CN102663764A - Image quality evaluation method based on structural distortion and spatial frequency index - Google Patents
Image quality evaluation method based on structural distortion and spatial frequency index Download PDFInfo
- Publication number
- CN102663764A CN102663764A CN2012101254006A CN201210125400A CN102663764A CN 102663764 A CN102663764 A CN 102663764A CN 2012101254006 A CN2012101254006 A CN 2012101254006A CN 201210125400 A CN201210125400 A CN 201210125400A CN 102663764 A CN102663764 A CN 102663764A
- Authority
- CN
- China
- Prior art keywords
- image
- sigma
- distorted
- image quality
- spatial
- 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.)
- Pending
Links
- 238000013441 quality evaluation Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 37
- 238000004364 calculation method Methods 0.000 abstract description 4
- 230000008447 perception Effects 0.000 abstract description 3
- 238000011156 evaluation Methods 0.000 description 15
- 238000012360 testing method Methods 0.000 description 11
- 230000000007 visual effect Effects 0.000 description 9
- 238000001303 quality assessment method Methods 0.000 description 7
- 230000015556 catabolic process Effects 0.000 description 4
- 238000006731 degradation reaction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000035945 sensitivity Effects 0.000 description 4
- 238000009792 diffusion process Methods 0.000 description 3
- 238000013178 mathematical model Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000007423 decrease Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000004438 eyesight Effects 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 1
- 238000013332 literature search Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 102220203103 rs1057521344 Human genes 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
- Image Processing (AREA)
Abstract
本发明公开了一种基于结构失真与空间频率指标的图像质量评价方法,该方法在基于结构失真的图像质量评价方法的基础上,提出结构失真与空间频率指标结合的图像质量评价方法,从而实现了从影响图像质量评价的主要因素即亮度、清晰度以及相关度对图像质量进行评价。本发明基于结构失真与空间频率指标的图像质量评价方法在客观评价图像时更能与人眼主观感觉相一致,并且由于所选用的计算关系式较为简单,只涉及到原始图像与失真图像像素之间的计算,所以该方法能够快速有效的对图像质量进行评价。
The invention discloses an image quality evaluation method based on structural distortion and spatial frequency index. On the basis of the image quality evaluation method based on structural distortion, the method proposes an image quality evaluation method combining structural distortion and spatial frequency index, thereby realizing This paper evaluates the image quality from the main factors that affect the image quality evaluation, namely brightness, sharpness and correlation. The image quality evaluation method based on structural distortion and spatial frequency index of the present invention can be more consistent with the subjective perception of human eyes when evaluating images objectively, and because the selected calculation relation is relatively simple, it only involves the relationship between the original image and the distorted image pixels. Therefore, this method can quickly and effectively evaluate the image quality.
Description
技术领域 technical field
本发明属于图像质量评价领域,特别是涉及一种基于图像的亮度、清晰度和相关度的客观图像质量评价方法。The invention belongs to the field of image quality evaluation, in particular to an objective image quality evaluation method based on image brightness, definition and correlation.
背景技术 Background technique
图像质量评价是图像处理系统的关键技术之一,对图像的质量进行评价是一种复杂的心理活动,有很多因素影响到对图像质量好坏的判断。如果将这些因素分别提取并应用于对图像的客观评价中将能很好的对图像质量进行判定。这些因素主要有:①亮度,适当的亮度是人观察图像的基本条件,亮度过强或者过弱都会引起图像质量的下降;②清晰度,人们常用模糊,清晰等词汇来描述图像的质量。从频域角度看,只有当一幅图像的高低频信息比例适当时会有清晰的感觉;③相关度,即图像之间的相似程度。Image quality evaluation is one of the key technologies of image processing system. It is a complex psychological activity to evaluate image quality, and many factors affect the judgment of image quality. If these factors are extracted and applied to the objective evaluation of the image, the image quality can be well judged. These factors mainly include: ①Brightness. Proper brightness is the basic condition for people to observe images. If the brightness is too strong or too weak, it will cause the image quality to decline; ②Sharpness. People often use words such as blur and clarity to describe the quality of images. From the perspective of frequency domain, only when the ratio of high and low frequency information of an image is appropriate, there will be a clear feeling; ③ Correlation, that is, the degree of similarity between images.
