for test are calculated, and according to the quality prediction model built in the training stage, prediction is performed to obtain quality objective evaluation prediction values of the eye fundus images. The obtained eigenvector information can better reflect the quality change conditions of the eye fundus images, so that the correlation between objective evaluation results and subjective perception is effectively improved.

The invention relates to an encoder adaptive adjustment method based on subjective quality evaluation. The method comprises the following steps: step A, through analyzing influences exerted by encoding parameters and configuration conditions on a video stream code rate, establishing an HEVC encoder parameter model; step B, by use of subjective perception of a terminal user for received video quality, establishing a subjective QoE model integrally taking encoder arrangement and network loss factors into consideration; and step C, based on the established subjective QoE model, for video service of multiple users, constructing an adaptive adjustment algorithm maximizing total QoE of all the users under the condition of a limited channel capacity. Compared to the prior art, the method provided by the invention substantially improves user total satisfaction under the condition of multiple users, at the same time, supports resource distribution according to user priorities, and has great significance on improvement of system performance and optimization of network resource distribution.

The invention discloses a non-reference image objective quality evaluation method. The non-reference image objective quality evaluation method comprises the following steps: respectively applying Gaussian smoothing gradient filter and laplace operator Gaussian filter to a distorted image by deeply excavating perception characteristic of human vision to image structure to correspondingly obtain a Gaussian smoothing gradient filter image and a laplace operator Gaussian filter image, then respectively carrying out local binaryzation mode operation on the two filter images to obtain respective local binaryzation mode feature images, then calculating respective marginal probability feature and conditional probability feature of the two local binaryzation mode feature images, and finally forecasting on objective quality evaluation predicted value of the to-be-evaluated distorted images by adopting support vector regression according to the marginal probability feature and conditional probability feature. The obtained objective quality evaluation predicted value can accurately reflect subjective perceived quality of human vision and can be used for effectively improving correlation between objective evaluation result and subjective perception.

The invention discloses an image quality evaluation method based on a combination neural network and a classification neural network. In the training stage, an objective reality quality image of a distortion image obtained by adopting a full-reference image quality evaluation method is adopted as supervision, a normalized image of the distortion image is trained to obtain a combination neural network regression training model for different distortion types; a classification label of the distortion image is adopted as supervision, and the normalized image of the distortion image is trained to obtain a classification neural network training model; in the testing stage, the normalized image of the distortion image to be evaluated is input into the classification neural network training model,and a distortion type is obtained; according to the distortion type, the normalized image is input into the corresponding combination neural network regression training model to obtain an objective quality evaluation prediction quality map, and adopting a saliency map for performing weighing pooling on the objective quality evaluation prediction quality map, and obtaining an objective quality evaluation prediction value. The method has the advantage that the correlation between the objective evaluation result and subjective perception is effectively improved.

The present invention relates to a no-reference quality evaluation method of compression perception recovery images. The method comprises the steps of aiming at the distortion characteristics of the compression perception recovery images, extracting three local characteristics and five global characteristics of the images to evaluate the quality of the image, wherein the local characteristics comprise the image definition, the contrast and the homogeneity, and the global characteristics comprise four characteristics about the naturality of the images extracted by utilizing a generalized Gaussian model to fit the MSCN coefficients and one characteristic obtained by utilizing the singular value decomposition; and finally utilizing a support vector machine to train the above 8 characteristics to obtain a characteristic model, thereby applying the model to predict the quality scores of the images. The evaluation results of the method are better than the results of an international mainstream method, and have the very high consistency with the human eye subjective perception.

The invention relates to a texture feature-based image quality evaluating method without reference. The texture feature-based image quality evaluating method includes the steps of conducting three-scale wavelet transform decomposition for an image I to obtain nine high-frequency detailed subgraphs, calculating a local binary atlas of the image I by using a local binary mode, calculating the gray co-occurrence matrix of the sub-graphs and the local binary atlas respectively to obtain ten Gh, calculating the four feature parameters of the Gh, namely contrast, correlation, energy and homogeneity, and taking the average of the feature parameters of the horizontal detailed wavelet subgraphs and vertical detailed wavelet subgraphs under same wavelet transform decomposition scale to obtain a 24-dimensional eigenvector, and conducting network training to obtain a prediction model. The texture feature-based image quality evaluating method can effectively improve the correlation between the objective evaluation result and the subjective perception.

The invention provides a quantized optimization method used for high-efficiency video coding (HEVC). Through performing visual perception threshold calculation on each transform unit (TU), a perception threshold of each transformation coefficient is obtained. In a quantization period, based on the thresholds, candidate quantization values are screened for each transformation coefficient and quantized values are restrained. At last, in a TU division mode selection period, visual perception information of a transform block is used to perform pre-judgment on the division mode of the TUs. The quantized optimization method used for the HEVC does not need to make any change on a decoder and a coding compression efficiency is improved and the coding speed is increased on the premise that a video subjective perception quality is kept to be unchanged.