• <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 ?

    The Impact of Bias Row Noise to Photometric Accuracy:Case Study Based on a Scientific CMOS Detector

    2024-03-22 04:11:48LiShaoHuZhanChaoLiuHaonanChiQiuyanLuoHuaipuMuandWenzhongShi

    Li Shao , Hu Zhan,2 , Chao Liu,3,4, Haonan Chi, Qiuyan Luo, Huaipu Mu, and Wenzhong Shi

    1 National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100101, China; shaoli@nao.cas.cn

    2 Kavli Institute for Astronomy and Astrophysics, Peking University, Beijing 100871, China

    3 University of Chinese Academy of Sciences, Beijing 100049, China

    4 Institute for Frontier in Astronomy and Astrophysics, Beijing Normal University, Beijing 100875, China

    Abstract We tested a new model of CMOS detector manufactured by the Gpixel Inc, for potential space astronomical application.In laboratory, we obtain some bias images under the typical application environment.In these bias images, clear random row noise pattern is observed.The row noise also contains some characteristic spatial frequencies.We quantitatively estimated the impact of this feature to photometric measurements, by making simulated images.We compared different bias noise types under strict parameter control.The result shows the row noise will significantly deteriorate the photometric accuracy.It effectively increases the readout noise by a factor of 2–10.However, if it is properly removed, the image quality and photometric accuracy will be significantly improved.

    Key words: instrumentation: detectors – methods: statistical – techniques: image processing

    1.Introduction

    Since the time of digitalization, CCD (charge coupled device) detectors have played an essential role in astronomy(Mackay 1986;Janesick 2001;Howell 2006;Lesser 2015).But after the invention of CMOS (complementary metal oxide semiconductor) image detector (Fossum 1997), CMOS has attracted more and more attentions in astronomy due to its advantages(Qiu et al.2013;Wang et al.2014;Mendikoa et al.2016; Jorden et al.2017; Shugarov 2020; Ardilanov et al.2021; Greffe et al.2022; Liu et al.2022), especially in time domain area (Pratlong et al.2016; Niu et al.2022; Song et al.2022).It also shows the potential to replace CCD (see e.g.,Gach et al.2022).

    During the development of the Chinese Space Station Telescope(CSST),both CCD and CMOS were considered as possible options for the CSST Survey Camera (CSC).The CSC is designed for large area optical survey.It uses 30 detectors to form a giant 2.3 billion pixel array to cover a field of view of more than one square degree.The optical survey will cover wavelengths ranging from 250 to 1000 nm.The detectors are required to have high quantum efficiency,low readout noise, low dark current and large dynamic range to fit for the proposed wide field blind survey (Zhan 2011,2018, 2021).In order to test whether the state-of-art CMOS detectors can satisfy the scientific requirement, we perform series of tests in the laboratory at the National Astronomical Observatories,Chinese Academy of Sciences,in Beijing.The detector we tested is an HR9090BSI scientific CMOS(sCMOS) detector5The detector we tested is a customized one.The standard version of this model is officially named as “GSENSE1081BSI.” A Chinese exoplanet searching project called “Earth 2.0” (Song et al.2022) uses the standard version.manufactured by the Gpixel Inc.6official website: https://www.gpixel.com/.This model is under active development in last few years.The Gpixel Inc.also produces some other science level CMOS detectors for (potential) astronomical applications (Shugarov 2020; Gill et al.2022).

    The readout noise is one of the most important parameters to evaluate the performance of a detector.The bias noise properties do have subtle consequences to final photometry or spectroscopy quality(Jiménez Mu?oz et al.2021).Only a few studies were done on CMOS detectors in astronomy field (Basden 2015; Basden &Morris 2016; Greffe et al.2022), probably due to limited application cases.More can be found in non-astronomy scientific applications(Diekmann et al.2017;Matsiaka&Plakhotnik 2021).In this paper,we will study the bias noise properties of this detector and its photometric performance with simulated images.Since this model is still under development, the content of this paper only reflects the status before the end of 2022.

    2.The Noise Properties of CMOS Bias Image

    2.1.The CMOS Detector and the Bias Images

    Figure 1.The setup of the detector test bench.

    The bias images were taken from an HR9090BSI sCMOS detector (serial number HR9090BSI-A10LV-SV-22001JC-0027).This one is designed to optimize the performance in visible bands.The pixel physical size is 10 μm×10 μm.The detector supports on-chip 16 bit column-parallel analog-todigital converter (ADC), with up to four low voltage differential signalling (LVDS) channels for image output.The detector uses rolling shutter.The supported minimal exposure time is about 0.6 ms, which is used for taking bias images.The readout time is about 3 s when using two LVDS channels at frequency of 250 MHz.The raw image has 8976×9222 pixels.On both left and right sides, there are 34 columns at the edge of array designed as “electrical black”(EB) region.The readings of these pixels only come from electronics.In principle,they can be used to calculate bias level as well (similar to overscan region in CCD case).However in practice we found the values in EB region were exceptionally large, and therefore not useful for our analysis.In this paper,we only use the central 8900×9120 pixels (“active” region).

    We prepared a test bench (Figure 1) for the test of CSC detectors in an ISO 7 clean room.It can satisfy multiple test requirements (e.g., pixel response at different wavelengths).The optical system was kept in a light tight box, which was installed on a vibration controlled optical table.All devices emitting light(e.g.,running indicator)were carefully separated.Light intensity monitor system was also used when necessary.It makes sure the bias and dark images are not contaminated by any external light source.The detector was installed in a vacuum dewar,cooled with liquid nitrogen.The temperature of the detector was kept at about 188 K, with typical temperature fluctuation no more than 1 K, similar to the designed on-board environment.

    The bias pattern is not very sensitive to the temperature.The variation of temperature has negligible contribution to the bias noise.The backend test board was also provided by the Gpixel Inc.(serial number 008).The test was performed in full dark environment to avoid external light contamination to the bias images.Given the typical dark current of ≯0.02 e-/s/pixel and high readout frequency, the dark current contribution during readout is negligible.In this paper, we use 30 bias images that were continuously taken on 2022 September 5.We note that other data sets show similar properties.

    In this paper we will use ADU for simplicity.The gain value(in astronomical definition) is about 1.3 e-/ADU (measured from photon transfer curve7In the CMOS detector, each pixel has its own gain value.The traditional photon transfer curve technique only gives a typical value.).Readers can convert all the results with this gain if they prefer unit of electrons.

    Figure 2.Left: master bias.Middle: individual bias image.Right: noise residual image (middle–left).

