CN109242790A - A kind of image deblurring appraisal procedure towards fields of measurement - Google Patents

A kind of image deblurring appraisal procedure towards fields of measurement Download PDF

Info

Publication number
CN109242790A
CN109242790A CN201810957343.5A CN201810957343A CN109242790A CN 109242790 A CN109242790 A CN 109242790A CN 201810957343 A CN201810957343 A CN 201810957343A CN 109242790 A CN109242790 A CN 109242790A
Authority
CN
China
Prior art keywords
image
measurement
deblurring
evaluation
blur
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
CN201810957343.5A
Other languages
Chinese (zh)
Inventor
刘巍
潘翼
李肖
王福吉
贾振元
马建伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian University of Technology
Original Assignee
Dalian University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian University of Technology filed Critical Dalian University of Technology
Priority to CN201810957343.5A priority Critical patent/CN109242790A/en
Publication of CN109242790A publication Critical patent/CN109242790A/en
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/73—Deblurring; Sharpening
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30168—Image quality inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)

Abstract

本发明一种面向测量领域的图像去模糊评估方法属于计算机视觉测量技术领域，涉及一种面向测量领域的图像去模糊评估方法。该方法首先搭建双目视觉测量系统，驱动机床带动用于表征定位信息的标记点高速运动，同时采集序列图像。其次利用曝光时间、帧频已知先验信息求解模糊核，基于非盲复原算法实现图像去模糊。然后建立图像去模糊评估函数，结合精度指标、图像结构数值分析及客观评价方法将误差波动幅度，圆心距约束条件，及图像结构相似度三种评估手段归一化，对图像还原效果进行定性评价，最终实现面向测量领域的图像去模糊评估。该方法实现了对面向测量领域的图像去模糊的有效评估，方法可行性及鲁棒性好。

The invention relates to an image deblurring evaluation method oriented to the measurement field, belonging to the technical field of computer vision measurement, and relates to an image deblurring evaluation method oriented to the measurement field. The method firstly builds a binocular vision measurement system, drives the machine tool to drive the high-speed movement of the marker points used to characterize the positioning information, and collects sequence images at the same time. Secondly, the known prior information of exposure time and frame rate is used to solve the blur kernel, and the image deblurring is realized based on the non-blind restoration algorithm. Then, an image deblurring evaluation function is established, and the three evaluation methods of error fluctuation range, center distance constraint, and image structure similarity are normalized by combining the accuracy index, image structure numerical analysis and objective evaluation method, and the image restoration effect is qualitatively evaluated. , and finally realize the image deblurring evaluation for the measurement field. The method realizes the effective evaluation of image deblurring for the measurement field, and the method has good feasibility and robustness.

