A super-resolution effect evaluation method and device based on high-resolution measurement data

Abstract

The present invention discloses a super-resolution effect evaluation method and device based on high-resolution measurement data. The method comprises: obtaining a low-resolution image and a high-resolution image database, selecting the high-resolution image database to obtain resolution measurement data related to the low-resolution image; processing the resolution measurement data related to the low-resolution image to obtain a ground object dataset; performing super-resolution reconstruction processing on the low-resolution image using a super-resolution algorithm to be evaluated to obtain a super-resolution processing result; and performing resolution evaluation on the super-resolution processing result based on the ground object dataset to obtain the resolution of the super-resolution processing result. The method of the present invention can not only overcome the inefficiency and uncertainty of traditional subjective evaluation methods, but also improve the defects of existing objective quantitative evaluation methods, such as poor consistency with human visual perception on the one hand, and strong correlation with training datasets on the other hand, but insufficient generalization ability. It can achieve accurate quantitative, intuitive and efficient evaluation of super-resolution result images.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a super-resolution effect evaluation method and device based on high-resolution measurement data. Background Art

[0002] How to achieve accurate evaluation of super-resolution processing effects directly affects the performance optimization and algorithm improvement of super-resolution technology in the field of image processing, and determines the application effect and promotion scope of this technology.

[0003] Based on practical work experience, the most intuitive and accurate evaluation method is subjective evaluation. This method uses visual observation by participants, combined with cerebral empirical information, to compare subjective experience and feelings, resulting in vague qualitative evaluation conclusions. This method suffers from individual uncertainty and susceptibility to external interference. Furthermore, it is complex and time-consuming, requiring high labor costs. This makes subjective evaluation methods unsuitable for applications requiring high real-time performance or for ground processing systems requiring high batch standardization.

[0004] From a practical and efficiency perspective, only quantifiable, objective evaluation methods can be applied in the aforementioned scenarios. Existing objective evaluation methods are generally categorized into three types: full-reference, partial-reference, and no-reference quality assessment methods. Full-reference quality assessment methods require the full information of high-resolution ground-truth images, partial-reference quality assessment methods require partial information of ground-truth images, and no-reference quality assessment methods do not require ground-truth images as a reference. Since obtaining high-resolution ground-truth images as a reference is difficult in reality, no-reference quality assessment methods are relatively more important. Most research papers, both domestic and international, evaluate super-resolution images based on the full-reference metrics Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM). These metrics primarily measure the difference between the super-resolution result and the high-resolution ground-truth image, but in reality, there is a significant gap between these metrics and the true visual experience. Furthermore, full-reference metrics require a reference image for comparison. In practice, obtaining true high- and low-resolution data pairs is difficult. Consequently, low-resolution images (sequences) can only be simulated using a simplified degraded imaging model using an interpolation degradation algorithm and Gaussian noise. However, such data deviates significantly from the actual degradation process, especially in the case of non-Gaussian noise. Consequently, using full-reference metrics for super-resolution evaluation yields poor performance.

[0005] Current academic no-reference quality assessment methods still have certain limitations. These methods are primarily based on machine learning and perform well on small datasets, but their robustness and generalization capabilities are significantly insufficient. Generally speaking, they have three main limitations: 1) These models rely on handcrafted features, which require sufficient prior knowledge and practical skills in feature design; 2) Traditional machine learning algorithms lack generalization capabilities. While they can achieve good results on small datasets, their performance degrades significantly as the dataset increases. 3) Deep learning-based methods are limited by insufficient training data. Therefore, existing no-reference image quality assessment methods cannot be applied independently. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the current lack of accurate and efficient super-resolution effect evaluation methods in the field of image processing. This invention provides a super-resolution effect evaluation method and device based on high-resolution metric data. This method combines subjective evaluation with no-reference quality evaluation methods and proposes the use of high-resolution ground object images as resolution metric benchmark data. Operators can use this metric benchmark data to compare with the super-resolution processing results to complete the evaluation of the super-resolution reconstruction results. This method can overcome the inefficiency and uncertainty of traditional subjective evaluation methods and improve the shortcomings of existing objective quantitative evaluation methods, such as poor consistency with human visual perception and strong correlation with training datasets but insufficient generalization ability. This method can achieve accurate quantitative, intuitive, and efficient evaluation of super-resolution image results.

[0007] In order to solve the above technical problems, the first aspect of the embodiments of the present invention discloses a super-resolution effect evaluation method based on high-resolution measurement data, the method comprising:

[0008] S1, obtain low-resolution image I L ;

[0009] S2, obtaining a high-resolution image database, selecting the high-resolution image database, and obtaining resolution measurement data related to the low-resolution image;

[0010] S3, processing the resolution measurement data related to the low-resolution image to obtain a ground feature dataset

[0011] S4, the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ;

[0012] S5, based on the ground feature dataset The super-resolution processing result I sPerform resolution evaluation to obtain the super-resolution processing result I s Resolution s .

