KR20200110255A - Method and apparatus for assessing feature of image - Google Patents

Abstract

Disclosed are a method for measuring a feature of an image as a score, and an apparatus therefor. The measurement method comprises the steps of: calculating and extracting a plurality of pieces of feature information from a reference image and a comparison image; calculating distortion values between corresponding pieces of feature information from the extracted plurality of pieces of feature information; extracting a plurality of prediction models of correlations between the distortion values and subjective image quality and/or naturalness measurement values by using the calculated distortion values as an input; constructing a system by using the plurality of prediction models, and determining weights for the plurality of prediction models, respectively; and predicting the subjective image quality/neutrality measurement values by using the plurality of prediction models to which the weights are assigned. According to the present invention, accuracy of an image quality indicator can be improved.

Description

Method and apparatus for measuring the characteristics of an image {METHOD AND APPARATUS FOR ASSESSING FEATURE OF IMAGE}

The following embodiments relate to a method and apparatus for processing information on an image, and in more detail, a method and apparatus for measuring a characteristic of an image are disclosed.

In the field of image processing or image/video compression, an index for measuring image quality is required for verification of the performance of a developed technology or optimization of a developed technology. The better this indicator reflects the subjective quality performance, the more the developed technology can provide superior subjective quality performance.

Recently, in various image processing using the generated images, which are the output of Deep Neural Network (DNN), in addition to measuring the quality of the comparative image, an index for measuring naturalness has been developed and developed. It can be used for technology that has been developed.

The measurement of subjective quality/naturality of an image is a method of statistically processing individual evaluation values obtained through a subjective evaluation experiment involving a plurality of evaluators to obtain a measurement value of image quality and/or naturalness. This subjective measurement of quality/naturality is considered to be the most accurate method in measuring the quality and/or nature of an image. However, such subjective measurement of image quality/naturality has a disadvantage in that it requires high temporal and/or economical cost.

In this case, in the subjective quality/naturality measurement, each evaluator evaluates the quality/naturality of the reference image and the comparison image as values, and a quality/naturality measurement value that is an average value of differences between the evaluated values may be obtained. Here, the quality/naturality measurement value may be a Mean Opinion Score (MOS) or a Difference Mean Opinion Score (DMOS).

The purpose of the development of the automatic image quality/natural level measurement index is to measure the quality/natural level using relatively low economic costs, and to replace subjective evaluation.

In order to achieve this purpose, the automatic image quality/naturality measurement method must provide high measurement reliability. Typically, measurement reliability is evaluated through Pearson's Correlation Coefficient (PCC) or Spearman's Rank Correlation Coefficient (SRCC).

An embodiment may provide an index for automatically measuring the perceived quality of a natural image or a natural degree of a DNN-generated image.

In one embodiment, in order to provide high measurement reliability, a comparison image based on various feature information and a multi-stage machine learning system structure for measuring structural/statistical similarity between an original image and a comparison image It can provide a method for measuring the quality/naturality of

In one aspect, calculating and extracting a plurality of feature information from a reference image and a comparison image; Calculating distortion values between corresponding feature information from the extracted plurality of feature information; Extracting a plurality of predictive models of correlations between the distortion values and subjective image quality and/or naturalness measurement values by using the calculated distortion values as inputs; And configuring a system using the plurality of prediction models, and determining weights for each of the plurality of prediction models.

In addition to this, another method, apparatus, and system for implementing the present invention, and a computer-readable recording medium for recording a computer program for executing the method are further provided.

Various feature information can be used for prediction of image quality/naturality.

The accuracy of the quality index may be improved by measuring the accuracy and structural difference of the comparison image in pixel units compared to the original image by using structural similarity feature information among the feature information.

Image quality degradation that is not measured by structural similarity may be quantified by using statistical similarity feature information among the feature information. In addition, in determining the naturalness of the DNN-generated image, prediction accuracy may be improved compared to a conventional quality index.

By configuring a prediction model system using a multi-stage machine learning system, the accuracy of prediction can be improved by adaptively allocating weights for structural similarity and statistical similarity according to characteristics of an input image.