Description

A kind of image deblurring appraisal procedure towards fields of measurement
Technical field
The invention belongs to computer vision measurement technical fields, are related to a kind of image deblurring assessment towards fields of measurement Method.
Background technique
With the continuous reduction of hardware device, the fast development of software performance and measurement cost, vision measurement technology is because of tool There are non-contact, dynamic measurement, informative, it is considered to be that realizes live real-time measurement effectively has one of section.Depending on The principle for feeling measurement is to acquire two dimensional image by camera, and the three-dimensional spatial information of measured target is restored by image procossing, But in high dynamic fields of measurement, the image that the high-speed motion of measured target will lead to shooting generates motion blur, extreme influence Positioning measurement precision, scholar furthers investigate for image deblurring problem in recent years, but conventional images restore quality evaluation Means mainly include mean square deviation, signal-to-noise ratio, Y-PSNR, gray scale AVERAGE GRADIENT METHOD WITH etc., however because actual measurement environment is different, The factors such as grain background difference, noise all have larger impact to image restoration appraisal procedure, and above-mentioned evaluation measures are commonly view Feel the improvement in effect or calculate the method for objectively evaluating such as signal noise ratio (snr) of image, fields of measurement can not be applied to, because of the invention one kind Image deblurring appraisal procedure towards fields of measurement is of great significance.
Guangxi Normal University's beam knows the " a kind of to be based on image of the Patent No. CN 106296619A of inventions such as duckweed, Luo Xiaoshu The High-motion picture deblurring method of column gray probability consistency ", this method combination genetic algorithm establish image deblurring mould Type carries out image deblurring and output signal-to-noise ratio using Wiener filtering.But this method assesses image deblurring using signal-to-noise ratio As a result, being only applicable to static or quasi-static image deblurring and effect assessment.The inventions such as the yellowish green pretty young woman of University Of Nanchang, Wu Lushen it is special Benefit number is a kind of " the High-motion picture deblurring method based on image column gray probability consistency " of CN107945127A, should Method carries out image deblurring method based on column gray consistency and uses by being modeled according to camera propulsion blurring process Gray scale mean square error is as evaluation index.Image deblurring is assessed under the conditions of this method is suitable for camera motion, may not apply to Field of high-precision measurement.
Summary of the invention
The invention solves technical problem be to overcome the deficiencies of existing technologies problem, invent a kind of towards fields of measurement Image deblurring appraisal procedure.This method uses two CCD camera measure system, and control lathe drives mark point high-speed motion to adopt Collect sequence image, complete the solution of fuzzy core using known prior information and image high-precision is realized also based on non-blind restoration algorithm Then original proposes image deblurring valuation functions for fields of measurement, in conjunction with the crucial from fluctuating error amplitude, circle of fields of measurement The heart assesses image restoring effect away from three aspects of constraint condition and image structure similarity, and precision index, image is special Property analysis and method for objectively evaluating normalization, the final image deblurring assessment realized towards fields of measurement.This method combines visitor Evaluation method and images themselves characteristic are seen, can be to the moving image deblurring effect qualitative evaluation under fields of measurement, and feasibility It is good.
The technical scheme adopted by the invention is that a kind of image deblurring appraisal procedure towards fields of measurement, feature It is that this method drives the mark point high-speed motion for characterizing location information using two CCD camera measure system driving lathe, together When acquisition sequence image, secondly solve fuzzy core using time for exposure, prior information known to frame frequency, it is real based on non-blind restoration algorithm Existing image deblurring, then establishes image deblurring valuation functions, in conjunction with precision index, picture structure numerical analysis and objective comments Valence method is by fluctuating error amplitude, and three kinds of evaluation measures of distance of center circle constraint condition and image structure similarity normalize, to image Reduction effect carries out qualitative evaluation, the final image deblurring assessment realized towards fields of measurement；Specific step is as follows for method:
Step 1: building vision measurement system acquisition image
