• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Memristor-Based Genetic Algorithm for lmage Restoration

    2022-07-08 01:38:38YongBinYuChenZhouQuanXinDengYuanJingYangZhongManChengZhengFeiKang

    Yong-Bin Yu | Chen Zhou | Quan-Xin Deng | Yuan-Jing-Yang Zhong | Man Cheng | Zheng-Fei Kang

    Abstract—This paper explores a way of deploying the classical algorithm named genetic algorithm (GA) with the memristor.The memristor is a type of circuit device with both characteristics of storage and computing, which provides the similarity between electronic devices and biological components, such as neurons, and the structure of the memristor-based array is similar to that of chromosomes in genetics.Besides, it provides the similarity to the image gray-value matrix that can be applied to image restoration with GA.Thus, memristor-based GA is proposed and the experiment about image restoration using memristor-based GA is carried out in this paper.And parameters,such as the size of initial population and the number of iterations, are also set different values in the experiment,which demonstrates the feasibility of implementing GA with memristors.

    1.lntroduction

    Memristor, the abbreviation name of the memory resistor, was first proposed by Chua in 1971[1].From the point of view of the circuit theory, there should also be a circuit element to represent the relationship between the flux-linkage and current, then, the memristor was proposed.However, limited by the production technology at that time, it was not born until 2008 when the HP lab made the first memristor[2], arousing the research upsurge of memristors globally.Due to the special switching mechanism, memory, continuous input and output, and nano scale, memristors have great potential applications.For example, the memristorcomplementary metal-oxide-semiconductor transistor (memristor-CMOS) hybrid circuits could be used to realize neural networks[3].The memristor is also well-behaved in signal processing[4]and cyber-security[5].Recently, its value in image processing has been found with the application in image processing.The memristor bridge-based lowpass filter is designed along with an adaptive Gaussian filter, which is presented for denosing images[6]and has been applied to edge detection[7].Besides, it has good practical value in hand gesture recognition[8],[9].

    The genetic algorithm (GA) is a method that Holland[10]draws lessons from the basic principles of genetics to simulate natural evolution.It is used in designing the artificial adaptive system.A computational model is proposed based on natural selection that simulates Darwin’s theory of biological evolution.The basic idea of GA is to select individuals from the initial population, adopt the natural law of “survival of the fittest”, and then produce a new generation of population through hybridization and mutation, so as to evolve generation by generation until the goal is met.Nowadays, it works well in solving complex nonlinear combinatorial optimization problems, such as the traveling salesman problem[11]and image processing[12].

    In recent years, image restoration has become important in many fields, such as medical imaging and graphic printing.Digital images are often corrupted by degradation[13].The reason of image blur is complicated, the restoration cannot be processed by the regular linear function.So, it is proper to use GA to restore the image.

    Because of the advantage of memristors in storage, the memristor-based array is proposed, which also can be called the memristive network.This array is capable of solving more complex problems compared with the single memristor unit.Besides, it has the similarity to the image gray-value matrix, and its structure is similar to chromosomes and genes.In the memristor-based array, every pixel of the image has a corresponding memristor unit to store its value, making it possible to realize storage for binary, grayscale, and color images.So our goals are to verify the feasibility of deploying GA with memristors, explore mapping rules between GA and the memristor-based array, and deploy GA with memristors through the mapping rules in image restoration.

    2.Computing Model

    2.1.GA

    GA can be divided into three parts.1) The first part is initialization.Engineering problems are mapped into population evolution problems in the biological world.Coding and generating population and feasible solutions of problems can be defined by chromosomes and genes.2) The second part is to design the fitness function, that is, choosing the appropriate standard to quantify the individual’s adaptation to the living environment.Fitness is an important standard to evaluate the individual and the basis of other operations.Because the requirement of the fitness function is not particularly strict and it is not forced to be differentiable,derivable, and so on, GA is suitable to deal with nonlinear problems.3) The third part is the core operations of GA, including selection, crossover, and mutation.Selection is to select individuals with good characters in the current population, which is the basis of crossover and mutation operations.Crossover is the exchange of the segment of parents’ genes, that is, the process of reproduction in biology, which is an important part of GA.And the genes of parents can be continued because of crossover.Mutation is the idea of mutation in biology,sometimes, gene mutation occurs during the process that offspring inherits the parent gene, and mutation is beneficial for the population to jump out of the local optimal solution.The specific process of GA is showed inFig.1.