    2.2.Noise Residual Images and Row Noise

    We first stacked the 30 bias images to make a master bias.The stacking was performed pixel by pixel.In each pixel, we sorted the 30 values, excluded the top two and bottom two values to avoid any possible cosmic ray contamination, and then took the average of the rest.We note if we do not exclude any data or exclude more data,the result does not significantly change.On the master bias,we identified a pixel as“bad”if its value exceeded the 7σ threshold of typical(median)bias value.There are totally 52565 pixels marked as “bad,” including two rows.Most defected pixels are excluded in this way.The master bias has a strong feature of horizontal and vertical lines as well as square blocks.A similar feature was also observed in the other CMOS model with large arrays (e.g., Wang et al.2014; Karpov et al.2020).

    The master bias was then subtracted from each individual bias image to remove the fixed pattern,leaving only the noise fluctuation.Hereafter we call such a fixed-pattern-corrected bias image as “noise residual image.” On the right panel of Figure 2 there shows one example.The horizontal stripes can be clearly observed.We calculated the average value and the standard deviation in each row or column on the noise residual image.The bad pixels were excluded from the calculation.We merge all the row and column values from the 30 noise residual images together to make statistical plots(Figures 3 and 4).

    The result shows that the row-to-row variation(σ=3.59 ADU) is much larger than column-to-column variation(σ=0.05 ADU).It means the bias images contain spatially correlated noise, which is visually represented as horizontal stripes in noise residual image.The standard deviation in each column is comparable with the pixel value noise(which is based on Gaussian fitting of the pixel histogram,σ=4.46 ADU)and nominal“readout noise”(which is from the standard deviation of noise residual image, σ=4.77 ADU).Therefore the row-to-row fluctuation is the dominant component of the total readout noise.The standard deviation in each row is actually quite low (typically σ=2.84 ADU),implying that the detector may have the potential to decrease the overall readout noise to less than 3 ADU.

    We notice that the noise distribution is always non-Gaussian.The pixel-wise noise distribution is known to be somehow non-Gaussian.An internal technical report from the Gpixel Inc.(private communication) shows that the histogram of single pixel noise values has a long tail at the high end.A similar behavior was also reported in other CMOS models(e.g.,Ishida et al.2018).

    We calculated the pixel-to-pixel cross-correlation coefficient as a function of spatial shift(both x and y direction).The result is displayed in Figure 5.The result clearly shows the correlation in each row (i.e., x direction) is strong (>0.5).There are also many horizontal lines with weak but statistically significant correlations, positively or negatively.Hereafter in this paper we call such a feature as row noise (RN).We note that the pattern also shows some periodicity in the y direction,implying that the noise also contains characteristic frequency components.

    In order to confirm the existence of the characteristic spatial frequencies in the y direction, we calculated the spatial power spectra of the bias noise in Figure 6.The power spectrum is defined as the squared absolute amplitude of the Fourier transform component at given spatial frequency.We calculated the power spectrum in each row (or column) and derived the stacked spectrum with 20% trimmed mean value for all rows(or columns) in each individual bias image.Then we obtained 30 spectra(illustrated in gray lines)and derived the average of them(red line).In this way,we can reject any potential outliers(due to bad pixels) to get a robust estimation of the intrinsic power spectrum.

    Figure 3.Top right:pixel value histogram of all 30 noise residual images(fixed-pattern-corrected bias images).Bottom left:average value versus standard deviation in each row (the Pearson correlation coefficient is displayed).Top left: histogram of average (in each row)values.Bottom right: histogram of standard deviation (in each row) values.The red dashed line is Gaussian fitting.

    We find that the noise in each row(top panel of Figure 6)has a significant low frequency component, consistent with “1/f noise” (see e.g., Tian & El Gamal 2000).The noise in each column(bottom panel of Figure 6)has flat spectrum except that there are five characteristic frequencies (0.048, 0.133, 0,181,0.229, 0.410, in units of pixel-1).It may be caused by the power supply fluctuation, which is quite common in sCMOS detectors using column parallel readout architecture (i.e., each row is read in sequence, see Mikkonen 2016; Gilroy 2020).The readout could be affected by internal or external electrical circuits which has characteristic frequencies.However,we note that the white noise component is still the dominant noise source (the observed white noise component is equivalent to a noise source with rms of ≮4 ADU).

    2.3.Image Correction to Remove Row Noise

    The row noise, or more generally stripe noise, is commonly observed in images in remote sensing and/or taken by CMOS detectors (e.g., Simpson et al.1995; Rakwatin et al.2007;Gilroy 2020; Wei et al.2020).There are many developed algorithms to remove the stripe noise, for example, histogram matching with facet filter (Rakwatin et al.2007), midway histogram equalization (Tendero et al.2010), low-rank-based single-image decomposition (Chang et al.2016), group sparsity based regularization model (Chen et al.2017), wavelet deep neural network (Guan et al.2019), etc.It is also observed in some astronomical data (e.g., Schlawin et al.2020, for JWST NIRCam in time-series).The problem is not new and there are many possible solutions.Therefore it will be useful to investigate the properties of row noise corrected images as well.

    Figure 4.Same as Figure 3 but in columns.

    The origin of row noise in CMOS is complicated(Gilroy 2020).In this paper, only additive (or external signal independent) row noise is discussed (as only bias images are involved).8Actually, in another imaging test, we find there is row direction crosstalk,i.e.,signal dependent row noise,when there are many saturated pixels in a row.But it only appears when there is significantly saturated,i.e.,very bright,target.It does not affect the main conclusion of this paper,where we care more about faint objects with low signal-to-noise ratios.If we add the crosstalk in our simulation (presented in next section), the result will never be better.Our focus in this paper is not the correction method itself either.Therefore we use a very simple and straightforward method: subtracting the averaged empty pixel value in each row.

    To start, it is important to define the empty pixels, i.e., the image area not exposed to external light9Image area exposed to homogeneously distributed background light but not affected by astronomical objects may also be used for this purpose.However,in real application, such exercise has limitations: e.g., it is difficult for large extended objects (e.g., Magellanic Clouds) or over-crowded region (e.g., star cluster)where hardly any pixel is not contaminated by targets.In this paper,we do not discuss the algorithms to overcome such problems.Nevertheless, if used,it should be tested in real applications.On the other hand,our solution is the simplest way to avoid such problem..In case of HR9090BSI, the EB region (34 columns on each side) could be used for this purpose, after proper development in future(suggested by the Gpixel Inc.).Or alternatively,it is possible to mask some active pixels with opaque material to make an optically blind area at the edge of image.This option was proposed by the Gpixel Inc.(private communication) as well.In this paper,we assume the former option will be available.So we test the feasibility of this algorithm here.We use the leftmost 34 and rightmost 34 pixels of the active region (to mimic the size of EB)in each row to calculate the row average.We subtracted the average value of the 68 pixels in each row of the noise residual image.A similar algorithm is often used in CCD image processing, in which the reference empty area is from overscan instead of physical pixels (e.g., Wolf et al.2018).In Figure 7 the noise residual image before and after the correction are compared.Hereafter, we call these corrected images as “RN-corr” images.