[0013] As an optional implementation manner, in the first aspect of the embodiment of the present invention, the acquiring of a high-resolution image database and the selection of the high-resolution image database to obtain resolution measurement data related to the low-resolution image include:

[0014] S21, processing the low-resolution image to obtain a resolution range of the low-resolution image;

[0015] S22: selecting the high-resolution image database according to the resolution range of the low-resolution image to obtain resolution measurement data related to the low-resolution image.

[0016] As an optional implementation manner, in the first aspect of the embodiments of the present invention, processing the low-resolution image to obtain a resolution range of the low-resolution image includes:

[0017] S211, analyzing the low-resolution image to obtain a resolution estimation of the low-resolution image;

[0018] S212, processing the estimated resolution to obtain a resolution range of the low-resolution image;

[0019] The resolution range of the low-resolution image is L min ~L max ,in Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0020] As an optional implementation, in the first aspect of the embodiment of the present invention, the resolution measurement data related to the low-resolution image is processed to obtain a ground feature dataset. include:

[0021] S31, from k = L min Initially, according to preset ground feature characteristic requirements, the resolution measurement data related to the low-resolution image is searched to obtain ground feature data information to be selected;

[0022] S32, according to the preset feature selection requirements, search the data information of the feature to be selected to obtain the data subset T corresponding to the resolution k k ;

[0023] S33, k=k+1, execute S31~S33, select the next resolution level feature data, until k>Lmax, and obtain the feature data set

[0024] Among them, L min ~L max is the resolution range of the low-resolution image, Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0025] As an optional implementation manner, in the first aspect of the embodiment of the present invention, the preset feature selection requirement is:

[0026] Taking the estimated resolution l of the low-resolution image as a benchmark, min ~L max Select corresponding ground features in a group of 1 meter within the range;

[0027] Each group selected 3 to 5 features;

[0028] According to the resolution range L of the low-resolution image min ~L max Select the features;

[0029] The selected features include three images. The first image is the feature navigation information, which is used to show the location of the feature within a larger area. The second image is the mesoscale feature information, which is used to show the feature outline and assist in determining the feature's location. The third image is the feature detail image, which is required to show the feature's geometric outline information in detail and annotate the scale data reflecting the resolution characteristics for resolution assessment.

[0030] Name the feature data images according to the naming conventions.

[0031] As an optional implementation, in the first aspect of the embodiment of the present invention, the ground feature dataset The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s ,include:

[0032] S51, from i=L min Start with the feature dataset The feature data T i As a benchmark, according to the preset ground object navigation information and mesoscale information, in the super-resolution processing result I s Choose from the list and get the value of T icorresponding features;

[0033] S52, select M operators, where M is an odd number greater than or equal to 3, and i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding ground features;

[0034] The resolution evaluation result is distinguishable or indistinguishable;

[0035] S53, i=i+1, repeat S51 to S53 until i>L max , get the feature dataset The resolution evaluation results of the corresponding ground features;

[0036] S54, the ground feature dataset The resolution evaluation result of the corresponding ground object is processed to obtain the super-resolution processing result I s Resolution s .

[0037] As an optional implementation, in the first aspect of the embodiment of the present invention, the selection of M operators, M is an odd number greater than or equal to 3, i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding features include:

[0038] S521, if operator j, j=1,2,…,M thinks that in I s Information and datasets of Chinese features The information in is basically the same, so operator j determines that I s Zhong and T i The resolution evaluation result of the corresponding feature is distinguishable, otherwise operator j determines I s Zhong and T i The resolution evaluation result of the corresponding ground feature is indistinguishable;

[0039] S522, for a certain resolution l1, and T i There are multiple corresponding ground objects. If more than half of the ground objects in the resolution evaluation results of M operators can be distinguished, it is determined that the resolution reaches l1+1; otherwise, it is determined that the resolution does not reach l1+1.

[0040] A second aspect of an embodiment of the present invention discloses a super-resolution effect evaluation device based on high-resolution measurement data, the device comprising:

[0041] Low-resolution image acquisition module, used to acquire low-resolution image I L ;

[0042] a high-resolution image database acquisition module, configured to acquire a high-resolution image database, select the high-resolution image database, and obtain resolution measurement data related to the low-resolution image;

[0043] A ground object dataset construction module is used to process the resolution measurement data related to the low-resolution image to obtain a ground object dataset

[0044] A super-resolution processing module is used to process the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ;

[0045] A super-resolution effect evaluation module is used to evaluate the effect of the ground object dataset. The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s .