1 is a structural diagram of a measuring device according to an exemplary embodiment.
2 is a flowchart of a measurement method according to an embodiment.
3 illustrates use of feature information of a subband image generated using SVD transformation according to an example.
4 illustrates frame-based feature calculation according to an example.
5 illustrates block-based feature calculation according to an example.
6 illustrates generation of subjective quality/naturality values using a linear or nonlinear model according to an example.
7 shows two-stage prediction models according to an example.
8 shows n-stage prediction models according to an example.
9 illustrates a structure of an electronic device implementing a measuring device according to an exemplary embodiment.

For a detailed description of exemplary embodiments described below, reference is made to the accompanying drawings, which illustrate specific embodiments as examples. These embodiments are described in detail sufficient to enable a person skilled in the art to practice the embodiments. It should be understood that the various embodiments are different from each other but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the present invention in relation to one embodiment. In addition, it is to be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the embodiment. Accordingly, the detailed description to be described below is not intended to be taken in a limiting sense, and the scope of exemplary embodiments, if properly described, is limited only by the appended claims, along with all scope equivalents to those claimed by the claims.

Like reference numerals in the drawings refer to the same or similar functions over several aspects. The shapes and sizes of elements in the drawings may be exaggerated for clearer explanation.

The terms used in the examples are for describing the examples and are not intended to limit the present invention. In embodiments, the singular also includes the plural unless specifically stated in the text. As used in the specification, "comprises" and/or "comprising" refers to the presence of one or more other components, steps, actions and/or elements, and/or elements, steps, actions and/or elements mentioned. Or, it does not exclude addition, it means that the additional configuration may be included in the scope of the technical idea of the exemplary embodiments or implementation of the exemplary embodiments. When a component is referred to as being "connected" or "connected" to another component, the two components may be directly connected to each other or may be connected, but the above 2 It should be understood that other components may exist in the middle of the components.

Terms such as first and second may be used to describe various components, but the above components should not be limited by the above terms. The above terms are used to distinguish one component from another component. For example, without departing from the scope of the rights, a first component may be referred to as a second component, and similarly, a second component may be referred to as a first component.

In addition, components shown in the embodiments are shown independently to represent different characteristic functions, and it does not mean that each component is composed of only separate hardware or one software component unit. That is, each component is listed as each component for convenience of description. For example, at least two of the components may be combined into one component. Also, one component may be divided into a plurality of components. An integrated embodiment and a separate embodiment of each of these components are also included in the scope of the rights unless departing from the essence.

In addition, some of the components are not essential components that perform essential functions, but may be optional components only for improving performance. The embodiments may be implemented including only components essential to implement the essence of the embodiments, and structures excluding optional components, such as components used only for improving performance, are also included in the scope of the rights.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings in order to enable those of ordinary skill in the art to easily implement the embodiments. In describing the embodiments, when it is determined that a detailed description of a related known configuration or function may obscure the subject matter of the present specification, a detailed description thereof will be omitted.

In the description of the specification, the symbol "/" may be used as an abbreviation of "and/or". That is, "A/B" may mean "A and/or B" or "at least one of A and B". have.

Full reference (FR) quality such as conventional mean squared error (MSE), peak signal-to-noise ratio (PSNR), or structural similarity SSIM (Structural SIMilarity; SSIM) The indicators are a method of measuring a difference between the reference image and the comparison image, and may have an advantage of being intuitive.

On the other hand, these FR quality indicators may have a large difference in the trend of the measured value compared to the perceived quality, and may have low measurement reliability in comparison between images of different characteristics.

The method of measuring these FR quality indicators is designed with a focus on maintaining fidelity in the pixel unit of the image, so it can show low accuracy for structural or statistical distortion of the image. In the measurement of the natural degree of, it can show a lower accuracy.

In the following embodiments, a method of measuring the quality or naturalness of an image is described, and an index for measuring the quality or naturalness of a comparison image is described as compared to a reference image.

1 is a structural diagram of a measuring device according to an exemplary embodiment.

A reference image and a comparison image may be input to the measurement device 100.

The reference image may be an original input image. The comparison image may be a comparison input image or a test image.

The measuring apparatus 100 may calculate a predicted quality/natural degree measurement value of the comparison image compared to the reference image by comparing the reference image and a plurality of extracted information values extracted from the comparison image. Here, the measured value may mean a score. Picture quality can mean perceptual quality.