    The row noise correction significantly reduces the readout noise (from 4.77 ADU to 2.89 ADU).The horizontal stripes become much less prominent visually.Similar power spectrum and correlation coefficient analysis were performed on RN-corr images for comparison (see Figures 8 and 9).

    Although there is still some small pixel-to-pixel correlation residual left in each row, the amplitude is greatly suppressed(Figure 8).The strength of the power spectrum in vertical direction(~6×104,Figure 9)is about one third of the original one (~2×105, Figure 6), becoming comparable with the horizontal one.Though as side effect,the power spectrum is no longer flat.Some characteristic frequencies are still there (e.g.,at 0.048 pixel-1),but their amplitudes are also greatly reduced.All these results demonstrate that the correction is very helpful.In next section, we will include the corrected images into our analysis as well.

    3.The Impact of Row Noise to Photometry

    3.1.The Simulated Noise Residual Images

    Here we use simulated images with galaxies and stars to better quantify the effect of row noise in astronomical application.We always simulate master bias corrected images(i.e.,object added to the noise residual image).

    For RN images, we use cutouts from original CMOS noise residual images.We randomly pick sub-regions of 196×196 pixels.Each random block must contain no more than 10 bad pixels and all bad pixels were set to zero.Random blocks should never overlap with each other so that the same noise pattern would not appear twice.There are totally 30446 random cutouts taken from 30 images.We also made cutouts for RN-corr images at the same positions for controlled comparison.

    For comparison, we also simulated pure Gaussian noise images.We made two sets of noise residual images with σ=4.77 ADU (Gauss-high) and σ=2.89 ADU (Gauss-low)respectively.They were used to compare with RN and RN-corr images respectively, in sense of nominal readout noise.We do not only observe the effect of overall readout noise reduction,but also the effect of RN itself.Figure 10 compares the different noise residual types.

    Figure 5.Pixel value correlation coefficient as a function of spatial shift.

    3.2.Real Galaxy Models

    Real galaxies were added to the simulated noise residual images.The image cutouts of real galaxies were directly taken from the website of the Legacy Survey10https://www.legacysurvey.org/(Dey et al.2019).15 galaxies were randomly picked.For simplicity and better signal-to-noise ratio, all of the three bands were merged together.Source detection and segmentation were performed to remove all nearby neighboring objects (except one galaxy α=187.5740, δ=73.0354 where a possible foreground star was kept to mimic the effect of possible confusion in real pipeline photometry).All pixels out of the target galaxy segment were set to zero.The cutout images were then normalized to unity total flux.Figure 11 shows all the galaxies we used.

    When adding the galaxy model to noise residual cutout, we randomly picked one model from the 15 models,multiplied by a random flux scale between 0.1 and 10.We use the same set of models for each type of noise residual image,to make sure that the difference only comes from noise residual.The left column of Figure 10 shows one example of such an image set.

    3.3.Sérsic Models

    Figure 6.Spatial power spectra of bias noise in rows (top)and columns (bottom).Grey lines are from individual bias images.Red lines are the average of them.In bottom panel, the values of five characteristic frequencies are noted.

    Sérsic model is a good way to parameterize the light distribution of elliptical galaxies, often used in simplified simulations (e.g., Holzschuh et al.2022).We use python package astropy to make 2D Sérsic model images,with random parameters listed in Table 1.Some examples are shown in Figure 12.Since the Sérsic model extends to infinite radius,we made a segment (consistent with an elliptical aperture) which included 90% of the total flux for further photometric measurement.Different from real galaxy models,Sérsic models have larger variability in galaxy size, probably more representative for field galaxies in a blind survey.The middle column of Figure 10 shows one example for different noise residual background.

    3.4.Star Models

    To simulate stars, we use 2D Gaussian model with σ=1.13 pixels (FWHM 2.66 pixels), which is the designed optical resolution (dynamic) of CSC.The model values were first assigned to a 0.05 pixel resolution sub-grid and then summed together to recover the true pixel value.Random amplitudes (between 10 and 103.5) and x/y center offsets(between -0.5 and 0.5) were used for each simulated star.Some examples were displayed in Figure 13.A circular aperture with radius of 4 pixels was used as segment for further photometric measurement.The right column of Figure 10 shows one example for different noise residual background.

    3.5.The Photometry and Basic Parameters

    We used the python package photutils v1.5 to measure several basic photometric parameters: position, elongation,position angle of the major axis, segment flux, Kron flux and circular aperture flux.

    The position measurement is very important in astrometry,i.e., determining the actual position of the object.Many other more complicated photometric parameters are sensitive to position measurement accuracy.The elongation is defined as the ratio between the length of major axis and the length of minor axis.It is often used as an indicator of the inclination angle of spiral galaxy.The position angle of the major axis is defined as the angle between the major axis and y-axis direction.It is also an important parameter to describe how the galaxy is positioned.Both the elongation and position angle are important morphological parameters,especially useful in weak lensing shear measurement (although the actual algorithm is more complicated).

    Figure 7.Before (left) and after (right) the row noise correction, displayed in the same scale.

    The segment flux is the sum of pixel values within the segment.It is a model-independent flux measurement,which is believed to be a robust estimation of the total flux.The Kron flux(Kron 1980)is another parameter to calculate the total flux.It sums up the flux within an elliptical aperture (i.e., assuming the galaxy shape can be represented by an elliptical).The parameters of the ellipse are determined from the first-order and second-order moment values.We use the default parameter in photutils.Comparing with segment flux, Kron flux is more sensitive to position and morphological measurements, but more reasonable in sense of measuring the total flux of a(generally symmetric)galaxy.It is widely used in many optical surveys (e.g., Guo et al.2013; Chambers et al.2016).The circular aperture flux was also calculated, which was preferred for stars.We always used the circular aperture with radius of 1.5 FWHM (3.99 pixels), which was sufficiently large to include most flux of a point source (more than 99.9% in our case) and was not too large to include too many noisy empty pixels.Most astrophysical studies require accurate flux measurement.It is important to check the reliability of the flux measurement.