[0046] As an optional implementation manner, in the second aspect of the embodiment of the present invention, the obtaining of a high-resolution image database and the selection of the high-resolution image database to obtain resolution measurement data related to the low-resolution image include:

[0047] S21, processing the low-resolution image to obtain a resolution range of the low-resolution image;

[0048] S22: selecting the high-resolution image database according to the resolution range of the low-resolution image to obtain resolution measurement data related to the low-resolution image.

[0049] As an optional implementation manner, in the second aspect of the embodiments of the present invention, processing the low-resolution image to obtain a resolution range of the low-resolution image includes:

[0050] S211, analyzing the low-resolution image to obtain a resolution estimation of the low-resolution image;

[0051] S212, processing the estimated resolution to obtain a resolution range of the low-resolution image;

[0052] The resolution range of the low-resolution image is L min ~L max ,in Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0053] As an optional implementation, in the second aspect of the embodiment of the present invention, the resolution measurement data related to the low-resolution image is processed to obtain a ground feature dataset. include:

[0054] S31, from k = L min Initially, according to preset ground feature characteristic requirements, the resolution measurement data related to the low-resolution image is searched to obtain ground feature data information to be selected;

[0055] S32, according to the preset feature selection requirements, search the data information of the feature to be selected to obtain the data subset T corresponding to the resolution k k ;

[0056] S33, k=k+1, execute S31~S33, select the next resolution level feature data, until k>L max , get the feature dataset

[0057] Among them, L min ~L max is the resolution range of the low-resolution image, Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0058] As an optional implementation manner, in the second aspect of the embodiment of the present invention, the preset feature selection requirement is:

[0059] Taking the estimated resolution l of the low-resolution image as a benchmark, min ~L max Select corresponding ground features in a group of 1 meter within the range;

[0060] Each group selected 3 to 5 features;

[0061] According to the resolution range L of the low-resolution image min ~L max Select the features;

[0062] The selected features include three images. The first image is the feature navigation information, which is used to show the location of the feature within a larger area. The second image is the mesoscale feature information, which is used to show the feature outline and assist in determining the feature's location. The third image is the feature detail image, which is required to show the feature's geometric outline information in detail and annotate the scale data reflecting the resolution characteristics for resolution assessment.

[0063] Name the feature data images according to the naming conventions.

[0064] As an optional implementation, in the second aspect of the embodiment of the present invention, the ground feature dataset The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s ,include:

[0065] S51, from i=L min Start with the feature dataset The feature data T i As a benchmark, according to the preset ground object navigation information and mesoscale information, in the super-resolution processing result I s Choose from the list and get the value of T i corresponding features;

[0066] S52, select M operators, where M is an odd number greater than or equal to 3, and i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding ground features;

[0067] The resolution evaluation result is distinguishable or indistinguishable;

[0068] S53, i=i+1, repeat S51 to S53 until i>L max , get the feature dataset The resolution evaluation results of the corresponding ground features;

[0069] S54, the ground feature dataset The resolution evaluation result of the corresponding ground object is processed to obtain the super-resolution processing result I s Resolution s .

[0070] As an optional implementation, in the second aspect of the embodiment of the present invention, the selection of M operators, M is an odd number greater than or equal to 3, for the T i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding features include:

[0071] S521, if operator j, j=1,2,…,M thinks that in I s Information and datasets of Chinese features The information in is basically the same, so operator j determines that I s Zhong and T iThe resolution evaluation result of the corresponding feature is distinguishable, otherwise operator j determines I s Zhong and T i The resolution evaluation result of the corresponding ground feature is indistinguishable;

[0072] S522, for a certain resolution l1, and T i There are multiple corresponding ground objects. If more than half of the ground objects in the resolution evaluation results of M operators can be distinguished, it is determined that the resolution reaches l1+1; otherwise, it is determined that the resolution does not reach l1+1.

[0073] A third aspect of the present invention discloses another super-resolution effect evaluation device, comprising:

[0074] a memory storing executable program code;

[0075] a processor coupled to the memory;

[0076] The processor calls the executable program code stored in the memory to execute part or all of the steps in the super-resolution effect evaluation method based on high-resolution measurement data disclosed in the first aspect of the embodiment of the present invention.

[0077] The fourth aspect of the present invention discloses a computer-storable medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the super-resolution effect evaluation method based on high-resolution measurement data disclosed in the first aspect of an embodiment of the present invention.