Here, the plurality of extracted information values may be feature information.

For example, the measuring apparatus 100 may utilize feature information related to various Singular Value Decomposition (SVD) to measure structural similarity between the reference image and the comparison image. Here, the feature information related to singular value decomposition may form one feature information group.

For example, the measuring apparatus 100 may utilize feature information related to a histogram distance in order to measure statistical similarity between the reference image and the comparison image. Here, feature information related to a distance between histogram distributions may form one feature information group.

For example, the measurement apparatus 100 may combine two feature information group groups, and calculate a final quality/natural degree measurement value using the combined two feature information group groups. The two feature information group groups may mean a feature information group group generated by feature information related to singular value decomposition and a feature information group group generated by feature information related to a distance between the histogram distribution.

The measuring apparatus 100 may use a multi-stage machine learning system to calculate a final image quality/natural degree measurement value.

The measuring apparatus 100 for providing a model for measuring image quality/natural degree includes a feature information extraction unit 110, a distortion calculation unit 120, a prediction model generation unit 130, a model optimization unit 140, and a prediction unit ( 150) may be included.

The feature information extraction unit 110 may calculate and extract a plurality of feature information from the reference image and the comparison image.

The distortion calculator 120 may compare and calculate difference values or distortion values between corresponding feature information from the plurality of feature information extracted from the original image and the comparison image.

The prediction model generator 130 may extract a plurality of prediction models of correlations between distortion values and subjective quality/naturality measurement values by using a plurality of feature information and/or calculated distortion values as inputs.

The prediction model generation unit 130 may learn a prediction model between a distortion value of each feature information and a subjective quality/natural degree measurement value for each feature information of a plurality of feature information.

The model optimizer 140 may configure a system using a plurality of prediction models according to learning, and may determine weights for each of the plurality of prediction models.

The predictor 140 may predict a subjective quality/naturality measurement value using a plurality of prediction models to which weights are assigned.

2 is a flowchart of a measurement method according to an embodiment.

In step 210, the feature information extractor 110 may calculate and extract a plurality of feature information from the reference image and the comparison image.

In step 220, the distortion calculator 120 may compare and calculate difference values or distortion values between corresponding feature information from the plurality of feature information extracted from the reference image and the comparison image.

In step 230, the prediction model generation unit 130 uses a plurality of feature information and/or calculated distortion values as inputs, and a plurality of prediction models of correlations between distortion values and subjective quality/naturality measurement values Can be extracted.

The prediction model generation unit 130 may learn a prediction model between a distortion value of each feature information and a subjective quality/natural degree measurement value for each feature information of a plurality of feature information.

In step 240, the model optimizer 140 may configure a system using a plurality of prediction models according to training, and determine weights for the plurality of prediction models, respectively.

In step 250, the prediction unit 140 may predict a subjective quality/naturality measurement value using a plurality of prediction models to which weights are assigned.

The measured value can be a score.

In the following, the operation of the above-described steps will be described in more detail.

Extraction of feature information

Refer back to FIG. 2.

In step 210, the feature information extraction unit 110 may calculate and extract a plurality of feature information from the reference image and/or the comparison image.

The feature information extraction unit 110 may extract one or more feature information from the reference image and/or the comparison image as described below in order to measure the quality and/or the natural degree of the comparison image.

Hereinafter, the image may be a reference image or a comparison image.

The image size or resolution may be WxH. W may mean the width of an image, that is, the number of pixels on the horizontal axis. H may mean the height of the image, that is, the number of pixels on the vertical axis.

For example, the feature information extracting unit 110 may use a pixel value (ie, intensity) of an image as feature information of an image. In other words, the feature information of the image may include a pixel value of the image.

For example, the feature information extraction unit 110 may use a coefficient value generated by performing a 2D transformation on an image as feature information of an image. In other words, the feature information of the image may include a coefficient value of the image.

The second-order transform is Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Discrete Fourier Transform (DFT), Discrete Wavelet Transform (DWT) and singular values. It may be one of Singular Value Decomposition (SVD) and the like. In addition, other conversion equations may be used for the second-order conversion.

3 illustrates use of feature information of a subband image generated using SVD transformation according to an example.