图像质量评价方法可分为主观质量评价方法和客观质量评价方法两类。主观质量评价方法是指让观察者根据一些事先规定好的评价尺度或自己已有的经验,对待测试图像按视觉或主观印象提出质量判断,并给出质量分值。最常用的主观质量评价方法有主观平均分(MOS),该方法已经应用多年。然而,MOS方法往往受到观察者本身因素的影响,而且进行视觉心理测试经常需要花费的时间较长,对观察环境有一定的限制。因此,目前在图像质量评价中主要使用客观质量评价方法。这几乎在所有有关图像质量评价的文献和研究论文中均有论述。图像的客观质量评价方法是指用再现图像偏离原始图像的误差来衡量再现图像的质量。其主要是应用数学模型来表示视觉对图像的主观感受。数学模型的应用使得图像的客观质量评价具有快速、稳定、易于被量化的特点,传统的图像客观质量评价通常基于与标准图像的灰度差异越大质量退化越严重的思想,具有代表性的方法有均方误差(MSE)、峰值信噪比(PSNR)、图像清晰度、信息熵(H)等。这些常用的客观质量评价方法虽然看起来简单直观,数学表达严格,但是其评价结果往往与人的主观感觉不一致。主要原因在于PSNR值和MSE值都是从信息数据的失真出发,并没有考虑人眼对同样的信息数据失真程度有着不一样的视觉感受,也没有考虑信息数据的空间位置关系。Image quality assessment methods can be divided into two categories: subjective quality assessment methods and objective quality assessment methods. The subjective quality evaluation method refers to allowing the observer to judge the quality of the test image according to the visual or subjective impression according to some predetermined evaluation scales or their own experience, and give a quality score. The most commonly used subjective quality evaluation method is subjective mean score (MOS), which has been used for many years. However, the MOS method is often affected by the observer's own factors, and it often takes a long time to perform a visual psychological test, which has certain restrictions on the observation environment. Therefore, objective quality evaluation methods are mainly used in image quality evaluation at present. This is discussed in almost all literature and research papers on image quality evaluation. The objective quality evaluation method of the image refers to measuring the quality of the reproduced image by the deviation of the reproduced image from the original image. It mainly uses mathematical models to represent the subjective perception of vision on images. The application of mathematical models makes the objective quality evaluation of images fast, stable, and easy to be quantified. Traditional image objective quality evaluation is usually based on the idea that the greater the grayscale difference from the standard image, the more serious the quality degradation. This is a representative method. There are mean square error (MSE), peak signal-to-noise ratio (PSNR), image clarity, information entropy (H) and so on. Although these commonly used objective quality evaluation methods seem simple and intuitive, and their mathematical expressions are strict, their evaluation results are often inconsistent with people's subjective feelings. The main reason is that both the PSNR value and the MSE value start from the distortion of the information data, and do not take into account that the human eye has different visual perceptions of the same degree of distortion of the information data, nor do they consider the spatial position relationship of the information data.
Zhou Wang等学者提出了一种基于结构失真的图像质量评价方法。该图像质量评价指标将图像质量降低模拟为结构失真而不是误差,通过计算原始图像和失真图像结构的相似度来评价图像质量。该方法是一个简单有效的质量评价算法,这种评价模型的优点是对图像处理过程中的相关性、亮度失真和反差等因素进行分析研究,而这些因素正是人们评价图像输出质量的重要依据。结构失真的图像质量评价指标Q的动态范围为[-1,1]。只有当i=1,2,...,N都有yi=xi时,Q才能达到最佳值1。然而,基于结构失真的图像质量评价模型在评价失真图像质量时,也有一定的缺陷性。其主要原因在于该模型没有考虑人眼视觉系统中的对比敏感度对图像评价的影响。人眼视觉系统的对比敏感度主要反映在图像的清晰度等方面。Scholars such as Zhou Wang proposed an image quality evaluation method based on structural distortion. The image quality evaluation index simulates the degradation of image quality as structural distortion rather than error, and evaluates image quality by calculating the similarity between the original image and the distorted image structure. This method is a simple and effective quality evaluation algorithm. The advantage of this evaluation model is to analyze and study factors such as correlation, brightness distortion and contrast in the process of image processing, and these factors are the important basis for people to evaluate image output quality. . The dynamic range of the image quality evaluation index Q of structural distortion is [-1, 1]. Only when i =1, 2 , . However, the image quality evaluation model based on structural distortion also has certain defects when evaluating the quality of distorted images. The main reason is that the model does not consider the impact of contrast sensitivity in the human visual system on image evaluation. The contrast sensitivity of the human visual system is mainly reflected in the clarity of the image and so on.
图像的清晰度是指刻画图像特征的清晰程度,也就是特征(边缘)与其背景区域的对比度。清晰度是用来反映图像边缘或点特征的扩散程度的度量,扩散程度越大,清晰度越低,扩散程度越小,清晰度越高。定量评价清晰度的指标较多。空间频率指标通过计算空间行频率和空间列频率,综合两者得到整体空间频率。空间频率指标反映评价图像空间的总体活跃程度,可以有效地反映图像的清晰度。但是,对于引进噪声的失真图像,并非图像清晰度值越高图像的主观视觉越好,原因在于图像的清晰度是对图像本身像素之间的计算,它并没有考虑到失真图像与原始图像的相关性。The sharpness of an image refers to the degree of clarity that characterizes image features, that is, the contrast between features (edges) and their background regions. Sharpness is a measure used to reflect the degree of diffusion of image edge or point features. The larger the degree of diffusion, the lower the definition, and the smaller the degree of diffusion, the higher the definition. There are many indicators to quantitatively evaluate clarity. The spatial frequency index calculates the spatial row frequency and spatial column frequency, and integrates the two to obtain the overall spatial frequency. The spatial frequency index reflects the overall activity of the evaluation image space, which can effectively reflect the clarity of the image. However, for a distorted image that introduces noise, it does not mean that the higher the image clarity value, the better the subjective vision of the image. The reason is that the image clarity is calculated between the pixels of the image itself, and it does not take into account the difference between the distorted image and the original image. Correlation.