    2.2.Memristor-Based Array

    Limited by the computing ability of a single memristor, the memristor-based array was designed for more complex problems.The memristor-based array has the structure similarity to a two-dimensional matrix, which is composed of multiple memristors.Its structure is shown inFig.2.Each memristor can be added to this array through two connecting wires, and the deletion of the memristor only needs to disconnect the two connecting wires between the memristor and array[14].

    Fig.1.Process of GA.

    Fig.2.Structure of memristor-based array.

    There have been applications of using the memristive network to implement other swarm intelligence algorithms.For example, Pershin and Ventra[15]compared the memory characteristics of memristors with the pheromones left by ants on the path, proposed a mapping relationship, and used memristors to deploy the ant colony optimization, which inspires us to explore memristor-based GA.

    2.3.Memristor-Based GA

    To solve optimization problems, the analogy and mapping between GA and the memristor array is then conducted.The structure of the memristor-based array is similar to that of the chromosome, so a single memristor in the array can be abstracted into a gene unit.Then the operation in GA is mapped to the change of the resistance value of the memristor in the memristor-based array.The mapping relationship is shown inTable 1.The memristor is the basic unit of the memristor network and the gene is the basic unit of the chromosome; the sum of information of each pixel constitutes an image, so the memristor is mapped to genes in GA, which corresponds to the gray value information of a pixel in an image.Multiple genes form a chromosome or individual, so the chromosome corresponds to a set of memristors, that is, a memristor network and a set of gray value information of pixels in the memristor network.

    Table 1: Mapping rules among nature, GA, memristive network, and image restoration

    3.lmage Restoration Based on Memristor-Based GA

    This section will introduce the theory of image restoration, detailed design of image restoration based on GA, including the way of coding and generating initial population, the selection of the fitness function, and the selection of the basic operation of GA.

    3.1.Theory of lmage Restoration

    In the process of image generation, transportation, and processing, the phenomenon of image blur caused by man-made and natural factors, such as the camera jitter, inappropriate light, and haze, is called image degradation.Except image blur, the degradation is often manifested as distortion, additional noise, etc.,resulting in the decline in image quality.In the image degradation model, image degradation is modeled by using the degradation function and adding the external noise.The model is showed inFig.3,xandydenote coordinates in the spatial domain,f(x,y)represents the original clear image in the spatial domain,h(x,y) is the degradation function,n(x,y)represents the noise, andg(x,y) denotes the degraded image.

    The mathematical expression of the degenerate model can be expressed as[16]

    Fig.3.Image degradation model.

    where * denotes the convolution operation.After the Fourier transform, (1) in the frequency domain is

    It can be inferred that the restored images in the frequency domain is

    G(u,v) is defined in (2), and then substituted into (3):

    The restored image is obtained by using the inverse Fourier transform:

    wheref′(x,y) denotes the restored image.So, it is possible to restore a blurred image without nois e.

    Before the start of GA, the significance and design of image segmentation will be introduced.During the process of image restoration, the number of pixel in a picture is very big, so the computing burden will be heavy.Suppose that the size of the image to be restored is 100×30 pixels, and the image is not divided into blocks.At this time, an image is an individual has 100×30 = 3000 genes, and the range of each gene is an integer of 0 to 255.For one graph, there are 2563000different feasible solutions, which is a very big number.Usually, the size of the initial population is about 20 to 120.In this case, the size of the randomly generated initial population is only about 100/2563000, and the result of this ratio is close to zero.Therefore, the probability of getting individuals with good characters is very small, that is, the fitness of individuals is very low, and the use of improved selection, crossover, and mutation algorithms cannot get good results afterwards.And the speed of convergence is very slow, because the individual does not perform well from the beginning.Furthermore, in order to increase the probability of producing good initial individuals, we have to choose to increase the size of the initial population.However, increasing the size of the initial population means that more space is needed to store individual information, and each additional individual of the initial population needs more 3000 memories to store individual information, so the space cost is very huge.The most critical problem is that even if the size of the initial population is increased to thousands, tens of thousands, or even hundreds of thousands, it is still a very small number compared with 2563000.Moreover,when the size of the picture increases slightly, this number will grow exponentially, and the increasing population size will greatly increase the amount of memristors needed and the space cost.The effect on improving the efficiency of the algorithm is negligible.So, above all, we put forward to process the image by blocks.

    3.2.Divide lmage into Small Blocks

    In this situation, the size of the image block is set as 3×3 pixels, and the image block is moved according to the rules from left to right and from top to bottom.The 3×3-image block first traverses from left to right,moves one pixel down after traversing a whole line, and continues to traverse a whole row until all pixels have been traversed.