[0078] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0079] The present invention obtains resolution measurement data from a high-resolution image library, which can provide a reliable benchmark for operators to evaluate super-resolution processing results, ensuring that operators can accurately determine the image resolution. This forms a novel and practical super-resolution effect evaluation method. The method of the present invention can not only overcome the randomness problem of existing subjective evaluation methods, but also solve the distortion problem of current objective evaluation methods. In practical applications, the method of the present invention can measure the degree of resolution improvement of the super-resolution algorithm by comparing the changes in image resolution before and after super-resolution, thereby promoting the continuous optimization of a single algorithm; it can also compare the performance differences of different algorithms in improving resolution, providing a basis for the selection of super-resolution algorithms in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0081] Figure 1 This is a flow chart of a method for evaluating super-resolution effects based on high-resolution measurement data disclosed in an embodiment of the present invention;

[0082] Figure 2 1 is a flow chart of another super-resolution effect evaluation method based on high-resolution measurement data disclosed in an embodiment of the present invention;

[0083] Figure 3 This is a flow chart of constructing a resolution measurement dataset using high-resolution images disclosed in an embodiment of the present invention;

[0084] Figure 4 This is a schematic diagram of the naming of the ground feature data pictures disclosed in the embodiment of the present invention. Figure 1 ;

[0085] Figure 5 This is a schematic diagram of the naming of the ground feature data pictures disclosed in the embodiment of the present invention. Figure 2 ;

[0086] Figure 6 is a resolution evaluation flow chart disclosed in an embodiment of the present invention;

[0087] Figure 7 1 is a schematic structural diagram of a super-resolution effect evaluation device based on high-resolution measurement data disclosed in an embodiment of the present invention;

[0088] Figure 8 This is a structural diagram of another super-resolution effect evaluation device based on high-resolution measurement data disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0089] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0090] The terms "first," "second," and so on, in the description and claims of the present invention and the accompanying drawings are used to distinguish between different objects, not to describe a specific order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to the process, method, product, or device.

[0091] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0092] The present invention discloses a super-resolution effect evaluation method and device based on high-resolution measurement data, the method comprising: obtaining a low-resolution image I L ; Obtain a high-resolution image database, select the high-resolution image database, and obtain resolution measurement data related to low-resolution images; process the resolution measurement data related to low-resolution images to obtain a ground feature dataset For low-resolution image I L Perform super-resolution reconstruction to obtain super-resolution processing result I s ; Based on the feature dataset Super-resolution processing results I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s The method of the present invention can not only overcome the inefficiency and uncertainty of traditional subjective evaluation methods, but also improve the defects of existing objective quantitative evaluation methods, such as poor consistency with human visual perception on the one hand, and strong correlation with training data sets but insufficient generalization ability on the other hand. It can achieve accurate quantitative, intuitive and efficient evaluation of super-resolution image results.

[0093] This method utilizes image data with fixed ground scales from a high-resolution database as resolution measurement data. For example, these data include oil tanks of various apertures in large oil storage bases, airports, docks, and large-scale buildings, which have permanent geometric outlines and easily measurable scales. By subjectively comparing the super-resolution image with the same features in the high-resolution image, the resolution of the super-resolution image is measured, enabling quantitative evaluation of the super-resolution effect in a manner consistent with human visual perception. These are explained in detail below.

[0094] Example 1

[0095] See also Figure 1 , Figure 1 This is a flow chart of a method for evaluating super-resolution effects based on high-resolution measurement data disclosed in an embodiment of the present invention. Figure 1 The described super-resolution effect evaluation method based on high-resolution measurement data is applied to the field of image processing and is used to evaluate the performance of image super-resolution processing algorithms, which is not limited in the embodiments of the present invention. Figure 1 As shown, the super-resolution effect evaluation method based on high-resolution measurement data may include the following operations:

[0096] S1, obtain low-resolution image I L ;

[0097] S2, obtaining a high-resolution image database, selecting the high-resolution image database, and obtaining resolution measurement data related to the low-resolution image;

[0098] S3, processing the resolution measurement data related to the low-resolution image to obtain a ground feature dataset

[0099] S4, the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ;

[0100] S5, based on the ground feature dataset The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s .

[0101] Optionally, the acquiring of a high-resolution image database and selecting the high-resolution image database to obtain resolution measurement data related to the low-resolution image include:

[0102] S21, processing the low-resolution image to obtain a resolution range of the low-resolution image;

[0103] S22: selecting the high-resolution image database according to the resolution range of the low-resolution image to obtain resolution measurement data related to the low-resolution image.