发明内容 Contents of the invention
为了克服上述现有技术的不足,本发明提供一种基于结构失真与空间频率指标的图像质量评价方法。In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides an image quality evaluation method based on structural distortion and spatial frequency indexes.
本发明的技术方案为一种基于结构失真与空间频率指标的图像质量评价方法,其特征是在于,包括以下步骤:The technical solution of the present invention is an image quality evaluation method based on structural distortion and spatial frequency index, which is characterized in that it includes the following steps:
步骤1,计算原始图像与失真图像的均值利用关系式计算原始图像与失真图像的平均亮度逼近度,Step 1, calculate the mean of the original image and the distorted image Use relational Calculate the average brightness approximation of the original image and the distorted image,
其中,x,y分别代表原始图像与失真图像,图像中横向和纵向的像素数目分别为m和n,Among them, x and y represent the original image and the distorted image respectively, and the number of horizontal and vertical pixels in the image are m and n respectively,
x(i,j)和y(i,j)分别为原始图像和失真图像在第i行第j列的灰度值;x(i, j) and y(i, j) are the gray values of the original image and the distorted image in row i, column j, respectively;
步骤2,计算原始图像与失真图像的均方误差σx,σy及协方差σxy,利用关系式计算原始图像与失真图像的线性相关度,Step 2, calculate the mean square error σ x , σ y and covariance σ xy of the original image and the distorted image, using the relation Calculate the linear correlation between the original image and the distorted image,
其中,x,y分别代表原始图像与失真图像,图像在横向和纵向的像素数目分别为m和n,Among them, x and y represent the original image and the distorted image respectively, and the number of pixels in the horizontal and vertical directions of the image are m and n respectively,
x(i,j)和y(i,j)分别为原始图像和失真图像在第i行第j列的灰度值;x(i, j) and y(i, j) are the gray values of the original image and the distorted image in row i, column j, respectively;
步骤3,根据步骤2所得原始图像与失真图像的均方误差σx,σy,利用关系式计算原始图像与失真图像的相似度;Step 3, according to the mean square error σ x , σ y of the original image and the distorted image obtained in step 2, use the relation Calculate the similarity between the original image and the distorted image;
步骤4,计算失真图像的空间行频率RF以及空间列频率CF,通过空间行频率RF及空间列频率CF计算失真图像空间频率指标计算空间行频率RF和空间列频率CF的公式如下Step 4, calculate the spatial row frequency RF and spatial column frequency CF of the distorted image, and calculate the spatial frequency index of the distorted image through the spatial row frequency RF and spatial column frequency CF The formulas for calculating the spatial row frequency RF and the spatial column frequency CF are as follows
其中,m,n分别为图像在横向和纵向的像素数目,y(i,j)为失真图像在第i行第j列的灰度值;Wherein, m, n are respectively the number of pixels of the image in horizontal and vertical directions, and y(i, j) is the gray value of the distorted image in row i, column j;
步骤5,综合步骤1所得平均亮度逼近度、步骤2所得线性相关度、步骤3所得相似度及步骤4所得失真图像空间频率指标,计算基于结构失真与空间频率指标的图像质量评价指标根据计算所得图像质量评价指标数值从图像的亮度、清晰度以及相关性三方面对图像质量进行综合评价。Step 5, combining the average brightness approximation degree obtained in step 1, the linear correlation degree obtained in step 2, the similarity degree obtained in step 3 and the spatial frequency index of the distorted image obtained in step 4, and calculating the image quality evaluation index based on structural distortion and spatial frequency index According to the calculated image quality evaluation index value, the image quality is comprehensively evaluated from the three aspects of image brightness, clarity and correlation.
本发明所提供基于结构失真的图像质量评价方法将图像质量降低模拟为结构失真而不是误差。本发明在基于结构失真的图像质量评价方法的基础上,提出结合空间频率指标与结构失真的图像质量评价方法,从而实现了从影响图像质量评价的主要因素即亮度、清晰度以及相关度对图像质量进行评价。实验结果表明,基于结构失真与空间频率指标的图像质量评价方法在客观评价图像时更能与人眼主观感觉相一致。并且由于所选用的计算关系式较为简单,只涉及到原始图像与失真图像像素之间的计算,所以该算法能够快速有效的对图像质量进行评价。The image quality evaluation method based on structural distortion provided by the present invention simulates image quality degradation as structural distortion rather than error. On the basis of the image quality evaluation method based on structural distortion, the present invention proposes an image quality evaluation method combining spatial frequency index and structural distortion, thereby realizing the evaluation of image quality from the main factors that affect image quality evaluation, namely brightness, clarity and correlation. Quality is evaluated. Experimental results show that the image quality evaluation method based on structural distortion and spatial frequency index is more consistent with the subjective perception of human eyes when evaluating images objectively. And because the calculation relation selected is relatively simple, which only involves the calculation between the original image and the distorted image pixels, so the algorithm can quickly and effectively evaluate the image quality.