Two CCD camera measure system is built first and carries out camera calibration, and upper surface is contained to the measurement mesh of circular markers Mark is fixedly mounted on numerically-controlled machine tool, and two industrial cameras are fixed on above measurement target by profile suspension, in measurement process The track high-speed motion for driving numerically-controlled machine tool to drive measurement target by setting, while camera shooting measurement target is triggered, to obtain Mark point sequence image containing motion blur is simultaneously stored into computer.
Step 2: solving fuzzy core and carry out image deblurring
It is very short due to being influenced the time for exposure by environment and camera frame frequency during actual photographed, and become under certain condition Speed, non-rectilinear and periodic motion are all decomposed into multiple linear uniform motion, therefore the motion blur in the time for exposure can be considered line Property it is fuzzy, fuzzy core may be characterized as a line segment, be determined by two parameters of blur direction and Blur scale, it is of the invention in basis Relationship between adjacent two field pictures estimates accurate fuzzy core as known prior information:
Wherein, f is camera frame frequency, and t is time for exposure, (x1, y1) and (x2, y2) it is to mark null circle in the two field pictures of front and back The pixel coordinate of the heart sets previous frame image Blur scale as θ1, thus ergodic sequence image, which is asked, can acquire blur direction θ2And mould Scale L is pasted, every accurate fuzzy core of frame image can be solved, then deconvoluted using non-blind restoration algorithm, to realize each frame Image deblurring；
Step 3: establishing fluctuating error amplitude function
Since motion blur will affect the extraction accuracy of locating mark points information in image acquisition process, surveyed to reduce vision The measurement error of amount system, the appraisal procedure of image deblurring effect need to consider the influence of precision index, therefore to for measurement Error proposes function Q1:
Wherein, ei, i=(1,2 ..., n) indicates the Slight measurement errors of each frame image tagged point, i.e. the reality of mark point The deviation of border measurement position and theoretical position, E indicate the Slight measurement errors acquired in the n frame image shot by blurred picture blur(ei), i=(1,2 ..., n) and the Slight measurement errors deblur acquired by the clear image that image deblurring is restored (ei), (1,2 ..., set n), two o'clock obtain fluctuating error amplitude function Q in conjunction with expression to i=1, guarantee Q1∈ (0,1), Q1 And bigger, eiIt is smaller, indicate that Slight measurement errors are smaller, image is fuzzy lower, evaluates image deblurring processing for being with this The improvement of unified test accuracy of measurement index.
Step 4: establishing distance of center circle constraint condition function
Picture characteristics is considered in image processing process, according to the relationship under theoretical image between mark point, introduces single width figure As the constraint about circle center distance, function Q is proposed to for distance of center circle constraint condition2:
Wherein, d indicates the distance of center circle on every frame image in circular markers between every two adjacent marker point, the survey of design 7 × 7 circular markers, d are shared in amount targetmeanIt is average value of the d in every frame image.Component Q2It is worth bigger, expression dmean Closer to distance of center circle gauged distance, more accurate for the solution of distance, image blur effects are lower, evaluate image with this and obscure journey Degree and image deblurring treatment effect.
Step 5: establishing image structure similarity function
In the method for objectively evaluating of image deblurring, structural similarity is from the brightness of reference picture and clear image, right It is set out than the similitude between degree and structure to evaluate the fog-level of image, can be applied to evaluation image deblurring effect:
Wherein, μxAnd μyRespectively represent the mean value of two images, i.e. gray component, σxAnd σyPair the variance of image is indicated, i.e., Than degree component, σxyIndicate the covariance between two images, C1And C2Indicate constant.By multiple for fuzzy original image and deblurring Original image compared to the structural similarity of clear image, obtains Q respectively3The comparison of component, i.e. Q3It is worth bigger, expression two images More similar to clear image structure, image deblurring effect is better.
Step 6: establishing image deblurring valuation functions
In conjunction with the image deblurring valuation functions of step 3, step 4 and step 5, by fluctuating error amplitude, distance of center circle constraint Three kinds of evaluation measures normalizeds of condition and image structure similarity can obtain and assess letter for fields of measurement image deblurring Number Q:
Wherein, wiFor the respective weights of each function, which has comprehensively considered vision measurement system measurement accuracy The picture characteristics of index, single image itself is analyzed and to the method for objectively evaluating of picture quality, to image deblurring effect into Row qualitative evaluation, the final image deblurring assessment realized towards fields of measurement, that is, each frame image being directed to, respectively to fuzzy Original image and de-blurred image seek valuation functions Q value, and the value of valuation functions Q is relatively bigger, indicate that image deblurring effect is better.
The invention has the beneficial effects that for image deblurring appraisal procedure is established towards fields of measurement, in conjunction with vision The picture quality of measuring system measurement accuracy index, the structure numerical information for shooting image itself and traditional structural similarity is commented Valence method, by fluctuating error amplitude, the three kinds of evaluation measures normalization of distance of center circle constraint condition and image structure similarity can Qualitative evaluation is carried out to image deblurring effect.This method combination method for objectively evaluating and images themselves characteristic, realize towards The image deblurring of fields of measurement is assessed, so that solving fields of measurement does not have lacking for suitable image restoration quality evaluating method It falls into, method feasibility and robustness are good.
Detailed description of the invention
Fig. 1 is the flow diagram of the image deblurring appraisal procedure of the invention towards fields of measurement.
Fig. 2 is the measurement target image that circular markers are contained in upper surface.Wherein, d indicates circular mark on every frame image Distance of center circle in point between every two adjacent marker point.
Fig. 3 is the valuation functions Q comparative result figure of isogonism spiral trajectory in experiment.Wherein, figure a) feeds for 3m/min Valuation functions Comparative result under speed, figure are b) the valuation functions Comparative result under 5m/min feed speed.
Fig. 4 is the evaluation function Q comparative result figure of circular trace in experiment.Wherein, figure is a) under 3m/min feed speed Valuation functions Comparative result, figure is b) the valuation functions Comparative result under 5m/min feed speed.
Specific embodiment
Describe the specific embodiment of the method for the present invention in detail below in conjunction with technical solution and attached drawing.
The present invention uses two CCD camera measure system, and driving lathe drives the mark point high speed for characterizing location information to transport Dynamic and acquisition sequence image.Fuzzy core is solved using prior information known to time for exposure, frame frequency, is realized using non-blind restoration algorithm Image deblurring.Then in conjunction with precision index, picture characteristics analysis and method for objectively evaluating by fluctuating error amplitude, distance of center circle is about Three kinds of evaluation measures normalization of beam condition and image structure similarity, to establish image deblurring valuation functions.To image restoring Effect carries out qualitative evaluation, the final image deblurring assessment realized towards fields of measurement.Fig. 1 is the present invention towards fields of measurement Image deblurring appraisal procedure flow diagram, specific step is as follows for this method:
Step 1: building vision measurement system acquisition image
Two CCD camera measure system is built first, and embodiment builds experiment measurement system using two high resolution industrial cameras System, and carry out camera calibration.There is the measurement target of 7*7 circular markers to be fixedly mounted on numerically-controlled machine tool upper surface photoetching On, two industrial cameras are fixed on above measurement target by profile suspension, and the binocular camera time for exposure is set in measurement process And frame frequency is 20ms and 25fps, driving numerically-controlled machine tool drives measurement target to press respectively with the feed speed of 3m/min and 5m/min The track of setting is run, while triggering camera shooting measurement target acquisition sequence image, obtains the mark point containing motion blur Sequence image is simultaneously stored into computer；
Step 2: solving fuzzy core and carry out image deblurring
Since the motion blur in the time for exposure during actual photographed can be considered Linear Fuzzy, that is, solve each frame image Blur direction and Blur scale, according to the relationship between adjacent two field pictures as Given information, the combining camera time for exposure, Prior information known to frame frequency can solve accurate fuzzy core using formula (2), and being deconvoluted based on non-blind restoration algorithm, it is clear to restore Image traverses each frame image to realize each frame image deblurring；
Step 3: establishing fluctuating error amplitude function