[0104] Optionally, the processing the low-resolution image to obtain a resolution range of the low-resolution image includes:

[0105] S211, analyzing the low-resolution image to obtain a resolution estimation of the low-resolution image;

[0106] S212, processing the estimated resolution to obtain a resolution range of the low-resolution image;

[0107] The resolution range of the low-resolution image is L min ~L max ,in Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0108] Optionally, the resolution measurement data related to the low-resolution image is processed to obtain a ground feature dataset include:

[0109] S31, from k = L min Initially, according to preset ground feature characteristic requirements, the resolution measurement data related to the low-resolution image is searched to obtain ground feature data information to be selected;

[0110] S32, according to the preset feature selection requirements, search the data information of the feature to be selected to obtain the data subset T corresponding to the resolution k k ;

[0111] S33, k=k+1, execute S31~S33, select the next resolution level feature data, until k>L max , get the feature dataset

[0112] Among them, L min ~L max is the resolution range of the low-resolution image, Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

[0113] Optionally, the preset feature selection requirements are:

[0114] Taking the estimated resolution l of the low-resolution image as a benchmark, min ~L max Select corresponding ground features in a group of 1 meter within the range;

[0115] Each group selected 3 to 5 features;

[0116] According to the resolution range L of the low-resolution imagemin ~L max Select the features;

[0117] The selected features include three images. The first image is the feature navigation information, which is used to show the location of the feature within a larger area. The second image is the mesoscale feature information, which is used to show the feature outline and assist in determining the feature's location. The third image is the feature detail image, which is required to show the feature's geometric outline information in detail and annotate the scale data reflecting the resolution characteristics for resolution assessment.

[0118] Name the feature data images according to the naming conventions.

[0119] Optionally, the feature dataset The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s ,include:

[0120] S51, from i=L min Start with the feature dataset The feature data T i As a benchmark, according to the preset ground object navigation information and mesoscale information, in the super-resolution processing result I s Choose from the list and get the value of T i corresponding features;

[0121] S52, select M operators, where M is an odd number greater than or equal to 3, and i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding ground features;

[0122] The resolution evaluation result is distinguishable or indistinguishable;

[0123] S53, i=i+1, repeat S51 to S53 until i>L max , get the feature dataset The resolution evaluation results of the corresponding ground features;

[0124] S54, the ground feature dataset The resolution evaluation result of the corresponding ground object is processed to obtain the super-resolution processing result I s Resolution s .

[0125] Optionally, the operator is selected M, M is an odd number greater than or equal to 3, and ... i The corresponding ground features are compared and interpreted to obtain the iThe resolution evaluation results of the corresponding features include:

[0126] S521, if operator j, j=1,2,…,M thinks that in I s Information and datasets of Chinese features The information in is basically the same, so operator j determines that I s Zhong and T i The resolution evaluation result of the corresponding feature is distinguishable, otherwise operator j determines I s Zhong and T i The resolution evaluation result of the corresponding ground feature is indistinguishable;

[0127] S522, for a certain resolution l1, and T i There are multiple corresponding ground objects. If more than half of the ground objects in the resolution evaluation results of M operators can be distinguished, it is determined that the resolution reaches l1+1; otherwise, it is determined that the resolution does not reach l1+1.

[0128] Example 2

[0129] See also Figure 2 , Figure 2 This is a flow chart of another method for evaluating super-resolution effects based on high-resolution measurement data disclosed in an embodiment of the present invention. Figure 2 The described super-resolution effect evaluation method based on high-resolution measurement data is applied to the field of image processing and is used to evaluate the performance of image super-resolution processing algorithms, which is not limited in the embodiments of the present invention. Figure 2 As shown, the super-resolution effect evaluation method based on high-resolution measurement data may include the following operations:

[0130] During the operation, the low-resolution image I is input L On the one hand, according to I L The area where the image is located and the approximate resolution l of the image are selected. According to the principle of selecting resolution measurement data, the ground objects are selected in the public high-resolution image database, and the matching ground object local images are extracted to construct the resolution measurement dataset. On the other hand, the super-resolution algorithm to be evaluated is used to L Perform super-resolution reconstruction processing to generate super-resolution result I s ; Then use the resolution metric dataset to measure I s Perform resolution evaluation. The method includes the following steps:

[0131] Step S1: Input low-resolution image I L ;

[0132] Step S2: Using the high-resolution image database, select L Related resolution measurement data and constructing ground feature datasets

[0133] Step S3: Input low-resolution image I L , apply the preset super-resolution algorithm to perform super-resolution reconstruction processing and generate super-resolution processing result I s ;

[0134] Step S4: Using the object dataset Super-resolution processing results I s Perform resolution evaluation and identify I s Resolution s .

[0135] The super-resolution algorithm to be evaluated can be any self-developed or third-party developed algorithm, and the present invention does not impose any limitation thereto.