附图说明 Description of drawings
图1为本发明实施例的流程图。Fig. 1 is a flowchart of an embodiment of the present invention.
图2为本发明实施例的试验所用Lena测试图像,其中图2-1为原始图像,图2-2为加入均匀分布随机噪声的图像,图2-3为加入高斯噪声的图像,图2-4为模糊处理的图像,图2-5为压缩处理的图像。Fig. 2 is the used Lena test image of the experiment of the embodiment of the present invention, wherein Fig. 2-1 is original image, Fig. 2-2 is the image that adds uniformly distributed random noise, Fig. 2-3 is the image that adds Gaussian noise, Fig. 2- 4 is the blurred image, and Fig. 2-5 is the compressed image.
具体实施方式 Detailed ways
传统的图像质量评价方法可以分为主观图像质量评价方法和客观图像质量评价方法两大类。传统的客观图像质量评价方法,如均方误差(MSE)、峰值信噪比(PSNR)等方法通过每个像素与原像素的差值来判别图像质量的优劣,虽然数学表达简单、计算方便,但是其结果并不总是与人的主观感受相一致。而主观评价方法,如主观质量评分法(MOS)通过对观察者的评分归一化来判断图像质量,无法应用数学模型进行描述,在工程应用中费时费力,很难在实际应用中采用。Traditional image quality assessment methods can be divided into two categories: subjective image quality assessment methods and objective image quality assessment methods. Traditional objective image quality evaluation methods, such as mean square error (MSE), peak signal-to-noise ratio (PSNR), etc., use the difference between each pixel and the original pixel to judge the quality of the image, although the mathematical expression is simple and the calculation is convenient , but the results are not always consistent with people's subjective feelings. However, subjective evaluation methods, such as the subjective quality score method (MOS), judge the image quality by normalizing the observer's score, which cannot be described by a mathematical model.
人对图像质量评价是一种复杂的心理活动,有许多相关因素影响到人对图像质量好坏的判断。如果将这些因素分别提取最后加以综合将能很好地用于图像质量的判定。这些因素主要有:①亮度,适当的亮度是人观察图像的基本条件,亮度过强或者过弱都会引起图像质量的下降;②清晰度,人们常用模糊,清晰等词汇来描述主观视觉对图像质量的感受。从频域角度看,高频分量不足往往是模糊的原因,而高频分量过多又会造成图像的粗糙。只有当一幅图像的高低频信息比例适当时会有清晰的感觉;③相关度,即图像之间的相似程度。相似程度不能简单地理解为像素灰度的差值,而是图像的形态、内容相似。People's evaluation of image quality is a complex psychological activity, and there are many related factors that affect people's judgment on image quality. If these factors are extracted separately and finally synthesized, it will be well used for judging the image quality. These factors mainly include: ①Brightness. Proper brightness is the basic condition for people to observe images. If the brightness is too strong or too weak, it will cause the image quality to decline; ②Sharpness. feelings. From the perspective of the frequency domain, insufficient high-frequency components are often the cause of blur, while excessive high-frequency components will cause rough images. Only when the ratio of high and low frequency information of an image is appropriate will there be a clear feeling; ③ Correlation, that is, the degree of similarity between images. The degree of similarity cannot be simply understood as the difference in pixel grayscale, but the shape and content of the image are similar.
基于结构失真的图像质量评价方法。该图像质量评价指标将图像质量降低模拟为结构失真而不是误差,通过计算原始图像和失真图像结构的相似度来评价图像质量。该方法是一个简单有效的质量评价算法,这种评价模型的优点是对图像处理过程中的相关性、亮度失真和反差等因素进行分析研究,而这些因素正是人们评价图像输出质量的重要依据。然而,基于结构失真的图像质量评价模型在评价失真图像质量时,也有一定的缺陷性。其主要原因在于该模型没有考虑人眼视觉系统中的对比敏感度对图像评价的影响。人眼视觉系统的对比敏感度主要反映在图像的清晰度等方面。Image Quality Assessment Method Based on Structural Distortion. The image quality evaluation index simulates the degradation of image quality as structural distortion rather than error, and evaluates image quality by calculating the similarity between the original image and the distorted image structure. This method is a simple and effective quality evaluation algorithm. The advantage of this evaluation model is to analyze and study factors such as correlation, brightness distortion and contrast in the process of image processing, and these factors are the important basis for people to evaluate image output quality. . However, the image quality evaluation model based on structural distortion also has certain defects when evaluating the quality of distorted images. The main reason is that the model does not consider the impact of contrast sensitivity in the human visual system on image evaluation. The contrast sensitivity of the human visual system is mainly reflected in the clarity of the image and so on.