The feature information extractor 110 may generate sub-band images by dividing the image into N sub-bands, and may use the sub-band images as feature information. N may be a positive integer of 2 or more.

Segmentation can be applied to an image before (or to which no transformation has been applied). Alternatively, the segmentation may be applied to a transformed image generated by transforming the original image using one of the aforementioned transforms. Transformation may be one of the quadratic transformation methods.

In FIG. 3, an example of generating a sub-band image using SVD transformation and using the sub-band image as feature information is illustrated.

SVD transformation for an image X having a size of r × c can be defined as in Equation 1 below.

[Equation 1]

U may be a left singular matrix of size r × r .

u _i ( i = 1, 2, ..., r ) is a column vector constituting U , and may be an eigen vector of X * X ^T.

V may be a right singular matrix of size c × c .

v _i (i = 1, 2, ..., c ) is a column vector constituting V , and may be an eigen vector of X ^T * X.

σ may be a diagonal matrix composed of singular values.

t may be a smaller value of r and c ( t = min{ r , c }).

The ensemble image or EsbImg can be defined as in Equation 2 below. Here, it can be k < t .

[Equation 2]

Also, an eigen image may be defined as in Equation 3 below. Here, it can be k < t .

[Equation 3]

In FIG. 3, an example of dividing a reference image and/or a comparison image into three sub-band images using an ensemble image is illustrated.

In classification, the subbands may be a low band (LB), a mid band (MB), and a high band (HB).

The number of u _i and/or v _i for generating each sub-band image may be determined by one of the following methods. However, the sum t of the number of u _i and/or v _i of all subbands may be a smaller value among r and c .

-The number of u _i and/or v _i may be floor ( t / N ). Here, floor( x ) may be the largest integer less than or equal to x . For example, in FIG. 3, N may be 3,

-The number of u _i and/or v _i is floor{(1/6)* t } for the LB image, floor{(2/6)* t } for the MB image, and floor{ for the HB image It can be (3/6)* t }.

- u _i and / or v _i number is, floor {(w 1 * t } and, floor {(w 2 * t } for MB image on LB image and, HB for the image floor {w 3 * t of } In this case, w 1 + w 2 + w 3 = 1.

Refer back to FIG. 2.

The feature information extraction unit 110 may use a histogram generated from an image as feature information. In other words, the feature information of the image may include a histogram.

The histogram may be generated from an image before (or to which no transformation has been applied). Alternatively, the histogram may be generated from a transformed image generated by transforming the original image using one of the aforementioned transforms. Transformation may be one of the quadratic transformation methods.

The histogram may be generated based on one of the statistical indicators for the image. At this time, as the statistical index, a mean, a maximum value (max), a minimum value (min), a median value, and other statistical values for the image may be used.

The feature information extracting unit 110 may generate a histogram using a row value of an image. The row value Rho may be defined as in Equation 4 below.

[Equation 4]

The statistical index of an image used to generate the histogram may be generated by dividing the image into a plurality of unit blocks having a size of N × M and then using statistical values calculated for all the unit blocks. N and M may be positive integers in which N < H and M < W .

Calculation of distortion between feature information

In step 220, the distortion calculator 120 may compare and calculate difference values or distortion values between corresponding feature information from the plurality of feature information extracted from the reference image and the comparison image.

The distortion calculator 120 may extract feature information from the reference image. The distortion calculator 120 may extract feature information from the comparison image.

Feature information extracted from the reference image and feature information extracted from the comparison image may correspond to each other. In other words, a specific type of characteristic information among the characteristic information extracted from the reference image and the specific type of characteristic information among the characteristic information extracted from the comparison image may correspond to each other. Corresponding feature information of the same type may be referred to as the same feature information in the feature information.

The distortion calculation unit 120 may extract feature information from the reference image and the comparison image, and then calculate distortion between the same feature information in the extracted feature information. Distortion may be an index indicating an opposite tendency of similarity between values of the same feature information.

After extracting the feature information, the distortion calculator 120 may calculate distortion between the same feature information of the reference image and the comparison image as shown in Equation 5 below.

[Equation 5]

R may represent a reference image. T may represent a comparison image.

The function F ( R , T ) may be one of a pixel-based difference, an absolute difference, and an N-th square of the difference (N is an integer greater than or equal to 2).