[0136] Based on the approximate resolution l of the low-resolution image to be input, it is assumed that Indicates rounding down. Indicates rounding up, l is the estimated resolution of the low-resolution image. min ~L max Select the feature data with the corresponding resolution. Figure 3 This is a flow chart of constructing a resolution measurement dataset using high-resolution images disclosed in an embodiment of the present invention.

[0137] In the method for constructing a resolution measurement dataset using high-resolution images described in the above steps, step S2 includes the following steps:

[0138] Step S21: opening a high-resolution image database;

[0139] Step S22: Based on the input low-resolution image I L The resolution range is set to call out the corresponding high-resolution image;

[0140] Step S23: From k=L min At the beginning, according to the characteristics of the ground, C f Search for the objects to be selected, requiring at least 3 objects to be searched;

[0141] Step S24: Select the required C op Extract the data image of the selected object to form a data subset T corresponding to the resolution k k ;

[0142] Step S25: k=k+1, select the next resolution level feature data until k>L max ;

[0143] The relevant rules for selecting resolution measurement data as mentioned above are:

[0144] To achieve an intuitive evaluation of image resolution, it is necessary to select measurable large fixed-ground feature image slices from public high-resolution image data sources based on the satellite's resolution characteristics. Specifically, the following two rules must be followed:

[0145] 1) Terrain feature requirements C f :

[0146] (1) It should be a permanent feature, the geometric outline or diameter of its circular cross section will not change over a long period of time;

[0147] (2) It has a clear linear outline and a significant contrast between light and dark with the background in the image;

[0148] (3) From a bird's-eye view, the features should be relatively independent and not be confused with surrounding features;

[0149] (4) The horizontal height of the feature is not high, and the geometric characteristics do not change much when the observation angle is different. The top and side are clearly distinguishable and will not be confused;

[0150] (5) The scale of the object should be compatible with the resolution of the image data to be detected (see the requirements for object selection).

[0151] 2) Feature selection requirements C op :

[0152] (1) Based on the integer value of the resolution of the input image data, Select the corresponding ground features within the range in groups of 1 meter each;

[0153] (2) Each group should try to select 3-5 ground objects for alternative use in resolution testing;

[0154] (3) The features should be selected based on the area covered by the input image data;

[0155] (4) The selected feature data should include three pictures. The first picture is the feature navigation information, which shows the location of the feature within a larger area, so as to facilitate the correspondence with the feature location in the image to be measured; the second picture is the mesoscale feature information, which is used to roughly display the feature outline and assist in determining the feature location; the third picture is the feature detail image, which is required to display the geometric outline information of the feature in detail and mark the scale data reflecting the resolution characteristics (in meters, accurate to one decimal place) for resolution assessment.

[0156] (5) Name the feature data images according to the naming conventions. The naming convention for feature detail images is: feature location region name_serial number_scale (integer part)_detail; the naming convention for mid-scale feature information is: feature location region name_serial number_scale (integer part)_middle; the naming convention for feature navigation information is: feature location region name_serial number_scale (integer part)_overview; the serial number refers to the order during screening, starting from 1; the scale refers to the length of the selected feature information annotation, taking the integer part of the annotation length. Figure 4 As shown, the image comes from a public channel, Google Earth. The image is named: area name-7-11-detail and is in jpg format. When there are multiple annotations in the image, the scale number is determined according to the smaller scale. Figure 5 As shown, the map is named: area name-23-10-detail.jpg, and the image comes from the public channel, Google Earth.

[0157] In the method for evaluating super-resolution results using the resolution metric dataset constructed in step S2 described above, step S4 includes the following steps:

[0158] Step S41: From i=L min Start measuring T in the dataset with resolution i As a benchmark, find T based on the ground object navigation information and mesoscale information i In the results I s The corresponding features in

[0159] Step S42: Select at least 3 operators (the number of operators should be an odd number) to perform comparative interpretation. If operator j thinks that s The information of the features in the resolution measurement data is basically consistent with that in the resolution measurement data, then operator j determines that the features in the measurement data are T i in I s Otherwise, it is considered that the feature T i in I s Indistinguishable, operator j’s response to the object T i in I s Whether the identifiable information is recorded;

[0160] Step S43: i=i+1, use the resolution measurement data T in turn i in I s Perform resolution evaluation until i>L max ;

[0161] Step S44: Analyze and summarize the recorded information and determine the resolution according to the following method:

[0162] S441, if there are multiple objects corresponding to a certain resolution l, then if more than half of the objects in the evaluation record can be resolved, it can be determined that the result resolution reaches l+1; otherwise, it is determined that the result resolution does not reach l+1;

[0163] S442, take the minimum value (smallest value) of the resolution recognition result of operator i as the I recognized by the operator s Resolution s ;

[0164] S443, all operators identified I s The resolution of I s Certified resolution l s . Figure 6 This is a resolution evaluation flowchart disclosed in an embodiment of the present invention.