基于结构失真与空间频率指标的图像质量评价方法在基于结构失真的图像质量评价方法的基础上,引入空间频率指标对图像清晰度进行评价,从而实现了从亮度、清晰度、相关度三方面对图像的质量评价。该方法得出的结果更符合人眼的视觉感受。The image quality evaluation method based on structural distortion and spatial frequency index is based on the image quality evaluation method based on structural distortion, and the spatial frequency index is introduced to evaluate the image clarity, so as to realize the evaluation from the three aspects of brightness, clarity and correlation. Image quality evaluation. The results obtained by this method are more in line with the visual experience of the human eye.
本发明技术方案可采用计算机软件技术实现自动运行流程,以下结合实施例详细说明本发明技术方案。如图1,本发明实施例的流程包括以下步骤:The technical solution of the present invention can adopt computer software technology to realize the automatic operation process, and the technical solution of the present invention will be described in detail below in conjunction with the embodiments. As shown in Figure 1, the process of the embodiment of the present invention includes the following steps:
步骤1,计算原始图像与失真图像的均值利用关系式计算原始图像与失真图像的平均亮度逼近度,Step 1, calculate the mean of the original image and the distorted image Use relational Calculate the average brightness approximation of the original image and the distorted image,
其中,x,y分别代表原始图像与失真图像,图像中横向和纵向的像素数目分别为m和n,Among them, x and y represent the original image and the distorted image respectively, and the number of horizontal and vertical pixels in the image are m and n respectively,
x(i,j)和y(i,j)分别为原始图像和失真图像在第i行第j列的灰度值,平均亮度逼近度的取值范围为[0,1],只有当时,平均亮度逼近度值为1;x(i, j) and y(i, j) are the gray values of the original image and the distorted image at row i, column j, respectively, and the value range of the average brightness approximation is [0, 1]. Only when When , the average brightness approximation value is 1;
步骤2,计算原始图像与失真图像的均方误差σx,σy及协方差σxy,利用关系式计算原始图像与失真图像的线性相关度,Step 2, calculate the mean square error σ x , σ y and covariance σ xy of the original image and the distorted image, using the relation Calculate the linear correlation between the original image and the distorted image,
其中,x,y分别代表原始图像与失真图像,图像在横向和纵向的像素数目分别为m和n,Among them, x and y represent the original image and the distorted image respectively, and the number of pixels in the horizontal and vertical directions of the image are m and n respectively,
x(i,j)和y(i,j)分别为原始图像和失真图像在第i行第j列的灰度值,线性相关度的动态范围为[-1,1];x(i, j) and y(i, j) are the gray values of the original image and the distorted image in row i, column j, respectively, and the dynamic range of the linear correlation is [-1, 1];
步骤3,根据步骤2所得原始图像与失真图像的均方误差σx,σy,利用关系式计算原始图像与失真图像的相似度;Step 3, according to the mean square error σ x , σ y of the original image and the distorted image obtained in step 2, use the relation Calculate the similarity between the original image and the distorted image;
相似度的取值范围为[0,1],只有当σx=σy时,取得最佳值1;The value range of the similarity is [0, 1], only when σ x = σ y , the best value 1 is obtained;
步骤4,计算失真图像的空间行频率RF以及空间列频率CF,通过空间行频率RF及空间列频率CF计算失真图像空间频率指标计算空间行频率RF和空间列频率CF的公式如下Step 4, calculate the spatial row frequency RF and spatial column frequency CF of the distorted image, and calculate the spatial frequency index of the distorted image through the spatial row frequency RF and spatial column frequency CF The formulas for calculating the spatial row frequency RF and the spatial column frequency CF are as follows
其中,m,n分别为图像在横向和纵向的像素数目,y(i,j)为失真图像在第i行第j列的灰度值,y(i-1,j)即失真图像在第i-1行第j列的灰度值,y(i,j-1)即失真图像在第i行第j-1列的灰度值,图像空间频率指标的值域为[0,255],其值越大,图像的清晰度越高;Among them, m and n are the number of pixels in the horizontal and vertical directions of the image respectively, y(i, j) is the gray value of the distorted image in row i, column j, and y(i-1, j) is the distorted image in the The gray value of column j in row i-1, y(i, j-1) is the gray value of column j-1 in row i of the distorted image, and the value range of the image spatial frequency index is [0, 255] , the larger the value, the higher the image definition;
步骤5,综合步骤1所得平均亮度逼近度、步骤2所得线性相关度、步骤3所得相似度及步骤4所得失真图像空间频率指标,计算基于结构失真与空间频率指标的图像质量评价指标图像质量评价指标的值域为[0,255],根据计算所得图像质量评价指标数值从图像的亮度、清晰度以及相关性三方面对图像质量进行综合评价。Step 5, combining the average brightness approximation degree obtained in step 1, the linear correlation degree obtained in step 2, the similarity degree obtained in step 3 and the spatial frequency index of the distorted image obtained in step 4, and calculating the image quality evaluation index based on structural distortion and spatial frequency index The value range of the image quality evaluation index is [0, 255]. According to the calculated image quality evaluation index value, the image quality is comprehensively evaluated from the three aspects of image brightness, clarity and correlation.