    Thus, an individual is not a whole image, but the size of a block, and its structure is like that a chromosome has nine genes.To simplify the operation, the data matrix of a 3×3 block will be converted into a matrix of 1×9.Assuming that the number of the initial population is 20, then the size of the initial population is the two-dimensional matrix of 20×9.Besides, each individual also needs a memory to store the value of fitness, so the size of the matrix is extended to 20×10.This matrix is named as the operation matrix, in which the first nine genes in each row form an individual.The number of rows represents the size of the population,and the last one stores the individual fitness value.So two memristor-based arrays are required, one is to store the blurred image (called MNblur instead), and another is to store the restored image (called MNres instead).In addition, an operation-matrix memristor-based array is also needed, GA operations all happen in the operation-matrix memristor-based array.

    3.3.Select GA Operation for lmage Restoration

    3.3.1.Coding and Initializing Population

    Two-dimensional integer coding is adopted by this paper, because the picture can be seen as a twodimensional matrix storing pixel information.The gray value of the image is expressed as the genetic value directly, which not only retains the similarity between the data and the image, but also avoids the complex coding and decoding operations.It is helpful to improve the operation speed, and then stochastically generate initial population.

    3.3.2.Fitness Function

    The fitness function is a function to evaluate the fitness of an individual, which is generally required to be non-negative.The purpose of image restoration is to obtain the best original image, but the restoration quality of an image can only be known by comparing with the original image and restored image, especially, the original image is unknown.However, we discussed the image degradation model in subsection 3.1, and the degraded imageg(x,y) is degraded from the original imagef(x,y). If the restored imagef′(x,y) can also produce another degraded imageg′(x,y) after the same degradation function, then bothg(x,y) andg′(x,y)are known and can be compared.If the finalg′(x,y) is closer tog(x,y),f′(x,y) will be more similar tof(x,y),so the fitness function is designed by[8]

    3.3.3.Selection Function

    The selection function in this research adopts the roulette selection: The probability of each individual being selected is defined as

    wherexirepresents individual, andf(xi) denotes the fitness value of individualxi.

    3.3.4.Crossover Function

    As we discussed before, we designed image block processing, and one image block is an individual.The number of genes is very small, so instead of the complex way of crossover, one-point crossover is used.The crossing point is selected by a random function.According to the position of the selected crossing point, a chromosome is divided into two parts before and after the crossing point, where the matching parents exchange the front or back gene values of the crossing point.

    3.3.5.Mutation Function

    The mutation function imitates the mutation of biological genes in nature.First, a gene bit on the chromosome is selected randomly, and the decimal gray value on the gene position is converted into the binary code of 8 bits, then one bit is randomly selected to change, and finally the changed binary value is converted into the decimal gray value.

    4.Simulation Results

    4.1.Data of Experiment

    The experimental data is the captured picture shown inFig.4, and the size is 101×26 pixels.The degraded image is obtained from the original image by a 4×4-point spread function, as shown inFig.5.

    Fig.4.Test image.

    Fig.5.Degraded image.

    4.2.Experiment Results

    We discussed the design of dividing image into blocks, so in Experiment 1, the results of dividing the image into blocks or not are compared.The set of parameters is shown inTable 2and the results are shown inFig.6.

    Figs.6(a) and (b) are the image restoration results of Case 1 and Case 2, designed inTable 2,respectively.They are set with the same value of the initial population and iteration, and the only difference is the operation of dividing image into blocks or not.Obviously, as shown inFig.6(b), by dividing the image into blocks, Case 2 has a better restoration result.Because of the large number of pixels and the wide range of pixels in the image processed without blocks, it is difficult to get good individuals by randomly generating the initial population.As a result, the convergence speed is very slow in the later stage, and the individual performance is still very poor after the same number of iterations.And it is less likely to obtain more information through the restored image in Case 1.

    Then in Experiment 2, different values of the population size and iterations are set inTable 3.For all cases inTable 3, the dividing operation with the block size of 3×3 pixels is adopted.

    Table 2: Case design

    Table 3: Different parameter values

    Fig.6.Image restoration results of case design in Table 2:(a) result of Case 1 and (b) result of Case 2.

    Then, we get different results and their running time, as shown inFig.7.

    Fig.7.Experiment 2 results: (a) ID = 2 and t = 50 min, (b) ID = 3 and t = 22 min, (c) ID = 4 and t = 10.5 min, and (d) ID = 5 and t = 7 min.

    4.3.Comparative Study

    The contrastive study is shown in this subsection inTable 4.And experiment results are shown inFig.8.

    Fig.8.Comparative study results: (a) Gaussian filter, (b) median filter, and (c) our proposed model.