[0165] Example 3

[0166] See also Figure 7 , Figure 7 This is a schematic diagram of the structure of a super-resolution effect evaluation device based on high-resolution measurement data disclosed in an embodiment of the present invention. Figure 7 The described super-resolution effect evaluation device based on high-resolution measurement data is applied to the field of image processing to evaluate the performance of image super-resolution processing algorithms, which is not limited in the embodiments of the present invention. Figure 7 As shown, the super-resolution effect evaluation device based on high-resolution measurement data may include the following operations:

[0167] S301, a low-resolution image acquisition module, used to acquire a low-resolution image I L ;

[0168] S302, a high-resolution image database acquisition module, configured to acquire a high-resolution image database, select the high-resolution image database, and obtain resolution measurement data related to the low-resolution image;

[0169] S303, a ground object dataset construction module is used to process the resolution measurement data related to the low-resolution image to obtain a ground object dataset

[0170] S304, a super-resolution processing module for processing the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ;

[0171] S305, a super-resolution effect evaluation module is used to evaluate the effect of the super-resolution effect according to the ground object dataset. The super-resolution processing result Is Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s .

[0172] Example 4

[0173] See also Figure 8 , Figure 8 This is a schematic diagram of the structure of another super-resolution effect evaluation device based on high-resolution measurement data disclosed in an embodiment of the present invention. Figure 8 The described super-resolution effect evaluation device based on high-resolution measurement data is applied to the field of image processing to evaluate the performance of image super-resolution processing algorithms, which is not limited in the embodiments of the present invention. Figure 8 As shown, the super-resolution effect evaluation device based on high-resolution measurement data may include the following operations:

[0174] A memory 401 storing executable program code;

[0175] a processor 402 coupled to the memory 401;

[0176] The processor 402 calls the executable program code stored in the memory 401 to execute the steps of the super-resolution effect evaluation method based on high-resolution measurement data described in the first and second embodiments.

[0177] Example 5

[0178] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the super-resolution effect evaluation method based on high-resolution measurement data described in the first embodiment.

[0179] The device embodiments described above are merely illustrative. Modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules, i.e., they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0180] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by means of hardware. Based on this understanding, the above technical solution, in essence, or the portion that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0181] Finally, it should be noted that the super-resolution effect evaluation method and device based on high-resolution measurement data disclosed in the embodiment of the present invention are only preferred embodiments of the present invention, and are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, ordinary technicians in this field should understand that it is still possible to modify the technical solutions recorded in the aforementioned embodiments, or to replace some of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A super-resolution effect evaluation method based on high-resolution measurement data, characterized in that: The method comprises: S1, obtain low-resolution image I L ; S2, obtaining a high-resolution image database, selecting the high-resolution image database, and obtaining resolution measurement data related to the low-resolution image, including: S21, processing the low-resolution image to obtain a resolution range of the low-resolution image, including: S211, analyzing the low-resolution image to obtain a resolution estimation of the low-resolution image; S212, processing the estimated resolution to obtain a resolution range of the low-resolution image; The resolution range of the low-resolution image is L min ~L max ,in Indicates rounding down. represents rounding up, l is the estimated resolution of the low-resolution image; S22, selecting the high-resolution image database according to the resolution range of the low-resolution image to obtain resolution measurement data related to the low-resolution image; S3, processing the resolution measurement data related to the low-resolution image to obtain a ground feature dataset S4, the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ; S5, based on the ground feature dataset The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s ,include: S51, from i=L min Start with the feature dataset The feature data T i As a benchmark, according to the preset ground object navigation information and mesoscale information, in the super-resolution processing result I s Choose from the following and get the value of T i corresponding features; S52, select M operators, where M is an odd number greater than or equal to 3, and i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding features include: S521, if operator j, j=1,2,…,M thinks that in I s Information and datasets of Chinese features The information in is basically the same, so operator j determines that I s Zhong and T i The resolution evaluation result of the corresponding feature is distinguishable, otherwise operator j determines I s Zhong and T i The resolution evaluation result of the corresponding ground feature is indistinguishable; S522, for a certain resolution l1, and T i There are multiple corresponding objects. If more than half of the objects in the resolution evaluation results of M operators can be resolved, the resolution is determined to have reached l1+1. Otherwise, the resolution is determined to have not reached l1+1. The resolution evaluation result is distinguishable or indistinguishable; S53, i=i+1, repeat S51 to S53 until i>L max , get the feature dataset The resolution evaluation results of the corresponding ground features; S54, the ground feature dataset The resolution evaluation result of the corresponding ground object is processed to obtain the super-resolution processing result I s Resolution s .