In view of in image acquisition process since motion blur will affect the extraction accuracy of locating mark points information, for reduce The measurement error of vision measurement system, the appraisal procedure of image deblurring effect need to consider the influence of precision index, therefore to needle Valuation functions Q is proposed to measurement error1, as shown in formula (2), work as Q1And it is bigger, Slight measurement errors are smaller, image deblurring Effect is better；
Step 4: establishing distance of center circle constraint condition function
In conjunction with the structure numerical information of shooting image itself, i.e. relationship under theoretical image between mark point, single width is introduced Constraint of the image about circle center distance, as shown in Figure 2: 7 × 7 circular markers are shared in the measurement target of design, d indicates every Distance of center circle on frame image in circular markers between every two adjacent marker point, therefore proposed to for distance of center circle constraint condition Valuation functions Q2, as shown in formula (3), work as Q2It is worth bigger, expression dmeanCloser to distance of center circle gauged distance, i.e., for distance It is more accurate to solve, and image deblurring treatment effect is better；
Step 5: establishing image structure similarity function
According to the image quality evaluating method of traditional structural similarity, structural similarity is as evaluation index, from reference Similitude between the brightness of image and clear image, contrast and structure sets out to evaluate the fog-level of image, can answer Valuation functions Q is proposed for evaluating image deblurring effect, therefore to for image structure similarity3, as shown in formula (4), By, respectively compared to the structural similarity of clear image, obtaining Q for fuzzy original image and deblurring restored image3The knot of numerical value Fruit is compared, and Q is worked as3Value is bigger, indicates that image is more similar to clear image structure, image deblurring effect is better；
Step 6: establishing image deblurring valuation functions
In conjunction with three image deblurring valuation functions of step 3, step 4 and step 5, by fluctuating error amplitude, distance of center circle Three kinds of evaluation measures normalizeds of constraint condition and image structure similarity can be obtained and be commented for fields of measurement image deblurring Estimate function Q, as shown in formula (5), using qualitative evaluation can be carried out to image deblurring effect, realizes the figure towards fields of measurement As deblurring is assessed；
The weight chosen in experiment is respectively w1=0.5, w2=0.3 and w3=0.2, respectively using valuation functions Q to The trace image deblurring effect of isogonism spiral trajectory is assessed under 3m/min, 5m/min feed speed, i.e., for each Frame image seeks valuation functions Q value to fuzzy original image and de-blurred image respectively, as shown in Figure 3: horizontal axis indicates the sequence of shooting Image, the longitudinal axis indicate the calculated value of valuation functions Q, and wherein it is clear to represent original image, de-blurred image and theory by evaluation function value Q The degree of closeness of clear image.Fig. 3 a), b) in the Q value of de-blurred image be all larger than the Q value of original image, the two takes difference energy Qualitative evaluation deblurring effect illustrates the image deblurring appraisal procedure proposed by the present invention towards fields of measurement, can be right Image deblurring effect carries out qualitative evaluation, and method robustness is good.
The track of above-mentioned experimental image movement is the continually changing equiangular helical spiral of curvature, removes mould to verify image of the present invention Valuation functions Q is pasted to the Evaluated effect for acquiring image under the conditions of constant curvature, controls numerically-controlled machine tool respectively with 3m/min and 5m/ The feed speed operating radius of min is the circular trace of 50mm, selects experiment parameter identical with equiangular spiral line tracking, respectively Valuation functions Q value is sought to fuzzy original image and de-blurred image, as shown in Figure 4.By Fig. 4 a), b) it can be seen from de-blurred image Q value be all larger than the Q value of fuzzy original image, i.e. this method can effectively assess image deblurring effect.
This method do not have a suitable image restoration quality evaluation means for fields of measurement, proposition towards fields of measurement Image deblurring evaluation function, according to the measuring precision index, picture structure numerical information and traditional images quality evaluation side Method, by fluctuating error amplitude, opposite is realized in the three kinds of evaluation measures normalization of distance of center circle constraint condition and image structure similarity To effective assessment of the image deblurring of fields of measurement, method feasibility and robustness are good.