具体实施时,前4个步骤可以并行处理,本领域技术人员可以自行调整。但为节约资源起见,步骤3可以直接利用步骤2所得原始图像与失真图像的均方误差σx,σy,因此适合在步骤2后进行。During specific implementation, the first four steps can be processed in parallel, and those skilled in the art can make adjustments by themselves. However, in order to save resources, step 3 can directly use the mean square error σ x , σ y of the original image and the distorted image obtained in step 2, so it is suitable to be performed after step 2.
为便于说明本发明效果起见,采用根据本发明实施例设计的软件进行试验:For the convenience of illustrating the effect of the present invention, the software designed according to the embodiment of the present invention is used to test:
(1)利用软件读入原始图像与失真图像,试验选用Lena图为测试图像,原始图像记为图2-1,加入均匀分布随机噪声的图像记为图2-2,加入高斯噪声的图像记为图2-3,模糊处理的图像记为图2-4,压缩处理的图像记为图2-5。此时,图像在软件中以矩阵的形式存储,若设图像大小为m×n,则矩阵的大小为m×n,在本试验中,令N=m×n,矩阵中的每个元素存储着对应图像的像素值。(1) Use the software to read in the original image and the distorted image. The Lena image is selected as the test image for the test. The original image is recorded as Fig. 2-1, the image with uniformly distributed random noise is recorded as Fig. 2-2, and the image with Gaussian noise is recorded as Fig. 2-2. Figure 2-3, the blurred image is marked as Figure 2-4, and the compressed image is marked as Figure 2-5. At this time, the image is stored in the form of a matrix in the software. If the size of the image is m×n, the size of the matrix is m×n. In this experiment, let N=m×n, and each element in the matrix is stored corresponds to the pixel value of the image.
(2)以下步骤中,令x,y分别代表原始图像与失真图像,对于各幅失真图像,分别与原始图像进行比较,计算原始图像与失真图像相应的σx,σy,σxy,通过公式得到各幅失真图像与原始图像相比的Q值。Q值反映了图像的平均亮度逼近度、线性相关度和相似度的指标。其中,令Q21、Q31、Q41、Q51分别代表失真图像2-2,2-3,2-4,2-5与原始图像2-1相比计算得到的Q值,在本试验中,Q21=0.647,Q 31=0.647,Q 41=0.647,Q 51=0.647。(2) In the following steps, let x and y represent the original image and the distorted image respectively, and compare each distorted image with the original image, and calculate the corresponding ratio of the original image and the distorted image σ x , σ y , σ xy , by the formula Get the Q value of each distorted image compared with the original image. The Q value reflects the index of the average brightness approximation, linear correlation and similarity of the image. Wherein, make Q21, Q31, Q41, Q51 represent distorted image 2-2, 2-3, 2-4, 2-5 and the Q value that the original image 2-1 calculates respectively, in this test, Q21= 0.647, Q 31 =0.647, Q 41 =0.647, Q 51 =0.647.
(3)编程计算失真图像空间行频率以及空间列频率
(4)本发明设定基于结构失真与空间频率指标的图像质量评价指标:SFQ可以从图像的亮度、清晰度以及相关性三方面对图像质量进行综合评价。令SFQ21、SFQ31、SFQ41、SFQ51分别代表失真图像2-2,2-3,2-4,2-5与原始图像2-1相比计算得到的PQ值,在本试验中,SFQ21=20.403,SFQ31=19.033,SFQ41=6.479,SFQ51=9.295。(4) The present invention sets the image quality evaluation index based on structural distortion and spatial frequency index: SFQ can comprehensively evaluate the image quality from the three aspects of image brightness, clarity and correlation. Let SFQ21, SFQ31, SFQ41, and SFQ51 represent the PQ values obtained by comparing the distorted image 2-2, 2-3, 2-4, and 2-5 with the original image 2-1 respectively. In this test, SFQ21=20.403, SFQ31=19.033, SFQ41=6.479, SFQ51=9.295.
(5)本试验还给出了失真图像2-2,2-3,2-4,2-5与原始图像2-1相比计算得到的传统图像质量评价参数PSNR值,PSNR值由以下公式得出其中,m×n为图像的大小,x(i,j),y(i,j)分别代表原始图像与失真图像的对应像素值。PSNR21=27.587,PSNR31=22.885,PSNR41=24.84,PSNR51=25.523。(5) This test also gives the PSNR value of the traditional image quality evaluation parameter calculated by comparing the distorted image 2-2, 2-3, 2-4, 2-5 with the original image 2-1, and the PSNR value is given by the following formula inferred Among them, m×n is the size of the image, and x(i, j), y(i, j) respectively represent the corresponding pixel values of the original image and the distorted image. PSNR21=27.587, PSNR31=22.885, PSNR41=24.84, PSNR51=25.523.