    Figs.8(a) and (b) are the results of applying the Gaussian and median filters, respectively, andFig.8(c) is the results by applying the proposed model, which is better than the other two.From bothTable 4andFig.8, it can be seen that image restoration results are not good by directly performing the filtering operation on the image, and our proposed model performs better, which verifies the improvement of our model.

    Table 4: Compare the proposed method with filters

    Compared with Case 4, Case 3 has a larger number of iterations, leading to longer running time and a better result.Compared with Case 5, Case 4 has a larger population size, leading to longer running time and a clearer result.Cases 1 to 5 all demonstrate the feasibility of deploying GA with memristors in image restoration.Besides, when the initial population and the number of iterations are too small (as shown inFig.7(d)), the restoration result of the image is the worst: The words and numbers are vague.After increasing the size of the initial population and the number of iterations, the restoration results are improved,but it is not that: The larger the size of population and the number of iterations are, the better the results.Therefore, the appropriate population size and iteration time should be chosen according to the actual needs,without blindly pursuing the large size of the initial population and iterations.Although the restoration result could be improved, the time cost is very large.So both the improvement of restoration and time cost should be taken into consideration.

    5.Conclusions

    In this paper, memristor-based GA for image restoration is proposed.The mapping rules between the memristor and GA are shown.According to the rules, memristor-based GA is designed and carried out successfully.The comparative study shows that the proposed method has better performance.It is demonstrated that using memristors to implement GA is possible and memristor-based GA for image restoration is feasible.

    Memristor, GA, and image restoration all have wide applications in different fields.There are still some other algorithms, like the ant colony optimization, which can be implemented in a similar way.But the existing methods are not completely unified yet.Once unified, more different algorithms can be implemented with the memristor-based array, which is helpful to increase the use of memristors.So more efforts are needed to design a unified system and enrich the application cases.

    Disclosures

    The authors declare no conflicts of interest.