Claims (1)

1.一种面向测量领域的图像去模糊评估方法，其特征是，该方法采用双目视觉测量系统驱动机床带动用于表征定位信息的标记点高速运动，采集序列图像；利用曝光时间、帧频已知先验信息求解模糊核，基于非盲复原算法实现图像去模糊；然后建立图像去模糊评估函数，结合精度指标、图像结构数值分析及客观评价方法将误差波动幅度，圆心距约束条件，及图像结构相似度三种评估手段归一化，对图像还原效果进行定性评价，最终实现面向测量领域的图像去模糊评估；方法的具体步骤如下：1. An image deblurring evaluation method oriented to the measurement field is characterized in that, the method adopts a binocular vision measurement system to drive the machine tool to drive the high-speed movement of the marking points used to characterize the positioning information, and collect sequence images; The prior information is known to solve the blur kernel, and the image deblurring is realized based on the non-blind restoration algorithm; then the image deblurring evaluation function is established, and the error fluctuation range, the center distance constraints, and The three evaluation methods of image structure similarity are normalized, and the image restoration effect is qualitatively evaluated, and finally the image deblurring evaluation for the measurement field is realized. The specific steps of the method are as follows: 步骤1：搭建视觉测量系统采集图像Step 1: Build a visual measurement system to collect images 首先搭建双目视觉测量系统并进行相机标定，将上表面含有圆形标记点的测量目标固定安装在数控机床上，两个工业相机通过型材悬挂固定在测量目标上方，测量过程中驱动数控机床带动测量目标按设定的轨迹高速运动，同时触发相机拍摄测量目标，来得到含有运动模糊的标记点序列图像并储存到计算机；Firstly, a binocular vision measurement system is built and the camera is calibrated. The measurement target with circular marks on the upper surface is fixedly installed on the CNC machine tool. Two industrial cameras are suspended and fixed above the measurement target through the profile. During the measurement process, the CNC machine tool is driven to drive The measurement target moves at a high speed according to the set trajectory, and at the same time triggers the camera to shoot the measurement target to obtain a sequence image of marker points with motion blur and store it in the computer; 步骤2：求解模糊核进行图像去模糊Step 2: Solve the blur kernel for image deblurring 由于实际拍摄过程中受环境及相机帧频的影响曝光时间很短，且在一定条件下变速、非直线及周期运动都分解为多个匀速直线运动，因此曝光时间内的运动模糊可视为线性模糊，其模糊核可表征为一条线段，由模糊方向和模糊尺度两个参数确定，本发明中根据相邻两帧图像之间的关系作为已知的先验信息，估计精确的模糊核：Due to the influence of the environment and the camera frame rate in the actual shooting process, the exposure time is very short, and under certain conditions, variable speed, non-linear and periodic motions are decomposed into multiple uniform linear motions, so the motion blur during the exposure time can be regarded as linear. Blur, its blur kernel can be represented as a line segment, which is determined by two parameters: blur direction and blur scale. In the present invention, the relationship between two adjacent frames of images is used as known prior information to estimate the accurate blur kernel: 其中，f为相机帧频，t是曝光时间，(x1，y1)和(x2，y2)为前后两帧图像中标记点圆心的像素坐标，设定前一帧图像模糊尺度为θ1，由此遍历序列图像求可求得模糊方向θ2及模糊尺度L，即可求解每帧图像精确的模糊核，然后利用非盲复原算法去卷积，来实现每一帧图像去模糊；Among them, f is the camera frame rate, t is the exposure time, (x 1 , y 1 ) and (x 2 , y 2 ) are the pixel coordinates of the center of the marked point in the two frames of images before and after, and set the blur scale of the previous frame as θ 1 , the blur direction θ 2 and the blur scale L can be obtained by traversing the sequence images, and the precise blur kernel of each frame of image can be solved, and then the non-blind restoration algorithm is used to deconvolute the image to achieve de-blurring of each frame of image. ; 步骤3：建立误差波动幅度函数Step 3: Build the Error Fluctuation Magnitude Function 图像采集过程中由于运动模糊会影响标记点定位信息的提取精度，为降低视觉测量系统的测量误差，图像去模糊效果的评估方法需考虑精度指标的影响，因此对针对测量误差提出了函数Q1：In the process of image acquisition, motion blur will affect the extraction accuracy of marker location information. In order to reduce the measurement error of the visual measurement system, the evaluation method of the image deblurring effect needs to consider the influence of the accuracy index. Therefore, a function Q 1 is proposed for the measurement error. : 其中，ei,i＝(1,2,...,n)表示每一帧图像标记点的视觉测量误差，即标记点的实际测量位置与理论位置的偏差，E表示拍摄的n帧图像中由模糊图像求得的视觉测量误差blur(ei),i＝(1,2,...,n)和由图像去模糊复原的清晰图像求得的视觉测量误差deblur(ei),i＝(1,2,...,n)的集合，两点结合表示得到误差波动幅度函数Q1，保证Q1∈(0,1)，Q1且越大，ei越小，表示视觉测量误差越小，图像模糊越低，以此评价图像去模糊处理对于系统测量精度指标的改善效果；Among them, e i ,i=(1,2,...,n) represents the visual measurement error of the marked point in each frame of image, that is, the deviation between the actual measured position of the marked point and the theoretical position, and E represents the captured n frames of images The visual measurement error blur(e i ), i=(1,2,...,n) obtained from