    We used the fixed segment, i.e., the segment of the object(either galaxy or star) was fixed to the model image segment mentioned in previous subsections.If we use free segment or use resampled image,the conclusion does not change.We also note that Poisson resampling of the input flux was not used in our simulated images.In this way we make sure the difference only comes from the bias noise fluctuation within the same area.We will make a direct comparison between different noise types at the same nominal“readout noise”level and“signal-tonoise ratio.”

    3.6.Results

    We define the measurement error as the measurement difference between simulated image (model + noise) and pure model image, i.e., we treat the pure model result as “true value.”If the measurement is flux or elongation,the difference will be normalized by the “true value” and presented in percentage.

    Figure 8.Similar to Figure 5 but for row noise corrected images.

    Figure 9.Similar to Figure 6 but for row noise corrected images.

    Table 1 Random Parameters for Sérsic Models

    Figure 14 shows the measurement error as a function of average pixel flux in segment,which is defined as the total flux(in ADU) within the detection segment divided by the number of pixels of the segment.Here the position offset is defined as the absolute position offset with the sign of Y offset (i.e.,Each data point is binned from 1000 measurements.Error bars show the variation within the bin.Statistically, the mean value is accurate even for RN.But at any input brightness, the uncertainty of RN measurements are clearly worse than the Gaussian counterparts (note they have the same “readout noise”).Real galaxies and Sérsic models have similar results.

    Figure 15 shows the distribution of measurement error of all real galaxy images for six parameters.It shows the performance of RN is the worst for all parameters.The measurement error is especially large for Y position, segment flux and Kron flux.After correction, the performance of RN-corr is generally between Gauss-high and Gauss-low cases.It suggests that the row noise correction is not perfect but still very useful.

    The distribution is sometimes not symmetric.It may be related to the underlying algorithm and properties of the input galaxy templates.Such asymmetry does not affect our conclusion.Interestingly, the Y position uncertainty in RN case is significantly worse than X position.It is somehow expected, given the anisotropic nature of row noise.

    To better quantify the difference between different noise types, we plot the measurement uncertainty (the standard deviation of measurement error within the bin)as a function of average pixel flux in segment.Figures 16–18 show the results of all four noise types combined with three different models.

    The result is similar.Except the position angle of Sérsic images,RN is significantly worse than all other cases.RN-corr is much improved, comparable with Gaussian (between the high and low noise cases).The effect is much prominent in galaxies(including real galaxies and Sérsic models) than stars.It is probably due to the fact that the flux measurement apertures of stars are usually much smaller than that of galaxies.More pixels mean higher noise contribution from bias noise, and hence more vulnerable to row noise.It also clearly demonstrates that the correction is useful but the corrected result is still not as good as pure white noise.

    Figure 10.The same object with different noise residual as background (displayed in the same color scale).Left column: real galaxy model; middle column: Sérsic model; right column: star model.The images are zoomed in for better visibility.

    Figure 11.Real galaxy models used in simulation.Right ascension and decl.of the image center are shown for each galaxy (some galaxies are not centered).

    Figure 12.Example Sérsic models used in simulation.Blue line shows the segment which contains 90% of total flux.

    By comparing the flux uncertainties with the same input brightness,it is found that the uncertainty of RN case is about 3–10 times larger than Gauss-high for galaxies,and about 2 times larger for stars.In this paper,our test strictly controls the variables,leaving only the bias noise as the only parameter.So the effect of row noise we measured here can be directly translated into the increase of effective readout noise,by a factor of 2–10,depending on the target type.If we compare RN with Gauss-low, the difference is larger.

    Figure 13.Example Gaussian models used in simulation.All panels are in the same scale.The images are zoomed in.

    4.Discussion and Summary

    In this paper, we studied the noise properties of bias images taken from an HR9090BSI sCMOS detector.We found a strong pixel cross-correlation in each row, i.e., the row noise,and some minor fluctuations with fixed vertical spatial frequencies in the column direction.A simple row noise correction was applied to the real bias image,leading to a much lower readout noise and much smaller pixel-to-pixel correlation.We made a set of simulated galaxies and stars with different noise residual images as background.A series of photometry tests were performed.The results suggest that the row noise feature can significantly deteriorate the overall photometry performance, especially for galaxies.It effectively increases the readout noise by a factor of 2–10, depending on the type of object.

    Figure 14.Measurement error as a function of average pixel flux in detection segment(see text)for simulated stars.The blue data points are slightly shifted along Xaxis for better visibility.

    We demonstrate that the existence of row noise pattern, i.e.,pixel-to-pixel correlation, could greatly increase the photometric uncertainty.Such an effect is not simply due to the increase of nominal readout noise.In fact,at the same nominal readout noise level,the row noise case performs much worse than its Gaussian noise counterpart.It again proves the importance of pixel-to-pixel independence in astronomical high-precision photometry.It also makes a clear warning to anybody working on observation planning or camera design.The nominal“readout noise”may be misleading when the detector shows strong pixel-to-pixel correlation (no matter what the origin of such a correlation is).It may underestimate the required observation time for a given photometric accuracy, leading to an ambiguous observation result, or overestimate the performance of camera, leading to a failure of achieving the original design goal.

    We also note that, given the fact that the row noise is very common in sCMOS detectors, it is necessary to seriously consider the destriping method, either implemented in hardware or adopted at the data processing stage with efficient and robust algorithm.Our result suggests even the simplest destriping algorithm can make prominent improvement.Any future astronomical project using sCMOS detector should be aware of these facts.

    Our analysis also has limitations,although we note that some of them are out of the scope of this paper.First,our analysis of row noise is totally phenomenal.We treat the camera as a black box (due to some reasons).In order to fully understand the origin of row noise in this specific CMOS model,we will need more knowledge of the detector system, including the manufacturing details of the chip and the readout electronics.Second, we only consider the additive signal-independent row noise.Other multiplicative and/or signal-dependent noise may also affect the photometry.All these additional noise components will further worsen the photometry result if not properly corrected.In order identify these noise components,we will need more data, especially images from real astronomical observations.We note that in near future the“Earth 2.0” project (Song et al.2022) may give us a good chance.Third, this paper is focused on bias noise only.Our result is a good quantitative evaluator of the image readout performance.However, as a side effect, it makes the result looks too “exaggerated” when comparing to real situation.In practice,the bias noise usually only contributes a small fraction of the total photometric uncertainty.Many noise sources, such as Poisson noise,dark current,sky background and stray light,will make major contributions.Readers should be cautious when interpreting our results.To make a more realistic analysis of the photometric uncertainties,data from a more sophisticated simulation (e.g., Fang et al., in preparation, specifically designed for CSST11CSST simulation v2.0 release: https://csst-tb.bao.ac.cn/code/csst_sim/csst-simulation/-/tree/release_v2.0.) should be used.Finally, we do not discuss too much about the algorithm for row noise correction.When more and more CMOS data appears in the field of optical astronomy, it may be interesting to systematically study the effect of existing methods to high-precision photometry based on real observation images.