When N subband images are extracted as feature information from the reference image and the comparison image through the division in step 210 described above, the distortion value is calculated by one or more of the methods of Equations 6 to 10 below. Can be.

[Equation 6]

-EsbRef _B and EsbTest _B may represent the ensemble image EsbImg of the B- th subband of the reference image and the comparison image, respectively.

[Equation 7]

-EigRef _B and EigTest _B may represent eigen images of the B- th subband of the reference image and the comparison image, respectively.

[Equation 8]

[Equation 9]

-XvecRef _B and XvecTest _B may be singular vector matrices of the B- th subband of the reference image and the comparison image, respectively. X can be U or V.

-U _B and V _B may represent the number of singular vectors included in the singular matrix of the corresponding subband.

-Operator ○ can represent the dot product of a matrix. Operator Θ may represent an element-wise difference of a matrix. The function abs (.) can mean absolute difference.

[Equation 10]

-DiagRef _B and DiagTest _B may be a diagonal matrix of the B- th subband of the reference image and the comparison image, respectively, and may represent the number of singular values in N _B and subband B.

In step 210, when each histogram is generated as feature information from the reference image and the comparison image, the distortion calculator 120 uses the distribution distance of the two histograms to calculate the distortion between the two histograms. I can.

In the calculation of the distribution distance, a Kullback Leibler (KL) distance of Equation 11 below or a Chi-square distance of Equation 12 may be used.

[Equation 11]

n can represent the number of bins.

P (i) may represent an i- th bin value on the first probability distance.

Q (i) may represent an i- th bin value on the second probability distance.

[Equation 12]

n can represent the number of bins.

x _i may represent an i- th bin value of the first histogram.

y _i may represent an i- th bin value of the second histogram.

[Equation 13]

Hist _int ^Ref and Hist _int ^Test may represent histograms in an intensity domain before (or to which transformation is not applied) the reference image and the comparison image, respectively.

KL can use the KL distance between two histograms as a distortion value.

[Equation 14]

Hist _freq ^Ref and Hist _freq ^Test may represent histograms in a transformed domain of the transformed reference image and the transformed comparison image after the reference image and the comparison image are transformed, respectively.

KL can use the KL distance between two histograms as a distortion value.

In Equations 13 and 14 described above, in addition to the KL distance, a Chi-square distance and other distribution distance indicators may be used to calculate the distortion value.

[Equation 15]

Hist _int ^EsbRefB and Hist _int ^EsbTestB may represent histograms of B- th subband ensemble images in the intensity domain before (or to which transformation is not applied) the reference image and the comparison image, respectively.

KL can use the KL distance between two histograms as a distortion value.

[Equation 16]

Hist _freq ^EsbRefB and Hist _freq ^EsbTestB may represent histograms of B- th subband ensemble images in a transformed domain of the transformed reference image and the transformed comparison image after the base image and the comparison image are transformed, respectively. .

KL can use the KL distance between two histograms as a distortion value.

4 illustrates frame-based feature calculation according to an example.

5 illustrates block-based feature calculation according to an example.

In step 220, the distortion calculator 120 may use the total size of the image or the unit block size as a unit of calculation in calculating the distortion between the same feature information of the original image and the comparison image. In other words, in calculating the distortion between the same feature information of the original image and the comparison image, the distortion calculation unit 120 may calculate the distortion between the same feature information for the entire image or a block of the image.

As shown in FIG. 4, the distortion calculator 120 may calculate distortion between the same feature information for the entire image.

As illustrated in FIG. 5, the distortion calculator 120 may calculate distortion between feature information for each of the unit blocks of an image.

When the size of the image is H × W , the size of the unit block may be N × M. N < H and M < W , and N , H , M and W may be positive integers.

The distortion calculator 120 may divide the entire image into a plurality of unit blocks, and may calculate a distortion value for each unit block of the plurality of divided unit blocks.

The distortion calculator 120 may calculate distortion values for all unit blocks, and use statistical values of distortion values of a plurality of unit blocks as a final distortion value (for an image).

The statistical value may be calculated using one or more of an average, a maximum value, a minimum value, and a median value.

In calculating the distortion values for the unit blocks, the distortion calculator 120 may calculate the distortion value for every M pixels on the horizontal axis and every N pixels on the vertical axis of the image. M and N can be positive integers.