    一级毛片我不卡| 极品人妻少妇av视频| 六月丁香七月| 国精品久久久久久国模美| 久久国产亚洲av麻豆专区| 2021少妇久久久久久久久久久| 精品国产国语对白av| 97精品久久久久久久久久精品| 最近的中文字幕免费完整| 国产无遮挡羞羞视频在线观看| 欧美成人午夜免费资源| 久久久a久久爽久久v久久| 久久久久久久亚洲中文字幕| 亚洲人成网站在线观看播放| 国产成人精品福利久久| 色网站视频免费| 日韩人妻高清精品专区| 69精品国产乱码久久久| 日韩大片免费观看网站| 夫妻午夜视频| 亚洲欧美日韩卡通动漫| 一本一本综合久久| 性色av一级| 国产高清三级在线| 大香蕉久久网| 制服丝袜香蕉在线| 日本av免费视频播放| 色婷婷av一区二区三区视频| 日本猛色少妇xxxxx猛交久久| 青青草视频在线视频观看| 精品卡一卡二卡四卡免费| 亚洲国产精品成人久久小说| 欧美 日韩 精品 国产| 又黄又爽又刺激的免费视频.| 蜜桃在线观看..| 人人妻人人澡人人爽人人夜夜| 亚洲电影在线观看av| 日产精品乱码卡一卡2卡三| 亚洲欧美成人综合另类久久久| 中文精品一卡2卡3卡4更新| 少妇裸体淫交视频免费看高清| 午夜老司机福利剧场| 最近的中文字幕免费完整| 高清毛片免费看| av专区在线播放| 国产精品.久久久| 边亲边吃奶的免费视频| 亚洲精品日韩av片在线观看| 各种免费的搞黄视频| 丰满迷人的少妇在线观看| videossex国产| 午夜激情福利司机影院| 乱人伦中国视频| 两个人免费观看高清视频 | 免费不卡的大黄色大毛片视频在线观看| 国产日韩一区二区三区精品不卡 | 亚洲精品亚洲一区二区| 久久热精品热| 97精品久久久久久久久久精品| 在线免费观看不下载黄p国产| 超碰97精品在线观看| 国产亚洲一区二区精品| 国产91av在线免费观看| 免费人成在线观看视频色| 国产欧美亚洲国产| 午夜福利网站1000一区二区三区| 亚洲美女黄色视频免费看| 我的女老师完整版在线观看| 丰满迷人的少妇在线观看| 国产精品嫩草影院av在线观看| 欧美 日韩 精品 国产| 久久这里有精品视频免费| 欧美激情极品国产一区二区三区 | 97精品久久久久久久久久精品| 国产乱人偷精品视频| 99热这里只有是精品50| 亚洲国产最新在线播放| 这个男人来自地球电影免费观看 | 黑丝袜美女国产一区| 亚洲av男天堂| 黄色视频在线播放观看不卡| 18+在线观看网站| 国产精品国产三级专区第一集| 狂野欧美激情性xxxx在线观看| 人妻系列 视频| 亚洲,欧美,日韩| 国产熟女午夜一区二区三区 | 熟妇人妻不卡中文字幕| 永久网站在线| 久久午夜综合久久蜜桃| 特大巨黑吊av在线直播| a 毛片基地| 久久这里有精品视频免费| 日日啪夜夜撸| 国产免费视频播放在线视频| 国产色婷婷99| 自拍偷自拍亚洲精品老妇| 在线观看美女被高潮喷水网站| 成人二区视频| 全区人妻精品视频| 日韩av免费高清视频| 久久精品夜色国产| 性高湖久久久久久久久免费观看| 日本wwww免费看| 亚洲精品国产av成人精品| 久久国产精品男人的天堂亚洲 | 国产精品久久久久久久电影| 国产一区亚洲一区在线观看| a级毛色黄片| 免费大片黄手机在线观看| 99热国产这里只有精品6| 亚洲成人手机| 国产淫语在线视频| 亚洲精品乱码久久久久久按摩| 91久久精品国产一区二区三区| 韩国高清视频一区二区三区| .国产精品久久| 亚洲自偷自拍三级| 免费看av在线观看网站| h日本视频在线播放| 桃花免费在线播放| a级毛色黄片| 岛国毛片在线播放| 国产亚洲欧美精品永久| 久久99热6这里只有精品| 日本黄色日本黄色录像| 爱豆传媒免费全集在线观看| 国产乱来视频区| 欧美日本中文国产一区发布| 在线观看三级黄色| 超碰97精品在线观看| 看十八女毛片水多多多| 日韩av免费高清视频| 天堂俺去俺来也www色官网| 免费看光身美女| 亚洲怡红院男人天堂| 亚洲四区av| 秋霞伦理黄片| 精品国产乱码久久久久久小说| 亚洲高清免费不卡视频| 久久这里有精品视频免费| 欧美精品一区二区免费开放| 春色校园在线视频观看| 人妻一区二区av| 国产极品粉嫩免费观看在线 | 国产视频内射| 精品国产一区二区三区久久久樱花| 久久久国产一区二区| 国产乱来视频区| 看十八女毛片水多多多| 亚洲精品国产成人久久av| 免费看日本二区| av福利片在线观看| 狂野欧美激情性xxxx在线观看| 