    Figure 15.The distribution of measurement errors for real galaxy models.Different bias noise types are displayed in different colors.

    Figure 16.The measurement uncertainty as a function of average pixel flux in detection segment for real galaxy models.Different bias noise types are displayed in different colors.

    Acknowledgments

    We thank the anonymous referee for her/his comments which help to improve this paper.

    This work is support by the National Key R&D Program of China No.2022YFF0503400.

    Figure 17.Same as Figure 16 but for simulated Sérsic images.

    The Legacy Surveys consist of three individual and complementary projects: the Dark Energy Camera Legacy Survey (DECaLS; Proposal ID #2014B-0404; PIs: David Schlegel and Arjun Dey), the Beijing-Arizona Sky Survey(BASS; NOAO Prop.ID #2015A-0801; PIs: Zhou Xu and Xiaohui Fan), and the Mayall z-band Legacy Survey (MzLS;Prop.ID #2016A-0453; PI: Arjun Dey).DECaLS, BASS and MzLS together include data obtained, respectively, at the Blanco telescope, Cerro Tololo Inter-American Observatory,NSF?s NOIRLab; the Bok telescope, Steward Observatory,University of Arizona; and the Mayall telescope, Kitt Peak National Observatory, NOIRLab.Pipeline processing and analyses of the data were supported by NOIRLab and the Lawrence Berkeley National Laboratory (LBNL).The Legacy Surveys project is honored to be permitted to conduct astronomical research on Iolkam Du’ag(Kitt Peak),a mountain with particular significance to the Tohono O’odham Nation.

    Figure 18.Same as Figure 16 but for simulated stars.

    NOIRLab is operated by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation.LBNL is managed by the Regents of the University of California under contract to the U.S.Department of Energy.

    This project used data obtained with the Dark Energy Camera (DECam), which was constructed by the Dark Energy Survey(DES)collaboration.Funding for the DES Projects has been provided by the U.S.Department of Energy, the U.S.National Science Foundation, the Ministry of Science and Education of Spain, the Science and Technology Facilities Council of the United Kingdom,the Higher Education Funding Council for England, the National Center for Supercomputing Applications at the University of Illinois at Urbana-Champaign,the Kavli Institute of Cosmological Physics at the University of Chicago, Center for Cosmology and Astro-Particle Physics at the Ohio State University, the Mitchell Institute for Fundamental Physics and Astronomy at Texas A&M University,Financiadora de Estudos e Projetos, Fundacao Carlos Chagas Filho de Amparo, Financiadora de Estudos e Projetos,Fundacao Carlos Chagas Filho de Amparo a Pesquisa do Estado do Rio de Janeiro, Conselho Nacional de Desenvolvimento Cientifico e Tecnologico and the Ministerio da Ciencia,Tecnologia e Inovacao, the Deutsche Forschungsgemeinschaft and the Collaborating Institutions in the Dark Energy Survey.The Collaborating Institutions are Argonne National Laboratory, the University of California at Santa Cruz, the University of Cambridge, Centro de Investigaciones Energeticas, Medioambientales y Tecnologicas-Madrid, the University of Chicago, University College London, the DES-Brazil Consortium, the University of Edinburgh, the Eidgenossische Technische Hochschule (ETH) Zurich, Fermi National Accelerator Laboratory, the University of Illinois at Urbana-Champaign, the Institut de Ciencies de l’Espai (IEEC/CSIC),the Institut de Fisica d’Altes Energies, Lawrence Berkeley National Laboratory, the Ludwig Maximilians Universitat Munchen and the associated Excellence Cluster Universe, the University of Michigan, NSF?s NOIRLab, the University of Nottingham, the Ohio State University, the University of Pennsylvania, the University of Portsmouth, SLAC National Accelerator Laboratory, Stanford University, the University of Sussex, and Texas A&M University.

    BASS is a key project of the Telescope Access Program(TAP), which has been funded by the National Astronomical Observatories,the Chinese Academy of Sciences(the Strategic Priority Research Program “The Emergence of Cosmological Structures” grant # XDB09000000), and the Special Fund for Astronomy from the Ministry of Finance.The BASS is also supported by the External Cooperation Program of Chinese Academy of Sciences (grant # 114A11KYSB20160057), and the National Natural Science Foundation of China (grants #12120101003 and # 11433005).

    The Legacy Survey team makes use of data products from the Near-Earth Object Wide-field Infrared Survey Explorer(NEOWISE), which is a project of the Jet Propulsion Laboratory/California Institute of Technology.NEOWISE is funded by the National Aeronautics and Space Administration.

    The Legacy Surveys imaging of the DESI footprint is supported by the Director, Office of Science, Office of High Energy Physics of the U.S.Department of Energy under Contract No.DE-AC02-05CH1123, by the National Energy Research Scientific Computing Center, a DOE Office of Science User Facility under the same contract; and by the U.S.National Science Foundation, Division of Astronomical Sciences under Contract No.AST-0950945 to NOAO.