the blurred image and the visual measurement error deblur( ei ) obtained from the clear image restored from the image deblurring The set of i=(1,2,...,n), the combination of two points indicates that the error fluctuation range function Q 1 is obtained, which ensures that Q 1 ∈(0,1), and the larger Q 1 is, the smaller e i is, indicating that The smaller the visual measurement error is, the lower the image blur is, so as to evaluate the improvement effect of image deblurring on the system measurement accuracy index; 步骤4：建立圆心距约束条件函数Step 4: Establish the center-to-center distance constraint function 图像处理过程中考虑图像特性，根据理论图像下标记点之间的关系，引入单幅图像关于圆心距离的约束，因此对针对圆心距约束条件提出函数Q2：In the process of image processing, the characteristics of the image are considered. According to the relationship between the marked points under the theoretical image, the constraint of the distance between the center of a single image is introduced. Therefore, a function Q 2 is proposed for the constraint condition of the distance between the center of the circle: 其中，d表示每帧图像上圆形标记点中每两个相邻标记点之间的圆心距，设计的测量目标上共有7×7个圆形标记点，dmean是d在每帧图像中的平均值；分量Q2值越大，表示dmean越接近圆心距标准距离，对于距离的求解越准确，图像模糊效果越低，以此评价图像模糊程度及图像去模糊处理效果；Among them, d represents the center-to-center distance between each two adjacent marker points in the circular marker points on each frame of image, there are 7×7 circular markers on the designed measurement target, and d mean is d in each frame of image The larger the value of the component Q2 , the closer the d mean is to the standard distance from the center of the circle, the more accurate the solution of the distance, and the lower the image blurring effect, so as to evaluate the image blurring degree and image deblurring processing effect; 步骤5：建立图像结构相似度函数Step 5: Establish image structure similarity function 在图像去模糊的客观评价方法中，结构相似度从参考图像与清晰图像的亮度、对比度以及结构之间的相似性出发来评价图像的模糊程度，可应用于评价图像去模糊效果：In the objective evaluation method of image deblurring, the structural similarity evaluates the blur degree of the image from the similarity between the brightness, contrast and structure of the reference image and the clear image, and can be applied to evaluate the image deblurring effect: 其中，μx和μy分别代表两幅图像的均值，即灰度分量，σx和σy表示图像的方差，即对比度分量，σxy表示两幅图像之间的协方差，C1和C2表示常量；通过对于模糊原图和去模糊复原图像分别相比于清晰图像的结构相似度，得到Q3分量的比较，即Q3值越大，表示两幅图像与清晰图像结构越相似，图像去模糊效果越好；Among them, μ x and μ y represent the mean value of the two images, namely the gray component, σ x and σ y represent the variance of the image, that is, the contrast component, σ xy represents the covariance between the two images, C 1 and C 2 represents a constant; by comparing the structural similarity of the blurred original image and the deblurred restored image to the clear image, the comparison of the Q3 component is obtained, that is, the larger the Q3 value, the more similar the two images are to the clear image in structure. The better the image deblurring effect; 步骤6：建立图像去模糊评估函数Step 6: Build Image Deblurring Evaluation Function 结合步骤3、步骤4及步骤5的图像去模糊评估函数，将误差波动幅度，圆心距约束条件，及图像结构相似度三种评估手段归一化处理，可得针对测量领域图像去模糊评估函数Q：Combined with the image deblurring evaluation functions of step 3, step 4 and step 5, the three evaluation methods of error fluctuation range, center distance constraint, and image structure similarity are normalized, and the image deblurring evaluation function for the measurement field can be obtained. Q: 其中，wi为每个函数的对应权重，该评价函数综合考虑了视觉测量系统测量精度指标、单幅图像自身的图像特性分析及对图像质量的客观评价方法，对图像去模糊效果进行定性评价，最终实现面向测量领域的图像去模糊评估，即针对的每一帧图像，分别对模糊原图及去模糊图像求取评估函数Q值，评估函数Q的值相对越大，表示图像去模糊效果越好。Among them, w i is the corresponding weight of each function. This evaluation function comprehensively considers the measurement accuracy index of the visual measurement system, the image characteristic analysis of a single image itself, and the objective evaluation method for the image quality, and qualitatively evaluates the image deblurring effect. , and finally realize the image deblurring evaluation for the measurement field, that is, for each frame of image, the evaluation function Q value is obtained for the blurred original image and the deblurred image respectively. The larger the value of the evaluation function Q is, the better the image deblurring effect is. the better.
CN201810957343.5A 2018-08-22 2018-08-22 A kind of image deblurring appraisal procedure towards fields of measurement Withdrawn CN109242790A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810957343.5A CN109242790A (en) 2018-08-22 2018-08-22 A kind of image deblurring appraisal procedure towards fields of measurement

