A system for reference-free image quality assessment using the Ripple II transformation and SVR

Abstract

A system (100) for reference-free image quality assessment, wherein the system (100) comprises: an input module (102) comprising an image input mechanism that enables the user to provide input images for quality assessment; a feature extraction module (104) comprising an extraction processor (104a) configured to extract perceptual features from a provided input image in the Ripplet II transformation, wherein 2D singularities present in the image are well-defined by the NSS-based feature in the Ripplet domain; a machine learning module (106) comprising a dedicated processor (106a) configured by the Support Vector Regression (SVR) technique to map the extracted perceptual features to an image quality score;and an output module (108) that is connected to the machine learning module and is configured to display the predicted image quality value as output to the user via a user interface (108a).

Description

FIELD OF THE INVENTIONThe present disclosure relates to a referenceless image quality evaluation (NR-IQA) system, and more particularly to a referenceless image quality evaluation system using Ripplet-II transformation and Support Vector Regression (SVR), wherein the system is used to determine the image quality value based on the features extracted from the image.BACKGROUND OF THE INVENTIONImage quality assessment (IQA) has become an important area of research in various areas, from optics and biomedical imaging to image capture and display systems. Accurate assessment of image quality is of decisive importance for many real applications. The IQA can be performed by subjective assessment by human observers or objective assessment using mathematical models.Subjective evaluations, while reliable, are time consuming and impractical for many applications. Objective Non-Reference (NR) IQA methods, in which the original image is not available, have evolved into an active research area. NR-IQA techniques typically include two steps: extracting perceptual features from the input image and then associating these features with a predicted quality value.Although various NR-IQA approaches have been proposed, there is still room for the development of new and improved models. The quality value estimated with the NR-IQA technique should correlate well with subjective human perception. Machine learning techniques such as support vector regression (SVR) and artificial neural networks have proven promising in coping with this challenge. In this regard, the present invention provides a system for predicting image quality using a novel Ripplet SVR model that uses the Ripplet II transform to detect 2D singularities in the input image and then associates the extracted features with an image quality value using an SVR framework.SUMMARY OF THE INVENTIONThe present disclosure relates to a system for referenceless image quality assessment. The proposed referenceless image quality evaluation system is configured to extract features from the input images and then associate these features with the quality value using SVR. The proposed system operates in two phases, wherein, in the first phase, perception features are extracted from the image and, in the second phase, these perception features are assigned by means of support vector regression (SVR) in order to determine the image quality value. The proposed method, Ripplet SVR, was tested using a popular LIVE database for image quality assessment that exhibits a strong correlation between Ripplet SVR and human perception. It has also been applied to various distortions such as JPEG compression, JPEG200 compression, blurring, noise, and fast fading. The experimental result demonstrates that Ripplet SVR outbreaks and can compete with the most advanced NR IQA metrics.The present disclosure is intended to provide a system for referenceless image quality assessment. The system comprises: an input module having an image input mechanism that allows the user to provide input images for quality assessment; a feature extraction module having an extraction processor configured to extract perception features from the provided input image in the ripplet-II transformation, wherein 2D singularities present in the image are well defined by the NSS-based feature in the ripplet domain; a machine learning module having a dedicated processor configured by the support vector regression (SVR) technique to associate the extracted perception features with an image quality value; and an output module connected to the machine learning module and configured to display the predicted image quality value to the user as output via a user interface.An object of the present disclosure is to provide a system for referenceless image quality assessment using Ripplet II transformation and Support Vector Regression (SVR).Another object of the present disclosure is to provide a referenceless image quality evaluation system that uses the Ripplet II transform to effectively detect 2D singularities such as edges along curves in the input image.Another object of the present disclosure is to develop a new statistical measure called "Ripplet-SVR" that converts the perceptual characteristics extracted using the Ripplet-II transformation into an image quality value through the use of a Support Vector Regression (SVR).Another object of the present disclosure is to show that the proposed Ripplet SVR method, when tested against a common LIVE database for image quality assessment, can exceed and compete with the most sophisticated referenceless image quality assessment metrics.Another object of the present disclosure is to apply the ripplet SVR method to various image distortions including JPEG compression, JPEG2000 compression, blurring, noise, and fast fading, and to exhibit its effectiveness