    ORCID iDs

    Li Shao https://orcid.org/0000-0003-2015-777X

    Hu Zhan https://orcid.org/0000-0003-1718-6481

    香蕉精品网在线| 国产免费视频播放在线视频| 欧美高清性xxxxhd video| 亚洲人与动物交配视频| 久久久久国产精品人妻一区二区| 亚洲最大成人手机在线| 久久久久久久午夜电影| 两个人的视频大全免费| 国产黄色免费在线视频| 免费播放大片免费观看视频在线观看| 日韩制服骚丝袜av| 色吧在线观看| 亚洲国产欧美在线一区| 亚洲一级一片aⅴ在线观看| 久久人人爽人人爽人人片va| av一本久久久久| 观看美女的网站| 成人无遮挡网站| 国产成人aa在线观看| 亚洲欧美日韩另类电影网站 | 在线免费观看不下载黄p国产| 禁无遮挡网站| 99久国产av精品国产电影| 精品少妇久久久久久888优播| 欧美性猛交╳xxx乱大交人| 色哟哟·www| 精品一区在线观看国产| 一级片'在线观看视频| 在线观看免费高清a一片| 大香蕉97超碰在线| 丝袜脚勾引网站| 国产亚洲av嫩草精品影院| 下体分泌物呈黄色| 亚洲精品国产av蜜桃| 日日啪夜夜爽| 一级爰片在线观看| 赤兔流量卡办理| 一本一本综合久久| 欧美国产精品一级二级三级 | 久久99热6这里只有精品| 国产欧美日韩一区二区三区在线 | 听说在线观看完整版免费高清| 精品一区二区免费观看| 丰满人妻一区二区三区视频av| 国产亚洲午夜精品一区二区久久 | 18+在线观看网站| 日韩欧美一区视频在线观看 | 国产综合精华液| 午夜亚洲福利在线播放| 精品久久久精品久久久| 男人爽女人下面视频在线观看| 亚洲国产精品成人综合色| av在线播放精品| 热99国产精品久久久久久7| 亚洲欧美清纯卡通| 在现免费观看毛片| 麻豆乱淫一区二区| 我要看日韩黄色一级片| 一级毛片 在线播放| av国产免费在线观看| 国产乱人视频| 亚洲经典国产精华液单| 国产av码专区亚洲av| 欧美+日韩+精品| 国产毛片a区久久久久| 精品国产露脸久久av麻豆| 神马国产精品三级电影在线观看| 亚洲第一区二区三区不卡| 97人妻精品一区二区三区麻豆| 丰满人妻一区二区三区视频av| 五月伊人婷婷丁香| 免费电影在线观看免费观看| 久久人人爽人人爽人人片va| 91久久精品国产一区二区三区| 夫妻午夜视频| 嫩草影院新地址| 亚洲欧美清纯卡通| 97精品久久久久久久久久精品| 国产精品久久久久久精品古装| 狠狠精品人妻久久久久久综合| 国产精品麻豆人妻色哟哟久久| 久久久欧美国产精品| 国产毛片在线视频| 久久鲁丝午夜福利片| 亚洲经典国产精华液单| 精品酒店卫生间| 身体一侧抽搐| av网站免费在线观看视频| 看十八女毛片水多多多| 亚洲经典国产精华液单| 国产女主播在线喷水免费视频网站| 日韩一本色道免费dvd| 少妇人妻一区二区三区视频| 18禁裸乳无遮挡动漫免费视频 | 校园人妻丝袜中文字幕| 亚洲婷婷狠狠爱综合网| 2021少妇久久久久久久久久久| 在线免费十八禁| 中文字幕亚洲精品专区| 日本一本二区三区精品| av在线亚洲专区| 人体艺术视频欧美日本| 男女边摸边吃奶| 成人国产麻豆网| 亚洲国产欧美在线一区| 亚洲最大成人av| 卡戴珊不雅视频在线播放| 在线 av 中文字幕| 日韩一区二区三区影片| 又爽又黄无遮挡网站| 又爽又黄a免费视频| 欧美高清成人免费视频www| 免费看不卡的av| 99视频精品全部免费 在线| 2018国产大陆天天弄谢| 99热6这里只有精品| 亚洲自偷自拍三级| 久久99热这里只频精品6学生| 最近中文字幕高清免费大全6| 一级毛片我不卡| 欧美3d第一页| 国产大屁股一区二区在线视频| 欧美人与善性xxx| 国产高潮美女av| 久久久久久久久久成人| 国产男女超爽视频在线观看| av福利片在线观看| 天天躁夜夜躁狠狠久久av| 国产成人91sexporn| 99久久中文字幕三级久久日本| 国产精品三级大全| 国产亚洲91精品色在线| 七月丁香在线播放| xxx大片免费视频| 69av精品久久久久久| 99热6这里只有精品| 久久精品国产鲁丝片午夜精品| 免费av观看视频| 波野结衣二区三区在线| 日韩av不卡免费在线播放| 久久99热这里只有精品18| 视频中文字幕在线观看| 午夜爱爱视频在线播放| 亚洲图色成人| 国产午夜精品一二区理论片| 夫妻性生交免费视频一级片| 91午夜精品亚洲一区二区三区| 免费电影在线观看免费观看| 最近中文字幕高清免费大全6| 亚洲人与动物交配视频| 国产一区有黄有色的免费视频| 大香蕉久久网| 国产极品天堂在线| 精品国产三级普通话版| 久久ye,这里只有精品| 亚洲精品久久午夜乱码| 午夜爱爱视频在线播放| 99热这里只有是精品在线观看| 最近的中文字幕免费完整| 久久久久久久大尺度免费视频| 精品人妻熟女av久视频| 晚上一个人看的免费电影| 欧美+日韩+精品| 男插女下体视频免费在线播放| 在线 av 中文字幕| 内地一区二区视频在线| 精品久久久噜噜| 国产黄片美女视频| 街头女战士在线观看网站| xxx大片免费视频| 国语对白做爰xxxⅹ性视频网站| 国产片特级美女逼逼视频| 在线亚洲精品国产二区图片欧美 | 精品视频人人做人人爽| 免费大片黄手机在线观看| 婷婷色麻豆天堂久久| 久久久成人免费电影| 观看美女的网站| 免费在线观看成人毛片| 极品少妇高潮喷水抽搐| 国产成人freesex在线| 亚洲精品日韩在线中文字幕| 国产精品一及| 亚洲欧美日韩卡通动漫| 久久人人爽人人片av| 欧美一级a爱片免费观看看| 国产黄片美女视频| 久久精品国产自在天天线| 观看美女的网站| 婷婷色av中文字幕| 男女无遮挡免费网站观看| 国产av码专区亚洲av| 亚洲国产精品专区欧美| 嫩草影院新地址| 嘟嘟电影网在线观看| 赤兔流量卡办理| 亚洲精品aⅴ在线观看| 亚洲自偷自拍三级| 最后的刺客免费高清国语| 