Learning about the predictive model between the distortion value of each feature information and the measured value of subjective quality/naturality

Referring back to FIG. 2, distortion values may be calculated for all feature information according to methods as described above.

Through the subjective evaluation experiment, a subjective evaluation (Mean Opinion Score; MOS) value for the quality/naturality of all comparative images can be obtained.

6 illustrates generation of subjective quality/naturality values using a linear or nonlinear model according to an example.

As shown in FIG. 6, in step 230, the prediction model generator 130 may generate N distortion values from one feature information, and a linear or nonlinear relationship model between N distortion values and MOS values Can be created using one or more of the following methods. N can be a positive integer.

The prediction model generator 130 may generate a relationship model by determining weight values of a plurality of distortion values as shown in Equation 17 below.

[Equation 17]

Here, w ₁ + w ₂ + ... + w _N May be = 1.

The predictive model generator 130 may generate a nonlinear relationship between distortion values and MOS values through a machine learning technique.

Machine learning techniques include Support Vector Regression (SVR), random forest, gradient descent, naive bayes classifier, hidden markov model, and One or more of the K-nearest neighbors may be used.

The prediction model generator 130 may generate a nonlinear relationship between distortion values and MOS values through a deep neural network or deep learning technique.

Configuration of a system using multiple prediction models and determination of weights of prediction models

7 shows two-stage prediction models according to an example.

After the relationship models between the distortion value and the MOS representing the image quality/naturality for all feature information are secured, the model optimizer 140 builds a multi-prediction system that calculates the final image quality/naturality prediction value. I can.

For the reference image and/or the comparison image, when N pieces of feature information are used and the distortion values that can be calculated from each feature information are M _N pieces, the model optimizer 140 uses one or more of the following methods. Can be used to construct multiple prediction systems.

The model optimization unit 140 may configure a plurality of prediction models in two stages, as illustrated in FIG. 7.

The final prediction model of the second stage receives as input the predicted quality/naturality scores from all the prediction models in the first stage, and uses the predicted quality/naturality scores from all the prediction models in the first stage to be final. It may be a linear or nonlinear model that outputs a predicted score for the quality/naturality of.

8 shows n-stage prediction models according to an example.

As illustrated in FIG. 8, the model optimization unit 140 may configure a plurality of prediction models in three stages.

The K intermediate prediction models of the second stage receive as inputs the predicted quality/naturality scores from the L ₁ , L ₂ , ..., L _K prediction models of the first stage, and the prediction of the first stage It may be a linear or nonlinear model that outputs an intermediate picture quality/natural degree prediction score using the picture quality/natural degree scores predicted from the models. It can be L ₁ + L ₁ + ... + L ₁ = N.

The final prediction model of the third stage receives as input the intermediate quality/naturality scores predicted from all the prediction models of the second stage, and the intermediate quality/naturality scores predicted from the prediction models of the second stage. It may be a linear or nonlinear model that outputs the final picture quality/naturality prediction score by using.

In a method similar to the method described with reference to FIG. 8, a plurality of prediction models may be configured in N stages.

The method of measuring quality/naturality of the comparison image described in the embodiments may be extended to a video. The quality/naturality score of the video may be determined using one or more of the statistic values of quality/naturality scores of all images constituting the video. The statistical value may be one or more of a mean (mean), a maximum value (max), a minimum value (min), and a median value.

9 illustrates a structure of an electronic device implementing a measuring device according to an exemplary embodiment.

The measuring device 100 may be implemented as the electronic device 900 illustrated in FIG. 8. The electronic device 900 may be a general-purpose computer system operating as the measuring device 100.

As shown in FIG. 9, the electronic device 900 may include at least some of a processing unit 910, a communication unit 920, a memory 930, a storage 940, and a bus 990. Components of the electronic device 900 such as the processing unit 910, the communication unit 920, the memory 930, and the storage 940 may communicate with each other through the bus 990.

The processing unit 910 may be a semiconductor device that executes processing instructions stored in the memory 930 or the storage 940. For example, the processing unit 910 may be at least one hardware processor.

The processing unit 910 may process a task required for the operation of the electronic device 900. The processing unit 910 may execute a code of an operation or step of the processing unit 910 described in the embodiments.