in accurately judging the image quality.Another object of the present disclosure is to provide a robust and efficient reference-free image quality evaluation system that can be easily adopted for various real applications where accurate evaluation of image quality is of vital importance.In order to further clarify the advantages and features of the present disclosure, a more detailed description of the invention will be given with reference to specific embodiments thereof illustrated in the accompanying drawings. It should be noted that these drawings illustrate only typical embodiments of the invention and are therefore not to be considered as limiting the scope thereof. The invention will be described and explained with additional details and details with reference to the accompanying drawings.BRIEF DESCRIPTION OF THE FIGURESThese and other features, aspects, and advantages of the present disclosure will become more fully understood when the following detailed description is read with reference to the accompanying drawings, in which like characters represent like parts throughout. The following applies here: FIG. 1 is a block diagram of a non-reference image quality evaluation system according to an embodiment of the present disclosure; FIG. 2 is a diagram illustrating the general scope of the proposed system for referenceless image quality evaluation using Ripplet and SVR (Ripplet-SVR NR-IQA) according to an embodiment of the present disclosure.Moreover, those skilled in the art will appreciate that elements are shown in the drawings for simplicity and need not necessarily be drawn to scale. For example, the flowcharts illustrate the method using the major steps to improve understanding of aspects of the present disclosure. Moreover, with respect to the construction of the apparatus, one or more components of the apparatus may have been represented in the drawings by conventional symbols, and the drawings may only show the specific details relevant to understanding the embodiments of the present disclosure so as not to obscure the drawings with details readily apparent to those skilled in the art having the benefit of the description herein.DETAILED DESCRIPTION:In order to aid in the understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and described in specific language. It is to be understood, however, that no limitation of the scope of the invention is intended, since such changes and further modifications to the illustrated system and such further applications of the principles of the invention as illustrated therein are contemplated, as would normally occur to one skilled in the art to which the invention pertains.It will be understood by those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not intended to be limiting thereof.References throughout this specification to "one aspect," "another aspect," or similar language mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrase "in one embodiment," "in another embodiment," and similar language in this specification may or may not all refer to the same embodiment.The terms "comprises," "comprising," or other variations thereof are intended to cover a non-exclusive inclusion, such that a process or method comprising a list of steps not only comprises those steps, but may also comprise other steps not expressly listed or inherent in such a process or method. Likewise, one or more devices or subsystems or elements or structures or components preceded by "comprises...a" do not exclude, without further limitations, the existence of other devices or other subsystems or other elements or other structures or other components or additional devices or additional subsystems or additional elements or additional structures or additional components.Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The system, methods, and examples provided herein are illustrative only and are not to be considered limiting.Embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.The functional units described in this specification have been referred to as devices. A device may be implemented in programmable hardware devices such as processors, digital signal processors, central processing units, field programmable gate arrays, programmable array logic, programmable logic devices, cloud processing systems, or the like. The devices may also be implemented in software for execution by various types of processors. An identified device may include executable code and may include, for example, one or more physical or logical blocks of computer instructions, which may be organized, for example, as an object, procedure, function, or other construction. Nevertheless, the executable of an identified device need not be physically arranged together, but may include various instructions stored in various locations that, when logically interconnected, form the device and fulfil the stated purpose of the device.Indeed, executable code of a device or module may be a single instruction or multiple instructions, and may even be distributed across multiple different code segments, among different applications, and across multiple storage devices. Similarly, operational data may be identified and illustrated herein within the device and embodied in any suitable form and organized in any suitable type of data structure. The operational data may be acquired as a single dataset or distributed across different locations including different storage devices and may be present at least partially as electronic signals in a system or network.References throughout this specification to "a selected embodiment," "an embodiment," or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosed subject matter. Therefore, the appearances of the phrases "a selected embodiment," "in one embodiment," or "in one embodiment" in various places in this specification are not necessarily