(6)可以明显看出,图2-2,2-3,2-4,2-5的视觉效果并不一致,图2-2和图2-3的主观感觉要明显优于图2-4和图2-5。附表1中的SFQ值正好反映了这一点,然而Q21=Q31=Q41=Q51=0.647。Q值以及PSNR值不能将各失真图像的差别体现出来,所以基于结构失真与空间频率指标的图像质量评价方法得出的结果更符合人眼的视觉感受。(6) It can be clearly seen that the visual effects of Figure 2-2, 2-3, 2-4, and 2-5 are not consistent, and the subjective feeling of Figure 2-2 and Figure 2-3 is obviously better than that of Figure 2-4 and Figures 2-5. The SFQ values in Appendix 1 reflect exactly this, however Q21 = Q31 = Q41 = Q51 = 0.647. The Q value and PSNR value cannot reflect the difference of each distorted image, so the result obtained by the image quality evaluation method based on structural distortion and spatial frequency index is more in line with the visual experience of the human eye.
附表1Lena图Q值、PSNR值、SFQ值测试结果Attached Table 1 Lena chart Q value, PSNR value, SFQ value test results
本发明还可采用软件模块化设计技术实现为基于结构失真与空间频率指标的图像质量评价系统,包括:The present invention can also be implemented as an image quality evaluation system based on structural distortion and spatial frequency indicators by using software modular design technology, including:
平均亮度逼近度模块,用于计算原始图像与失真图像的均值利用关系式计算原始图像与失真图像的平均亮度逼近度,计算所得平均亮度逼近度输入综合评价模块;Average brightness approximation module for computing the mean of the original image and the distorted image Use relational Calculate the average brightness approximation degree of the original image and the distorted image, and input the calculated average brightness approximation degree into the comprehensive evaluation module;
线性相关度模块,用于计算原始图像与失真图像的均方误差σx,σy及系数σxy,利用关系式计算原始图像与失真图像的线性相关度,计算所得线性相关度输入综合评价模块;相似度模块,用于计算原始图像与失真图像的均方误差σx,σy,利用关系式计算原始图像与失真图像的相似度;计算所得相似度输入综合评价模块;Linear correlation module, used to calculate the mean square error σ x , σ y and coefficient σ xy of the original image and the distorted image, using the relationship Calculate the linear correlation between the original image and the distorted image, and input the calculated linear correlation into the comprehensive evaluation module; the similarity module is used to calculate the mean square error σ x , σ y of the original image and the distorted image, using the relationship Calculate the similarity between the original image and the distorted image; the calculated similarity is input into the comprehensive evaluation module;
失真图像空间频率指标模块,用于计算失真图像的空间行频率RF以及空间列频率CF,通过空间行频率RF及空间列频率CF计算失真图像空间频率指标计算所得失真图像空间频率指标输入综合评价模块;The distorted image spatial frequency index module is used to calculate the spatial row frequency RF and spatial column frequency CF of the distorted image, and calculate the distorted image spatial frequency index through the spatial row frequency RF and spatial column frequency CF The calculated spatial frequency index of the distorted image is input into the comprehensive evaluation module;
综合评价模块,用于综合平均亮度逼近度、线性相关度、相似度及失真图像空间频率指标,计算基于结构失真与空间频率指标的图像质量评价指标根据计算所得图像质量评价指标数值从图像的亮度、清晰度以及相关性三方面对图像质量进行综合评价。The comprehensive evaluation module is used to synthesize the average brightness approximation degree, linear correlation degree, similarity degree and distorted image spatial frequency index, and calculate the image quality evaluation index based on structural distortion and spatial frequency index According to the calculated image quality evaluation index value, the image quality is comprehensively evaluated from the three aspects of image brightness, clarity and correlation.
本文中所描述的具体实施例仅仅是对本发明精神作举例说明。本发明所属技术领域的技术人员可以对所描述的具体实施例做各种各样的修改或补充或采用类似的方式替代,但并不会偏离本发明的精神或者超越所附权利要求书所定义的范围。The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which the present invention belongs can make various modifications or supplements to the described specific embodiments or adopt similar methods to replace them, but they will not deviate from the spirit of the present invention or go beyond the definition of the appended claims range.