亚洲国产成人一精品久久久| 高清视频免费观看一区二区| 婷婷色麻豆天堂久久| 又粗又硬又长又爽又黄的视频| 国产在视频线精品| 久久女婷五月综合色啪小说 | 校园人妻丝袜中文字幕| 色婷婷久久久亚洲欧美| 日日啪夜夜爽| 免费av不卡在线播放| 亚洲熟女精品中文字幕| 最近最新中文字幕免费大全7| 性插视频无遮挡在线免费观看| 欧美丝袜亚洲另类| 男女无遮挡免费网站观看| 久久鲁丝午夜福利片| 欧美日韩视频精品一区| 精品一区在线观看国产| 日韩 亚洲 欧美在线| 欧美潮喷喷水| 波野结衣二区三区在线| 日本色播在线视频| 狂野欧美激情性xxxx在线观看| 国产人妻一区二区三区在| 99久久精品一区二区三区| 日韩人妻高清精品专区| 国产一区二区亚洲精品在线观看| 春色校园在线视频观看| 99热这里只有是精品50| 看黄色毛片网站| 国产亚洲5aaaaa淫片| 在线观看国产h片| 一级毛片我不卡| 国产日韩欧美亚洲二区| 插逼视频在线观看| 伦理电影大哥的女人| 国产午夜福利久久久久久| 国产精品福利在线免费观看| 亚洲内射少妇av| 亚洲成人久久爱视频| 日韩在线高清观看一区二区三区| 精品人妻偷拍中文字幕| 国产亚洲最大av| 亚洲无线观看免费| av福利片在线观看| eeuss影院久久| 欧美一区二区亚洲| 91精品国产九色| 大香蕉久久网| 黄片wwwwww| 91在线精品国自产拍蜜月| 亚洲精品456在线播放app| 特级一级黄色大片| 国产 一区精品| 嫩草影院新地址| 久久久久久久国产电影| 国产成人精品福利久久| 日本免费在线观看一区| 色婷婷久久久亚洲欧美| 国产综合懂色| 18+在线观看网站| 大话2 男鬼变身卡| 亚洲国产欧美人成| 秋霞在线观看毛片| 国产精品人妻久久久久久| 我的老师免费观看完整版| 国产亚洲午夜精品一区二区久久 | 黄色一级大片看看| 日韩制服骚丝袜av| 日韩免费高清中文字幕av| 观看免费一级毛片| 尾随美女入室| 午夜日本视频在线| 在线观看美女被高潮喷水网站| 色视频www国产| 亚洲在久久综合| 在线观看一区二区三区| 好男人视频免费观看在线| 干丝袜人妻中文字幕| 久久久久国产网址| 看非洲黑人一级黄片| 欧美国产精品一级二级三级 | 身体一侧抽搐| 一级二级三级毛片免费看| 成年免费大片在线观看| 不卡视频在线观看欧美| 2022亚洲国产成人精品| 久久99热这里只频精品6学生| 久久久久久久久久久丰满| 岛国毛片在线播放| 成人免费观看视频高清| 国产乱人偷精品视频| 免费看a级黄色片| 99九九线精品视频在线观看视频| 人妻少妇偷人精品九色| 欧美变态另类bdsm刘玥| 自拍偷自拍亚洲精品老妇| 午夜福利视频精品| 99热这里只有是精品50| 日本黄大片高清| 青春草亚洲视频在线观看| 91在线精品国自产拍蜜月| 九九爱精品视频在线观看| 精品久久久久久久末码| 欧美xxxx性猛交bbbb| 国产淫语在线视频| 日日啪夜夜撸| 久久韩国三级中文字幕| 天天躁狠狠躁夜夜躁狠狠躁| 亚洲七黄色美女视频| 青春草亚洲视频在线观看| 国产爽快片一区二区三区| 51午夜福利影视在线观看| 午夜免费鲁丝| 亚洲国产精品国产精品| 在线看a的网站| 激情五月婷婷亚洲| 男人舔女人的私密视频| 男女下面插进去视频免费观看| 亚洲av电影在线观看一区二区三区| 国产1区2区3区精品| 日本91视频免费播放| 天堂中文最新版在线下载| 一级毛片电影观看| 国产一级毛片在线| 午夜福利免费观看在线| 十分钟在线观看高清视频www| 中文字幕av电影在线播放| 在线观看人妻少妇| 久久久亚洲精品成人影院| 国产精品一区二区在线不卡| 免费黄网站久久成人精品| 国产精品久久久人人做人人爽| 国产不卡av网站在线观看| 在线观看免费午夜福利视频| 亚洲欧美一区二区三区久久| 一区二区日韩欧美中文字幕| 欧美日韩视频精品一区| 一级毛片我不卡| 日韩一区二区视频免费看| 只有这里有精品99| 国产亚洲欧美精品永久| 亚洲国产欧美网| 黑人巨大精品欧美一区二区蜜桃| 精品少妇黑人巨大在线播放| 人人妻人人澡人人爽人人夜夜| 最近最新中文字幕免费大全7| 热re99久久国产66热| 丝袜在线中文字幕| 国产亚洲精品第一综合不卡| 国产男女超爽视频在线观看| 中文精品一卡2卡3卡4更新| 久久精品亚洲熟妇少妇任你| 日韩视频在线欧美| 亚洲国产精品999| 美女主播在线视频| 久久ye,这里只有精品| 少妇人妻精品综合一区二区| av有码第一页| 欧美中文综合在线视频| 老司机靠b影院| 亚洲欧美一区二区三区黑人| 欧美国产精品一级二级三级| 少妇人妻精品综合一区二区| 亚洲伊人色综图| 欧美亚洲 丝袜 人妻 在线| 王馨瑶露胸无遮挡在线观看| 在线观看免费视频网站a站| 亚洲欧洲国产日韩| 97人妻天天添夜夜摸| 欧美激情极品国产一区二区三区| 精品午夜福利在线看| 妹子高潮喷水视频| 中文字幕人妻丝袜制服| 久久久久人妻精品一区果冻| 女人高潮潮喷娇喘18禁视频| 亚洲av中文av极速乱| 天天躁日日躁夜夜躁夜夜| 久久av网站| 18禁裸乳无遮挡动漫免费视频| 亚洲成人一二三区av| 男女免费视频国产| 色视频在线一区二区三区| 美女午夜性视频免费| 波多野结衣一区麻豆| 青青草视频在线视频观看| 日本欧美视频一区| 波多野结衣一区麻豆| 亚洲一区中文字幕在线| 大香蕉久久网| 久久午夜综合久久蜜桃| 免费黄网站久久成人精品| 欧美日韩成人在线一区二区| 一区二区三区乱码不卡18| 欧美黑人欧美精品刺激| 亚洲综合色网址| 三上悠亚av全集在线观看| 亚洲激情五月婷婷啪啪| 午夜福利影视在线免费观看| 中文字幕人妻丝袜一区二区 | 亚洲成人一二三区av| 日韩av在线免费看完整版不卡| 2021少妇久久久久久久久久久| 国产成人精品福利久久| 亚洲欧洲精品一区二区精品久久久 | 久久青草综合色| 一区福利在线观看| 成人国语在线视频| 欧美日韩亚洲高清精品| 国产在线免费精品| 久久久久精品国产欧美久久久 | 