referring to the same embodiment.Moreover, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to provide a thorough understanding of the embodiments of the disclosed subject matter. However, one skilled in the art will recognize that the disclosed subject matter may be practiced without one or more of the specific details or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail in order to avoid obscuring aspects of the disclosed subject matter.According to the example embodiments, the disclosed computer programs or modules may be executed in many example ways, such as an application stored in a device's memory or a hosted application executing on a server and communicating with the device application or browser via a number of standard protocols such as TCP / IP, HTTP, XML, SOAP, REST, JSON, and other suitable protocols. The disclosed computer programs may be written in example programming languages that execute from memory on the device or from a hosted server, such as BASIC, COBOL, C, C++, Java, Pascal, or scripting languages such as JavaScript, Python, Ruby, PHP, Perl, or other suitable programming languages.Some of the disclosed embodiments include or otherwise include data transfer over a network, such as communication of various inputs or files over the network. The network may include, for example, one or more of the following: Internet, Wide Area Networks (WANs), Local Area Networks (LANs), analog or digital wired and wireless telephone networks (e.g., a PSTN, Integrated Services Digital Network (ISDN), a cellular network and digital subscriber line (xDS)), radio, television, cable, satellite, and / or any other transmission or tunneling mechanism for transmitting data. The network may include multiple networks or sub-networks, each of which may include, for example, a wired or wireless data path. The network may comprise a circuit switched voice network, a packet switched data network or any other network capable of transmitting electronic communication. For example, the network may comprise networks based on Internet Protocol (IP) or Asynchronous Transfer Mode (ATM), and may support voice using, for example, VoIP, voice over ATM, or other comparable protocols used for voice data communication. In one implementation, the network comprises a mobile telephone network configured to exchange text or SMS messages.Examples of the network include, but are not limited to, a personal area network (PAN), storage area network (SAN), home area network (HAN), campus area network (CAN), local area network (LAN), wide area network (WAN), metropolitan area network (MAN), virtual private network (VPN), enterprise private network (EPN), the Internet, global area network (GAN), and so forth.FIG. 1 is a block diagram of a non-reference image quality evaluation system according to an embodiment of the present disclosure.Referring to FIG. 1, the system (100) includes an input module (102) having an image input mechanism that allows the user to provide input images for quality evaluation; a feature extraction module (104) having an extraction processor (104a) configured to extract perception features from a provided input image in the ripplet-II transformation, wherein 2D singularities present in the image are well defined by the NSS-based feature in the ripplet domain; a machine learning module (106) having a dedicated processor (106a) configured by a support vector regression (SVR) technique to associate the extracted perception features with an image quality value; and an output module (108) connected to the machine learning module and configured to display the predicted image quality value to the user as output via a user interface (108a).In one embodiment, the extracted features comprise a generalized Gaussian model shape parameter, a frequency variation coefficient, an energy subband ratio measure, and an orientation model-based feature.In one embodiment, the extraction processor (104a) of the feature extraction module (104) is configured to split the input image into a plurality of patches, calculate contrast features in the ripplet II region by calculating an average ratio of ripplet II coefficients into non-DC to normalized DC coefficients for each patch, and calculate structural features in the ripplet II region using kurtosis and anisotropy.In one embodiment, the dedicated processor ( 106 a) of the machine learning module ( 106) uses an ε-SVR model with a radial base kernel (RBF) to map the perception features to the image quality value.In one embodiment, the system is configured to assess image quality for various distortions including JPEG compression, JPEG2000 compression, blurring, noise, and fast fading.The present invention relates to a system for performing referenceless image quality assessment using ripplelet-based feature extraction. Subsequently, these features are subjected to an SVR model to predict an image quality value. In the first phase, perceptual features are extracted from the image and in the second phase, these perceptual features are mapped using support vector regression (SVR) to determine the image quality value.The selection of significantly significant and relevant features is a key aspect of newly developed NR IQA algorithms. In order to improve the prediction task, the identification of the most relevant features is indispensable. In this study we propose NR-IQA algorithms using NSS features in the Ripplet II transformation domain. The NSS-based features in the ripplelet region effectively capture the 2D singularities present in images.FIG. 2 is a diagram illustrating the general scope of the proposed system for non-reference image quality evaluation using Ripplet and SVR (Ripplet-SVR NR-IQA) according to an embodiment of the present disclosure.The object of the present invention is to evaluate the image quality in the ripplet-II