亚洲av不卡在线观看| 在线天堂最新版资源| 亚洲欧洲日产国产| 99久久人妻综合| 高清黄色对白视频在线免费看 | 久久久久久久久久人人人人人人| 国产亚洲最大av| 欧美 日韩 精品 国产| 日本免费在线观看一区| 综合色丁香网| 日韩视频在线欧美| 免费高清在线观看视频在线观看| 国语对白做爰xxxⅹ性视频网站| 日韩精品免费视频一区二区三区 | 欧美激情国产日韩精品一区| 国产成人91sexporn| 国产淫片久久久久久久久| 国产毛片在线视频| 国产欧美另类精品又又久久亚洲欧美| 免费看av在线观看网站| 纵有疾风起免费观看全集完整版| 一二三四中文在线观看免费高清| 两个人免费观看高清视频 | 日本欧美视频一区| 日韩中字成人| 亚洲人与动物交配视频| 99视频精品全部免费 在线| 国国产精品蜜臀av免费| 高清在线视频一区二区三区| freevideosex欧美| 日本欧美国产在线视频| 国产精品一区二区三区四区免费观看| 精品久久国产蜜桃| 久久精品国产鲁丝片午夜精品| 天天操日日干夜夜撸| 人人妻人人添人人爽欧美一区卜| 中文字幕久久专区| 日韩制服骚丝袜av| 久久精品久久久久久久性| av天堂中文字幕网| 一级毛片久久久久久久久女| 又爽又黄a免费视频| 亚洲精品日韩在线中文字幕| 亚洲精品国产av成人精品| 尾随美女入室| 久久人人爽人人爽人人片va| 亚洲国产精品国产精品| 久久国产乱子免费精品| 欧美精品亚洲一区二区| 国产av码专区亚洲av| 噜噜噜噜噜久久久久久91| 99久久人妻综合| 亚洲精品日韩在线中文字幕| 日本欧美国产在线视频| 欧美最新免费一区二区三区| 午夜免费鲁丝| 久热这里只有精品99| 最新中文字幕久久久久| 青春草国产在线视频| 成人亚洲欧美一区二区av| 国产视频首页在线观看| 亚洲国产精品国产精品| 蜜臀久久99精品久久宅男| 国产免费一区二区三区四区乱码| 伦精品一区二区三区| 久久久久视频综合| 九九爱精品视频在线观看| 欧美精品一区二区大全| 亚洲精品国产av成人精品| 内射极品少妇av片p| 久久久久视频综合| 伊人久久国产一区二区| 国产一区二区在线观看日韩| 丰满饥渴人妻一区二区三| 全区人妻精品视频| 嫩草影院入口| 精品人妻熟女毛片av久久网站| 777米奇影视久久| 一级av片app| 午夜福利在线观看免费完整高清在| 国产综合精华液| 国产日韩欧美在线精品| 国产一区二区在线观看日韩| 啦啦啦啦在线视频资源| 国产永久视频网站| 18禁在线无遮挡免费观看视频| 人人妻人人看人人澡| 我的女老师完整版在线观看| 亚州av有码| 国产无遮挡羞羞视频在线观看| 三级国产精品欧美在线观看| 亚洲国产最新在线播放| 三级经典国产精品| 日韩成人av中文字幕在线观看| 一级爰片在线观看| 国产极品粉嫩免费观看在线 | 亚洲不卡免费看| 午夜日本视频在线| 国产淫片久久久久久久久| 亚洲av二区三区四区| 亚洲av电影在线观看一区二区三区| 亚洲国产毛片av蜜桃av| www.色视频.com| 免费看av在线观看网站| 国产精品国产三级专区第一集| 中文字幕精品免费在线观看视频 | 蜜桃在线观看..| 各种免费的搞黄视频| 国产欧美日韩一区二区三区在线 | av一本久久久久| 中文字幕人妻熟人妻熟丝袜美| 欧美日韩视频精品一区| 青春草国产在线视频| 成人毛片60女人毛片免费| 美女福利国产在线| 日韩,欧美,国产一区二区三区| xxx大片免费视频| 纯流量卡能插随身wifi吗| freevideosex欧美| 水蜜桃什么品种好| 交换朋友夫妻互换小说| 老司机影院成人| 久久久国产精品麻豆| 久久影院123| 精品一区二区三卡| 免费不卡的大黄色大毛片视频在线观看| 亚洲国产成人一精品久久久| 一区在线观看完整版| 狂野欧美激情性bbbbbb| 交换朋友夫妻互换小说| 男人爽女人下面视频在线观看| 国产高清不卡午夜福利| 国产免费视频播放在线视频| 岛国毛片在线播放| 日日爽夜夜爽网站| 日韩,欧美,国产一区二区三区| 国产视频首页在线观看| 另类亚洲欧美激情| 欧美少妇被猛烈插入视频| 成人美女网站在线观看视频| 色吧在线观看| 日日啪夜夜爽| 久久久久久伊人网av| 亚洲av不卡在线观看| 在线观看免费日韩欧美大片 | 少妇被粗大的猛进出69影院 | 国内少妇人妻偷人精品xxx网站| 午夜老司机福利剧场| 大香蕉久久网| 91久久精品国产一区二区成人| 