Claims (1)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012101254006A CN102663764A (en) | 2012-04-25 | 2012-04-25 | Image quality evaluation method based on structural distortion and spatial frequency index |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2012101254006A CN102663764A (en) | 2012-04-25 | 2012-04-25 | Image quality evaluation method based on structural distortion and spatial frequency index |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102663764A true CN102663764A (en) | 2012-09-12 |
Family
ID=46773242
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2012101254006A Pending CN102663764A (en) | 2012-04-25 | 2012-04-25 | Image quality evaluation method based on structural distortion and spatial frequency index |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102663764A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107784645A (en) * | 2016-08-26 | 2018-03-09 | 广州康昕瑞基因健康科技有限公司 | Measurement for Digital Image Definition and system, auto focusing method |
CN108122229A (en) * | 2017-12-25 | 2018-06-05 | 重庆大学 | A kind of anti-image quality evaluating method of full reference resistance |
CN108629293A (en) * | 2018-04-16 | 2018-10-09 | 西安交通大学 | A kind of adaptive near-infrared iris image acquiring method with feedback mechanism |
CN110473200A (en) * | 2019-08-22 | 2019-11-19 | 深圳大学 | The video image quality evaluation method referred to entirely |
CN114445303A (en) * | 2016-02-16 | 2022-05-06 | 6115187加拿大公司暨伊美景象公司 | Image distortion transformation method and apparatus |
CN116843683B (en) * | 2023-08-30 | 2024-03-05 | 荣耀终端有限公司 | Equipment imaging clarity evaluation methods, systems and devices |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101441768A (en) * | 2008-11-28 | 2009-05-27 | 武汉大学 | Image quality evaluating method based on structure distortion and image definition |
-
2012
- 2012-04-25 CN CN2012101254006A patent/CN102663764A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101441768A (en) * | 2008-11-28 | 2009-05-27 | 武汉大学 | Image quality evaluating method based on structure distortion and image definition |
Non-Patent Citations (3)
Title |
---|
NIRANJAN DAMERA-VENKATA ET AL: "Image quality assessment based on a degradation model", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
SONJA GRGIE ET AL: "Reliability of objective picture quality measures", 《JOURNAL OF ELECTRICAL ENGINEERING》 * |
王正友 等: "基于对比敏感度的图像结构失真测量", 《第十三届全国图象图形学学术会议》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114445303A (en) * | 2016-02-16 | 2022-05-06 | 6115187加拿大公司暨伊美景象公司 | Image distortion transformation method and apparatus |
CN107784645A (en) * | 2016-08-26 | 2018-03-09 | 广州康昕瑞基因健康科技有限公司 | Measurement for Digital Image Definition and system, auto focusing method |
CN108122229A (en) * | 2017-12-25 | 2018-06-05 | 重庆大学 | A kind of anti-image quality evaluating method of full reference resistance |
CN108122229B (en) * | 2017-12-25 | 2021-02-02 | 重庆大学 | Full-reference electrical impedance imaging image quality evaluation method |
CN108629293A (en) * | 2018-04-16 | 2018-10-09 | 西安交通大学 | A kind of adaptive near-infrared iris image acquiring method with feedback mechanism |
CN108629293B (en) * | 2018-04-16 | 2020-10-27 | 西安交通大学 | Self-adaptive near-infrared iris image acquisition method with feedback mechanism |
CN110473200A (en) * | 2019-08-22 | 2019-11-19 | 深圳大学 | The video image quality evaluation method referred to entirely |
CN116843683B (en) * | 2023-08-30 | 2024-03-05 | 荣耀终端有限公司 | Equipment imaging clarity evaluation methods, systems and devices |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102881010B (en) | Method for evaluating perception sharpness of fused image based on human visual characteristics | |
CN101441768A (en) | Image quality evaluating method based on structure distortion and image definition | |
CN103295191B (en) | Multiple scale vision method for adaptive image enhancement and evaluation method | |
CN104023230B (en) | A kind of non-reference picture quality appraisement method based on gradient relevance | |
CN101976444B (en) | An Objective Evaluation Method of Image Quality Based on Structural Similarity Based on Pixel Type | |
CN102663764A (en) | Image quality evaluation method based on structural distortion and spatial frequency index | |
CN102421007A (en) | Image Quality Evaluation Method Based on Weighted Synthesis of Multi-scale Structure Similarity | |
CN107464222B (en) | No-reference high dynamic range image objective quality assessment method based on tensor space | |
CN102170581A (en) | Human-visual-system (HVS)-based structural similarity (SSIM) and characteristic matching three-dimensional image quality evaluation method | |
CN105744256A (en) | Three-dimensional image quality objective evaluation method based on graph-based visual saliency | |
CN105118053B (en) | A kind of full reference picture assessment method for encoding quality based on compressed sensing | |
CN102945552A (en) | No-reference image quality evaluation method based on sparse representation in natural scene statistics | |
CN106920232A (en) | Gradient similarity graph image quality evaluation method and system based on conspicuousness detection | |
CN102222323A (en) | Histogram statistic extension and gradient filtering-based method for enhancing infrared image details | |
CN106408561A (en) | Texture feature-based image quality evaluating method without reference | |
CN105407349A (en) | No-reference objective three-dimensional image quality evaluation method based on binocular visual perception | |
CN106454350A (en) | A No-Reference Evaluation Method for Infrared Images | |
CN103945217B (en) | Based on complex wavelet domain half-blindness image quality evaluating method and the system of entropy | |
CN101489130A (en) | Complete reference image quality assessment method based on image edge difference statistical characteristic | |
CN110910347B (en) | A No-Reference Quality Assessment Method for Tone Mapping Images Based on Image Segmentation | |
CN102890820A (en) | Image denoising method based on shearlet transformation and Wiener filtering | |
Yang et al. | Blind assessment for stereo images considering binocular characteristics and deep perception map based on deep belief network | |
CN104318545A (en) | Foggy weather polarization image quality evaluation method | |
CN104346809A (en) | Image quality evaluation method for image quality dataset adopting high dynamic range | |
CN105976351A (en) | Central offset based three-dimensional image quality evaluation method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120912 |