range. Normally, the global properties of an image have weaker correlations than the local properties. To analyze the local statistical attributes of the image, the image is divided into multiple regions and both texture and contrast features are extracted in the Ripplet II region.The contrast characteristics in the ripplet region are calculated as the average ratio of ripplet II coefficients in non-DC components to the normalized DC coefficient of each patch. Structural information is represented in the ripplet range by kurtosis and anisotropy. The overall frame of the proposed Ripplet SVR-NR-IQA algorithm is shown in Figure 2. It can be seen therein that four types of features are extracted, including: a) generalized Gaussian model shape parameter, b) frequency variation coefficient, c) measure of the energy subband ratio, and d) orientation model-based feature, each feature being described below.Generalised Gaussian Model Shape Parameters:The ripplet II coefficients of the image are analyzed using the generalized Gaussian density function defined by the mean (μ), scaling (β) and shape (γ) parameters. The shape parameter (γ) is estimated across image blocks using a percentile pooling technique. In particular, the average of the lowest becomes 10thpercentile and 100. Percentile pooling values calculated. This approach evaluates the uniformity of distortions present in the image and uses the two-percentile pooling method to effectively detect variations.Frequency Variation Coefficient (CFV):The coefficient of frequency variation (CFV), denoted as ξ, is calculated as the ratio of the standard deviation to the mean of the Ripplet II coefficient histogram, XX. Mathematically, it is expressed as follows:The CFV is evaluated for all image blocks using the percentile pooling method. The highest become 10th and 100th. The frequency differences in the image can be quantified by calculating percentile values of the local block values.Energy Subband Ratio:This function uses spectral feature analysis to describe the properties of distorted images. The energy subband ratio is derived from three defined frequency bands (n = 1, 2, 3) within each block. The energy subband ratio (R n) quantitates the relationship between the average energy of a frequency band and the cumulative energy up to that band and is calculated as follows:The final measurement uses the average of 10th and 100. Percentile values of the image blocks to provide an aggregated representation of the spectral properties.Orientation model-based function:Image distortions often interfere with local energy along certain orientations, thus changing the perceived quality of the image. To detect these distortions, three alignments are used to evaluate the local alignment energy within each block. The Generalized Gaussian Density (GGD) function is applied at each orientation to estimate the shape parameter (γ). Subsequently, the CFV (z) is calculated for each alignment. The average variance of ξ across the three orientations is calculated using the highest 10thpercentile and 100thpercentile. The result is that percentile values are derived from the image blocks and thus offer a robust measure of orientation-based distortions.To establish the relationship between the feature space and the quality score, machine learning techniques such as support vector regression (SVR) and neural networks are used. In this study, the LIBSVM-3.18 package is used to associate the feature space with the quality score. In particular, ε-SVR is applied with a radial base function core (RBF) to develop the regression model for the provided test features. This model predicts the corresponding quality score by using the features of the test dataset as input.In one embodiment, the proposed system is tested against the widely used LIVE IQA database. The LIVE database contains a total of 982 images categorized into reference images and distorted images. More specifically, it contains 29 reference images and 779 distorted images, each evaluated by 161 observers for subjective quality values. The data set represents five types of distortions: JPEG compression, JPEG2000 compression, white noise, Gaussian blurring, and a Rayleigh fading channel. Each type of distortion is further divided into five stages, which differ in the percentage of distortion. To evaluate the image quality, the database is randomly divided into two subsets: a training set and a test set. The training set comprising 80% of the images is used to create and refine the support vector regression (SVR) model. This model is then applied to the test set containing the remaining 20% of the images to evaluate its performance.Support vector regression (SVR) and neural networks are common machine learning methods that are used to establish a relationship between the feature set and a quality score. This study uses the LIBSVM-3.18 package to associate the feature space with the quality score. In particular, the ε-SVR approach is applied with a radial base function (RBF) kernel to develop a regression model for the provided feature set of test data. The model uses features from the test dataset to predict a quality score.For an image i, the features are represented as vectors F i= [ F 1, F 2, F 3..... F 24], wherein each element denotes a particular attribute. The mean opinion score (MOS) for the same image is referred to as Si. The function Ŷ i= f(F i) is used to predict the small deviation score, ∈ oscillations i. Mathematically, the function is defined as Ŷ i= w T φ(F) + y, where φ(F) is the kernel function, w represents the weight vector, and γ !gammais the bias term. Both w and γ belonging are initially unknown and are estimated during the training phase to ensure that the prediction error remains below a predefined threshold ∈ \ epision.In the training phase, the SVR model obtains the input features F i