午夜福利,免费看| 国国产精品蜜臀av免费| 久热这里只有精品99| 精品午夜福利在线看| 乱人伦中国视频| 看非洲黑人一级黄片| 天堂俺去俺来也www色官网| 国产永久视频网站| 狂野欧美白嫩少妇大欣赏| 久久99一区二区三区| 一级毛片我不卡| 久久久国产精品麻豆| 国产高清有码在线观看视频| 久久久欧美国产精品| 久久人人爽人人爽人人片va| 精品久久久噜噜| av免费在线看不卡| 99久久人妻综合| 久久精品久久久久久噜噜老黄| 成人二区视频| 日本午夜av视频| 三级国产精品片| 日韩av不卡免费在线播放| 国产精品嫩草影院av在线观看| 高清午夜精品一区二区三区| 插阴视频在线观看视频| 欧美三级亚洲精品| 一本久久精品| 午夜福利在线观看免费完整高清在| 一级毛片我不卡| av天堂中文字幕网| 一本久久精品| 日本黄大片高清| 亚洲va在线va天堂va国产| 黄色日韩在线| 精品久久久精品久久久| 人体艺术视频欧美日本| 高清欧美精品videossex| 免费av不卡在线播放| 少妇猛男粗大的猛烈进出视频| 国产一区二区三区综合在线观看 | 成人影院久久| 中国美白少妇内射xxxbb| 97在线视频观看| 青春草视频在线免费观看| 中国三级夫妇交换| 国产精品麻豆人妻色哟哟久久| 亚洲综合色惰| 99热这里只有是精品50| 九色成人免费人妻av| 国产日韩欧美在线精品| 国产精品免费大片| 国产成人午夜福利电影在线观看| 偷拍熟女少妇极品色| 久久av网站| 搡老乐熟女国产| a级一级毛片免费在线观看| 色哟哟·www| 青春草视频在线免费观看| 国产成人精品婷婷| av视频免费观看在线观看| 有码 亚洲区| 青春草视频在线免费观看| 大片电影免费在线观看免费| 久热这里只有精品99| 97在线视频观看| 久久毛片免费看一区二区三区| 丰满乱子伦码专区| 亚洲精华国产精华液的使用体验| 99九九线精品视频在线观看视频| 欧美人与善性xxx| 国产综合精华液| 99re6热这里在线精品视频| 日韩成人av中文字幕在线观看| 精品少妇黑人巨大在线播放| 免费观看无遮挡的男女| 色网站视频免费| 一区二区三区乱码不卡18| 国产av精品麻豆| 日韩免费高清中文字幕av| 国产有黄有色有爽视频| 国内精品宾馆在线| 制服丝袜香蕉在线| 狂野欧美白嫩少妇大欣赏| 亚洲av国产av综合av卡| 国产精品伦人一区二区| 99热国产这里只有精品6| 亚洲图色成人| 亚洲一级一片aⅴ在线观看| 久久久久久久亚洲中文字幕| √禁漫天堂资源中文www| 久久毛片免费看一区二区三区| 搡老乐熟女国产| 久久av网站| 嫩草影院新地址| 五月开心婷婷网| 国产精品久久久久久av不卡| av福利片在线观看| 日韩亚洲欧美综合| 伊人亚洲综合成人网| 国产中年淑女户外野战色| 汤姆久久久久久久影院中文字幕| 欧美成人精品欧美一级黄| 免费播放大片免费观看视频在线观看| 涩涩av久久男人的天堂| 免费久久久久久久精品成人欧美视频 | 欧美精品人与动牲交sv欧美| 免费黄色在线免费观看| 国产 一区精品| 久久久久久人妻| 亚洲国产av新网站| 日韩视频在线欧美| 青春草亚洲视频在线观看| 麻豆成人av视频| 久久 成人 亚洲| .国产精品久久| 久久国内精品自在自线图片| 22中文网久久字幕| 免费看av在线观看网站| 亚洲怡红院男人天堂| 天堂中文最新版在线下载| 中文乱码字字幕精品一区二区三区| 精品人妻偷拍中文字幕| 啦啦啦啦在线视频资源| 草草在线视频免费看| tube8黄色片| 2018国产大陆天天弄谢| 色5月婷婷丁香| 啦啦啦视频在线资源免费观看| 国产男女超爽视频在线观看| 国产精品嫩草影院av在线观看| a级毛片在线看网站| 亚洲精品第二区| 亚洲美女黄色视频免费看| 国产爽快片一区二区三区| 最近最新中文字幕免费大全7| 国产成人精品一,二区| 久久久久视频综合| 蜜桃在线观看..