The processing unit 910 may generate, store, and output information to be described in embodiments to be described later, and may perform operations of steps performed in other electronic devices 900.

The communication unit 920 may be connected to the network 999. Data or information required for the operation of the electronic device 900 may be received, and data or information required for the operation of the electronic device 900 may be transmitted. The communication unit 920 may transmit data to another device through the network 999 and may receive data from the other device. For example, the communication unit 920 may be a network chip or a port.

The memory 930 and the storage 940 may be various types of volatile or nonvolatile storage media. For example, the memory 930 may include at least one of a ROM 931 and a RAM 932. The storage 940 may include an internal storage medium such as RAM, a flash memory, and a hard disk, and may include a removable storage medium such as a memory card.

A function or operation of the electronic device 900 may be performed as the processing unit 910 executes at least one program module. The memory 930 and/or the storage 940 may store at least one program module. At least one program module may be configured to be executed by the processing unit 910.

At least some of the above-described measurement devices 100 may be at least one program module.

Program modules may be included in the electronic device 900 in the form of an operating system, an application module, a library, and other program modules, and may be physically stored on various known storage devices. Also, at least some of these program modules may be stored in a remote storage device capable of communicating with the electronic device 900. Meanwhile, these program modules are routines, subroutines, programs, objects that perform specific operations or specific tasks or execute specific abstract data types according to the embodiment. (object), component (component) and data structure (data structure) may be included, and may not be limited to these.

The electronic device 900 may further include a user interface (UI) input device 950 and a UI output device 960. The UI input device 950 may receive a user input required for the operation of the electronic device 900. The UI output device 960 may output information or data according to an operation of the electronic device 900.

The apparatus described above may be implemented as a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the devices and components described in the embodiments include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), It can be implemented using one or more general purpose computers or special purpose computers, such as a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications executed on the operating system. In addition, the processing device may access, store, manipulate, process, and generate data in response to the execution of software. For the convenience of understanding, although it is sometimes described that one processing device is used, one of ordinary skill in the art, the processing device is a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of these, configuring the processing unit to behave as desired or processed independently or collectively. You can command the device. Software and/or data may be interpreted by a processing device or to provide instructions or data to a processing device, of any type of machine, component, physical device, virtual equipment, computer storage medium or device. , Or may be permanently or temporarily embodyed in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium.

The computer-readable recording medium may contain information used in embodiments according to the present invention. For example, a computer-readable recording medium may include a bitstream, and the bitstream may include information described in embodiments according to the present invention.

The computer-readable recording medium may include a non-transitory computer-readable medium.

The computer-readable medium may include program instructions, data files, data structures, and the like alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. -A hardware device specially configured to store and execute program instructions such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of the program instructions include not only machine language codes such as those produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

As described above, although the embodiments have been described by the limited embodiments and drawings, various modifications and variations are possible from the above description by those of ordinary skill in the art. For example, the described techniques are performed in a different order from the described method, and/or components such as a system, structure, device, circuit, etc. described are combined or combined in a form different from the described method, or other components Alternatively, even if substituted or substituted by an equivalent, an appropriate result can be achieved.

The apparatus described in the embodiments may include one or more processors and may include a memory. The memory may store one or more programs executed by one or more processors. One or more programs may perform the operation of the device described in the embodiments. For example, one or more programs of the device may perform the operations described in the steps associated with the device among the aforementioned steps. That is to say, the operation of the device described in the embodiment may be executed by one or more programs. One or more programs may include a program, an application, and an app of the device described above in the embodiment. For example, one of the one or more programs may correspond to a program, an application, and an app of the device described above in the embodiment.

100: measuring device
110: feature information extraction unit
120: distortion calculation unit
130: prediction model generation unit
140: model optimizer
150: prediction unit

Claims (1)

Calculating and extracting a plurality of feature information from the reference image and the comparison image;
Calculating distortion values between corresponding feature information from the extracted plurality of feature information;
Extracting a plurality of predictive models of correlations between the distortion values and subjective image quality and/or naturalness measurement values by using the calculated distortion values as inputs; And
Configuring a system using the plurality of prediction models, and determining weights for the plurality of prediction models, respectively
Measurement method comprising a.

Images