together with the corresponding predicted values Y i and calculates the optimum values of w and γ !gamma. During the test phase, this trained model uses the features of the test dataset to predict the image quality value.The performance of the proposed Ripplet SVR method is evaluated using metrics such as the mean square error (RMSE), the linear correlation coefficient according to Pearson (PLCC), the rank correlation coefficient according to Spaerman (SROCC), and the rank correlation coefficient according to Kendall (KROCC). In an image database, PLCC evaluates the linear correlation between objective and subjective quality values, SROCC measures the rank correlation, and KROCC evaluates match and discordance in paired scores. RMSE quantitates the average deviation between predicted and subjective scores. Better performance is indicated by higher PLCC, SROCC, and KROCC values and a lower RMSE.The invention analyzes these performance metrics using subjective evaluations such as DMOS / MOS and objective evaluations. Researchers typically prioritize LCC and SROCC in evaluating non-reference image quality evaluation (NRIQA) techniques. In this work, the Ripplelet SVR method is compared against established NRIQA techniques using the LIVE database, including SESANIA(S), SESANIA(N), SHANIA, CurveletQ, BRISQUE, BIQI, BLIINDS-II, and SSEQ. Repeated experiments with a train test split of 80%-20% take into account the mean power values from 1000 iterations.The ripplet SVR exhibits comparable or better performance compared to other approaches. The results show a strong consistency across all distortion categories such as JP2K, JPEG, noise, blurring, and fast fading. The Ripplet SVR will correlate robust with subjective scores, confirming its reliability. Moreover, it maintains favorable standard deviations for performance metrics, emphasizing stability.In summary, the system based on Ripplet-SVR, which uses Ripplet-II region statistical feature extraction and support vector regression (SVR), provides a novel, effective solution for non-referencing image quality assessment. It closely matches the human perception of image distortions and is thus a promising alternative to existing NRIQA methods.The drawings and the foregoing description provide examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be divided into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, orders of the processes described herein may be changed and are not limited to the manner described herein. Moreover, the actions of any flow chart need not be implemented in the order shown; nor do all actions necessarily need to be performed. Also, those actions that are not dependent on other actions may be performed in parallel with the other actions. The scope of the embodiments is in no way limited by these specific examples. Numerous variations, whether or not explicitly stated in the specification, such as differences in structure, dimension, and material usage, are possible. The scope of the embodiments is at least as broad as recited in the following claims.Advantages, other advantages, and solutions to problems have been described above with respect to certain embodiments. However, the advantages, benefits, solutions to problems, and any components that may result in an advantage, solution occurring or becoming clearer are not to be construed as a critical, required, or essential feature or component of any or all claims.REFERENCES100 A system for assessing the quality of referenceless images. 102 Input module 104 Module for feature extraction 104 a Extraktions processor 106 Module for machine learning 106 a Dedizierte processor 108 Output module 108 a Benutzer 202 Image database 204 Training data 206 Test data 208 Ripplet-II Transform 208 a Extrahierte features 208 b Merkmals extraction 210-SVR Model for point calculation 212 Image quality evaluation

Claims

A system (100) for non-reference image quality assessment, the system (100) comprising: an input module (102) comprising an image input mechanism that allows the user to provide input images for quality assessment; a feature extraction module (104) comprising an extraction processor (104a) configured to extract perception features from a provided input image in the Ripplet-II transformation, wherein 2D singularities present in the image are well defined by the NSS-based feature in the Ripplet domain; a machine learning module (106) comprising a dedicated processor (106a) configured by the Support Vector Regression (SVR) technique to associate the extracted perception features with an image quality value; and an output module (108) connected to the machine learning module and configured to display the predicted image quality value as output to the user via a user interface (108a).The system (100) of claim 1, wherein the extracted features comprise a generalized shape parameter of a Gaussian model, a frequency variation coefficient, a measure of the energy subband ratio, and a feature based on the orientation model.The system (100) of claim 1, wherein the extraction processor (104a) of the feature extraction module (104) is configured to split the input image into a plurality of patches, calculate contrast features in the ripplet II region by calculating an average ratio of ripplet II coefficients into non-DC to normalized DC coefficients for each patch, and calculate structure features in the ripplet II region using kurtosis and anisotropy.The system (100) of claim 1, wherein the dedicated processor (106a) of the machine learning module (106) uses an ε-SVR model with a radial basis function (RBF) kernel to map the perception features to the image quality value.The system (100) of claim 1, wherein the system is configured to assess image quality for various distortions including JPEG compression, JPEG2000 compression, blurring, noise, and fast fading.

Images