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810957343.5A CN109242790A (en) 2018-08-22 2018-08-22 A kind of image deblurring appraisal procedure towards fields of measurement

Publications (1)

Publication Number Publication Date
CN109242790A true CN109242790A (en) 2019-01-18

Family

ID=65071081

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810957343.5A Withdrawn CN109242790A (en) 2018-08-22 2018-08-22 A kind of image deblurring appraisal procedure towards fields of measurement

Country Status (1)

Country Link
CN (1) CN109242790A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111626946A (en) * 2020-04-23 2020-09-04 武汉理工大学 Motion blur kernel measurement method for high-speed material conveying visual inspection system
CN113327206A (en) * 2021-06-03 2021-08-31 江苏电百达智能科技有限公司 Image fuzzy processing method of intelligent power transmission line inspection system based on artificial intelligence

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408707A (en) * 2014-10-28 2015-03-11 哈尔滨工业大学 Rapid digital imaging fuzzy identification and restored image quality assessment method
CN106296619A (en) * 2016-10-08 2017-01-04 广西师范大学 A kind of based on genetic algorithm with the image deblurring method of Wiener filtering
CN107945127A (en) * 2017-11-27 2018-04-20 南昌大学 A kind of High-motion picture deblurring method based on image column gray probability uniformity

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104408707A (en) * 2014-10-28 2015-03-11 哈尔滨工业大学 Rapid digital imaging fuzzy identification and restored image quality assessment method
CN106296619A (en) * 2016-10-08 2017-01-04 广西师范大学 A kind of based on genetic algorithm with the image deblurring method of Wiener filtering
CN107945127A (en) * 2017-11-27 2018-04-20 南昌大学 A kind of High-motion picture deblurring method based on image column gray probability uniformity

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
刘巍等: "《机床动态检测中的高速图像运动去模糊还原》", 《仪器仪表学报》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111626946A (en) * 2020-04-23 2020-09-04 武汉理工大学 Motion blur kernel measurement method for high-speed material conveying visual inspection system
CN113327206A (en) * 2021-06-03 2021-08-31 江苏电百达智能科技有限公司 Image fuzzy processing method of intelligent power transmission line inspection system based on artificial intelligence
CN113327206B (en) * 2021-06-03 2022-03-22 江苏电百达智能科技有限公司 Image fuzzy processing method of intelligent power transmission line inspection system based on artificial intelligence

Similar Documents

Publication Publication Date Title
CN109308693B (en) Single-binocular vision system for target detection and pose measurement constructed by one PTZ camera
DE102015005267B4 (en) Information processing apparatus, method therefor and measuring apparatus
EP2757346B1 (en) Method for measuring a height profile of a vehicle passing on the street
CN106127145B (en) Pupil diameter and tracking
CN108198224B (en) Linear array camera calibration device and calibration method for stereoscopic vision measurement
CN106863014A (en) A kind of five-axle number control machine tool linear axis geometric error detection method
DE112011100652T5 (en) THREE-DIMENSIONAL MEASURING DEVICE, PROCESSING METHOD AND NON-VOLATILE COMPUTER-READABLE STORAGE MEDIUM
CN1743806A (en) Moving-object height determining apparatus
CN113324478A (en) Center extraction method of line structured light and three-dimensional measurement method of forge piece
DE102018108027A1 (en) Object detecting device
CN111524238B (en) Three-dimensional point cloud deformation method based on coding point driving
DE102020007045A1 (en) Object detection system with transit time sensor
CN113011401B (en) Face image posture estimation and correction method, system, medium and electronic equipment
CN110260795B (en) An absolute displacement detection method based on incremental absolute grating ruler
CN114152208B (en) DIC technology-based 4D printing flexible skin deformation efficiency evaluation method
CN106846372A (en) Human motion quality visual A+E system and method
CN109242790A (en) A kind of image deblurring appraisal procedure towards fields of measurement
US20240257304A1 (en) Tactile pattern super resolution reconstruction method and acquisition system
Tjaden et al. High-speed and robust monocular tracking
CN107423766B (en) A method for detecting the movement position and attitude of the end of a hybrid automobile electrophoresis coating conveying mechanism
CN118429218A (en) Flight test system
CN116160458B (en) Multi-sensor fusion rapid positioning method, equipment and system for mobile robot
DE69331360T2 (en) Image processing method and apparatus
CN113674360A (en) A Covariance-Based Calibration Method of Line Structured Light Light Plane
Liu et al. Outdoor camera calibration method for a GPS & camera based surveillance system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20190118