2. The super-resolution effect evaluation method based on high-resolution measurement data according to claim 1, characterized in that: The resolution measurement data related to the low-resolution image is processed to obtain a ground feature data set. include: S31, from k = L min Initially, according to preset ground feature characteristic requirements, the resolution measurement data related to the low-resolution image is searched to obtain data information of the ground feature to be selected; S32, according to the preset feature selection requirements, search the data information of the feature to be selected to obtain the data subset T corresponding to the resolution k k ; S33, k=k+1, execute S31~S33, select the next resolution level feature data, until k>L max , get the feature dataset Among them, L min ~L max is the resolution range of the low-resolution image, Indicates rounding down. represents rounding up, and l is the estimated resolution of the low-resolution image.

3. The super-resolution effect evaluation method based on high-resolution measurement data according to claim 2, characterized in that: The preset feature selection requirements are: Taking the estimated resolution l of the low-resolution image as a benchmark, min ~L max Select corresponding ground features in a group of 1 meter within the range; Each group selected 3 to 5 features; According to the resolution range L of the low-resolution image min ~L max Select the features; The selected features contain three pictures. The first one is the feature navigation information, which is used to show the location of the feature in a larger area. The second one is the mesoscale feature information, which is used to show the outline of the feature. Assist in determining the location of ground objects; The third image is a detailed image of the ground feature, which is required to show the geometric outline information of the ground feature in detail and annotate the scale data reflecting the resolution characteristics for resolution assessment; Name the feature data images according to the naming conventions.

4. A super-resolution effect evaluation device based on high-resolution measurement data, characterized in that: The device comprises: Low-resolution image acquisition module, used to acquire low-resolution image I L ; The high-resolution image database acquisition module is used to acquire a high-resolution image database, select the high-resolution image database, and obtain resolution measurement data related to the low-resolution image, including: S21, processing the low-resolution image to obtain a resolution range of the low-resolution image, including: S211, analyzing the low-resolution image to obtain a resolution estimation of the low-resolution image; S212, processing the estimated resolution to obtain a resolution range of the low-resolution image; The resolution range of the low-resolution image is L min ~L max ,in Indicates rounding down. represents rounding up, l is the estimated resolution of the low-resolution image; S22, selecting the high-resolution image database according to the resolution range of the low-resolution image to obtain resolution measurement data related to the low-resolution image; A ground object dataset construction module is used to process the resolution measurement data related to the low-resolution image to obtain a ground object dataset A super-resolution processing module is used to process the low-resolution image I L Use the super-resolution algorithm to be evaluated to perform super-resolution reconstruction processing and obtain the super-resolution processing result I s ; A super-resolution effect evaluation module is used to evaluate the effect of the ground object dataset. The super-resolution processing result I s Perform resolution evaluation to obtain the super-resolution processing result I s Resolution s ,include: S51, from i=L min Start with the feature dataset The feature data T i As a benchmark, according to the preset ground object navigation information and mesoscale information, in the super-resolution processing result I s Choose from the following and get the value of T i corresponding features; S52, select M operators, where M is an odd number greater than or equal to 3, and i The corresponding ground features are compared and interpreted to obtain the i The resolution evaluation results of the corresponding features include: S521, if operator j, j=1,2,…,M thinks that in I s Information and datasets of Chinese features The information in is basically the same, so operator j determines that I s Zhong and T i The resolution evaluation result of the corresponding feature is distinguishable, otherwise operator j determines I s Zhong and T i The resolution evaluation result of the corresponding ground feature is indistinguishable; S522, for a certain resolution l1, and T i There are multiple corresponding objects. If more than half of the objects in the resolution evaluation results of M operators can be resolved, the resolution is determined to have reached l1+1. Otherwise, the resolution is determined to have not reached l1+1. The resolution evaluation result is distinguishable or indistinguishable; S53, i=i+1, repeat S51 to S53 until i>L max , get the feature dataset The resolution evaluation results of the corresponding ground features; S54, the ground feature dataset The resolution evaluation result of the corresponding ground object is processed to obtain the super-resolution processing result I s Resolution s .

5. A super-resolution effect evaluation device based on high-resolution measurement data, characterized in that: The device comprises: a memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the super-resolution effect evaluation method based on high-resolution measurement data according to any one of claims 1 to 3.

6. A computer storable medium, characterized in that The computer storable medium stores computer instructions, which, when called, are used to execute the super-resolution effect evaluation method based on high-resolution measurement data according to any one of claims 1 to 3.

Images