欧美激情极品国产一区二区三区| svipshipincom国产片| 丝袜喷水一区| 看非洲黑人一级黄片| 免费黄色在线免费观看| 人人妻人人澡人人爽人人夜夜| 色视频在线一区二区三区| 日本91视频免费播放| av免费观看日本| 国产亚洲av高清不卡| 男女免费视频国产| 一区二区三区四区激情视频| 久久久久人妻精品一区果冻| 波多野结衣av一区二区av| 婷婷色综合大香蕉| a 毛片基地| 精品国产露脸久久av麻豆| 久久影院123| 飞空精品影院首页| 伊人亚洲综合成人网| 老司机深夜福利视频在线观看 | 国产精品久久久久成人av| 欧美在线黄色| 亚洲av成人精品一二三区| 国产精品国产三级国产专区5o| 国产成人精品久久二区二区91 | 成年av动漫网址| 午夜日本视频在线| 一边摸一边抽搐一进一出视频| 国产精品一区二区在线观看99| 人人妻人人添人人爽欧美一区卜| 久久精品熟女亚洲av麻豆精品| av片东京热男人的天堂| 免费看av在线观看网站| 国产精品久久久久久人妻精品电影 | 免费高清在线观看视频在线观看| 国产精品熟女久久久久浪| 国产av精品麻豆| 精品少妇久久久久久888优播| 久久毛片免费看一区二区三区| 一级片'在线观看视频| 日日啪夜夜爽| 午夜老司机福利片| 超碰97精品在线观看| 亚洲av欧美aⅴ国产| 精品国产国语对白av| 亚洲精品美女久久av网站| 国产精品一区二区在线不卡| 久久婷婷青草| 少妇的丰满在线观看| 精品人妻一区二区三区麻豆| 日本色播在线视频| 又大又爽又粗| 亚洲色图综合在线观看| 国产成人午夜福利电影在线观看| 色综合欧美亚洲国产小说| 国产野战对白在线观看| 亚洲欧洲国产日韩| 国产精品蜜桃在线观看| 十八禁人妻一区二区| 久久精品亚洲熟妇少妇任你| 天天添夜夜摸| av卡一久久| 欧美激情极品国产一区二区三区| 91精品国产国语对白视频| 搡老乐熟女国产| 久久精品aⅴ一区二区三区四区| 十分钟在线观看高清视频www| 天堂8中文在线网| 欧美变态另类bdsm刘玥| 免费av中文字幕在线| 欧美在线黄色| 成年女人毛片免费观看观看9 | 午夜av观看不卡| 少妇 在线观看| 亚洲精品自拍成人| 亚洲精品日本国产第一区| 精品国产乱码久久久久久男人| 好男人视频免费观看在线| 久久久国产精品麻豆| 国产1区2区3区精品| 久久鲁丝午夜福利片| 国产成人av激情在线播放| 欧美激情 高清一区二区三区| 亚洲精品国产色婷婷电影| 两个人免费观看高清视频| 国产欧美日韩综合在线一区二区| 蜜桃国产av成人99| 中文欧美无线码| 黄网站色视频无遮挡免费观看| 久久99一区二区三区| 国产精品熟女久久久久浪| 久久精品国产亚洲av高清一级| 欧美人与性动交α欧美软件| 亚洲第一区二区三区不卡| 老司机亚洲免费影院| 天堂中文最新版在线下载| 99久国产av精品国产电影| 男女之事视频高清在线观看 | 免费久久久久久久精品成人欧美视频| 国产亚洲精品第一综合不卡| e午夜精品久久久久久久| 国产日韩欧美亚洲二区| 女人精品久久久久毛片| 精品亚洲成a人片在线观看| 桃花免费在线播放| 在线观看人妻少妇| 一本色道久久久久久精品综合| 日日啪夜夜爽| 一级毛片 在线播放| 免费黄频网站在线观看国产| 日韩大片免费观看网站| 极品少妇高潮喷水抽搐| 免费不卡黄色视频| 男女午夜视频在线观看| 日韩制服骚丝袜av| 久久精品国产a三级三级三级| 99热网站在线观看| videos熟女内射| 成年女人毛片免费观看观看9 | 777米奇影视久久| 精品视频人人做人人爽| 亚洲欧洲日产国产| 精品视频人人做人人爽| 欧美日韩亚洲国产一区二区在线观看 | 女性生殖器流出的白浆| 久久精品久久精品一区二区三区| 亚洲国产日韩一区二区| 亚洲色图综合在线观看| 欧美激情极品国产一区二区三区| 在线观看免费高清a一片| 美女高潮到喷水免费观看| 中文字幕av电影在线播放| videosex国产| 99九九在线精品视频| 亚洲成人av在线免费| 精品久久久精品久久久| 成年美女黄网站色视频大全免费| 我要看黄色一级片免费的| 国产成人精品久久二区二区91 | 国产精品麻豆人妻色哟哟久久| 日韩大片免费观看网站| 国产精品麻豆人妻色哟哟久久| 国产精品人妻久久久影院| 美女中出高潮动态图| 中国国产av一级| 国产在视频线精品| 男女免费视频国产| 亚洲欧美精品综合一区二区三区| 日韩人妻精品一区2区三区| 97精品久久久久久久久久精品| 欧美精品一区二区大全| 水蜜桃什么品种好| 视频在线观看一区二区三区| 一区二区三区激情视频| 免费观看a级毛片全部| 飞空精品影院首页| 久久精品久久久久久噜噜老黄| 国产亚洲一区二区精品| 在线观看免费日韩欧美大片| 性色av一级| 久久99精品国语久久久| 亚洲av综合色区一区| 日韩精品有码人妻一区| 嫩草影视91久久| 18禁裸乳无遮挡动漫免费视频| 亚洲国产看品久久| 午夜久久久在线观看| 国产伦理片在线播放av一区| 校园人妻丝袜中文字幕| 捣出白浆h1v1| 久久韩国三级中文字幕| 亚洲av电影在线进入| 高清欧美精品videossex| 欧美精品一区二区大全| 精品少妇一区二区三区视频日本电影 | 狠狠婷婷综合久久久久久88av| 国产免费一区二区三区四区乱码| 亚洲欧美一区二区三区黑人| 欧美日韩av久久| 母亲3免费完整高清在线观看| 狂野欧美激情性bbbbbb| 制服诱惑二区| 99久久人妻综合| 国产男人的电影天堂91| 久久精品aⅴ一区二区三区四区| av天堂久久9| 啦啦啦 在线观看视频| av卡一久久| 在线天堂最新版资源| 在线 av 中文字幕| 色精品久久人妻99蜜桃| 亚洲国产欧美日韩在线播放| 亚洲av成人精品一二三区| 高清在线视频一区二区三区| 可以免费在线观看a视频的电影网站 | 欧美日韩综合久久久久久| 久久久久精品性色| 亚洲美女黄色视频免费看| 久久久久久久大尺度免费视频| 人人妻人人澡人人看| 国产又爽黄色视频| 天堂8中文在线网| 女性生殖器流出的白浆| www日本在线高清视频|