| 国产高清有码在线观看视频| 全区人妻精品视频| 亚洲欧美日韩另类电影网站| av福利片在线| 丁香六月天网| 男女边吃奶边做爰视频| 伊人久久精品亚洲午夜| 国产在视频线精品| 在线观看av片永久免费下载| 香蕉精品网在线| 久久久久久久久大av| 黄色欧美视频在线观看| 色哟哟·www| 国产淫语在线视频| 日韩三级伦理在线观看| 亚洲欧洲精品一区二区精品久久久 | 乱人伦中国视频| 日韩制服骚丝袜av| 国产91av在线免费观看| 亚洲精品日韩av片在线观看| 嘟嘟电影网在线观看| 国产一区二区三区av在线| 欧美日韩av久久| 在线天堂最新版资源| 国产熟女欧美一区二区| 有码 亚洲区| 久久精品熟女亚洲av麻豆精品| 国产精品国产三级国产av玫瑰| 在线观看人妻少妇| 日韩三级伦理在线观看| 在线观看人妻少妇| 国产成人91sexporn| 精品亚洲乱码少妇综合久久| 国产成人91sexporn| 老熟女久久久| 久久99蜜桃精品久久| 国产黄色视频一区二区在线观看| 大话2 男鬼变身卡| 婷婷色综合www| 国产高清不卡午夜福利| av在线播放精品| 亚洲av二区三区四区| 王馨瑶露胸无遮挡在线观看| 99精国产麻豆久久婷婷| 大香蕉97超碰在线| 两个人的视频大全免费| 亚洲国产精品成人久久小说| 亚洲精品乱码久久久久久按摩| 人妻系列 视频| 亚洲精品乱码久久久久久按摩| 欧美精品亚洲一区二区| freevideosex欧美| 亚洲成人一二三区av| 夜夜看夜夜爽夜夜摸| 亚洲欧美清纯卡通| 午夜福利网站1000一区二区三区| 少妇人妻精品综合一区二区| 国产精品一区二区性色av| 国产黄色视频一区二区在线观看| av在线观看视频网站免费| 成人二区视频| 3wmmmm亚洲av在线观看| 欧美成人精品欧美一级黄| 最新中文字幕久久久久| 久久精品国产a三级三级三级| 少妇人妻久久综合中文| 成人美女网站在线观看视频| 亚洲欧洲日产国产| 亚洲,欧美,日韩| 青青草视频在线视频观看| 十八禁网站网址无遮挡 | 国产精品不卡视频一区二区| 国产国拍精品亚洲av在线观看| 国产av一区二区精品久久| 欧美国产精品一级二级三级 | 久久ye,这里只有精品| 色5月婷婷丁香| 国产黄片视频在线免费观看| 欧美最新免费一区二区三区| 亚洲四区av| 国产黄色免费在线视频| 韩国高清视频一区二区三区| 国产熟女午夜一区二区三区 | 18+在线观看网站| 国产日韩一区二区三区精品不卡 | 免费黄频网站在线观看国产| 亚洲内射少妇av| 人人妻人人看人人澡| 观看美女的网站| 看十八女毛片水多多多| 性色avwww在线观看| 男女免费视频国产| a 毛片基地| 日产精品乱码卡一卡2卡三| 久久精品久久久久久久性| 日本午夜av视频| 老熟女久久久| 久久久久久久久大av| 一个人免费看片子| 啦啦啦在线观看免费高清www| 久热这里只有精品99| 丰满迷人的少妇在线观看| 女人久久www免费人成看片| 国产成人a∨麻豆精品| av天堂中文字幕网| 人妻人人澡人人爽人人| 欧美xxⅹ黑人| 曰老女人黄片| 亚洲欧美一区二区三区黑人 | 国产日韩欧美亚洲二区| 丰满人妻一区二区三区视频av| 又黄又爽又刺激的免费视频.| 久久久久视频综合| 国产免费又黄又爽又色| 一级片'在线观看视频| 日韩一区二区视频免费看| a级毛片免费高清观看在线播放| 在线观看免费视频网站a站| 国产亚洲91精品色在线| 亚洲第一av免费看| 男女边吃奶边做爰视频| 日韩在线高清观看一区二区三区| 成人影院久久| 国产毛片在线视频| 国产高清有码在线观看视频| a级片在线免费高清观看视频| 丰满乱子伦码专区| 午夜福利,免费看| 大香蕉97超碰在线| 26uuu在线亚洲综合色| 国产午夜精品一二区理论片| 亚洲va在线va天堂va国产| 亚洲熟女精品中文字幕| 国产欧美日韩精品一区二区| videossex国产| 国产午夜精品久久久久久一区二区三区| 国产精品麻豆人妻色哟哟久久| 国产高清有码在线观看视频| 日韩人妻高清精品专区| 乱码一卡2卡4卡精品| 国产综合精华液| 91久久精品电影网| 99久久人妻综合| 18禁在线播放成人免费| 国产成人精品无人区| 熟妇人妻不卡中文字幕| 亚洲av电影在线观看一区二区三区| 午夜福利在线观看免费完整高清在| 18禁动态无遮挡网站| 精品一区在线观看国产| 精品熟女少妇av免费看| 中文字幕免费在线视频6| 国产亚洲最大av| 欧美xxⅹ黑人| 蜜桃久久精品国产亚洲av| 国产亚洲最大av| 亚洲欧美精品自产自拍| 亚洲美女搞黄在线观看| 久久久久人妻精品一区果冻| 免费观看av网站的网址| 成年av动漫网址| 国产精品一区二区在线不卡| 中文字幕人妻丝袜制服| 中文字幕免费在线视频6| 久久影院123| 高清黄色对白视频在线免费看 | 国产亚洲午夜精品一区二区久久| 大陆偷拍与自拍| 国产一级毛片在线| 91久久精品国产一区二区三区| 永久网站在线| 亚洲av在线观看美女高潮|