设为首页| 收藏本站|

开源天文论坛  astrocn.org

 !forgotpw!
 !register!

QQ login

One step, quick start

!show!: 5117|!reply!: 4

后期入门学习记录-关于概念

[!share_url_copy!]
  • TA的每日心情
    开心
    2018-2-10 21:40
  • 签到天数: 1 天

    [LV.1]初来乍到

    165

    viewthread_userinfo_threads

    731

    viewthread_userinfo_posts

    8910k

    viewthread_userinfo_credits

    管理员

    Rank: 9Rank: 9Rank: 9

    viewthread_userinfo_credits
    893018

    最佳新人活跃会员热心会员推广达人宣传达人灌水之王突出贡献优秀版主荣誉管理论坛元老

    !poston!: 2017-2-5 14:39:59
    | !thread_show_all! |!read_mode!
    源自MDL Glossary,还有部分群内讨论内容,理解不准确、学的也不深入,记录一下欢迎指正~随时改慢慢加,未完待续

    Amount Amount表示作用量,PixInsight后期处理中的一个关键参数,公式:合成值=A×M+(1-A)×N,(A= Amount作用量参数、M=作用后的像素值、N=作用前的像素值)

    MDL Info
    info.JPG
    信息框中对应示例值:
    Cursor( X= 432, Y= 135 ), Rad= 5, Rad2= 13
    Pixel32846.000                       Magnitude 4.910
    Maximum32846.000               Intensity 317385.750
    Minimum222.000                   SNR 4835.745
    Median819.000
    Average4049.271                    Bgd Avg130.929Std
    Dev 6940.297                     Bgd Dev7.293
    Centroid( X= 432.059, Y= 135.041 )
    FWHM30.056"                        Flatness0.168
    在 Aperture 模式下
    第一行Cursor代表鼠标的位置也即内圆中心的位置,以及内圆Rad和外圆Rad2的半径。
    Pixel为中心点的值,Magnitude 经过 Calibration 选项的调节可以显示其准确值。
    Maximum、 Minimum、Median、 Average 、Std Dev 分别代表最大值、最小值、中值、平均值、标准差。都是用内圆计算。
    BgdAvg ,Bgd Dev 为由外环计算的背景平均值和标准差。
    Intensity为内圆各个象元值减去背景平均值后所有数值相加的总和。


    SNR的求法: Intensity 值除以 Bgd Dev 值后,再除以内圆象元的平方根。
    例如以上: SNR.png
    Centroid表示的是内圆中星象的重心位置。
    FWHM半高全宽,高斯函数拟合所得峰值1/2 的宽度。
    Flatness平滑度。0 代表星象为标准的圆,0.1 代表星象在一个方向(如 x 轴)比在另一个方向(如 y 轴)长 10%。这个值反应了星像的质量及焦距的好坏。


    BackG/Bgd
    BackG (The skyBackground/ Background):背景天光值,反映拍摄环境的好坏,判断照片质量的重要参数,城里上千,山上几十。


    Bias Frame 偏置帧
    A bias frame is a zero-length image taken using a CCDcamera with the shutter closed. It is used to compensate for pixel-specificoffsets in the CCD chip and camera electronics.
    偏置帧就是盖住镜头盖拍摄一张最短曝光时间的图像。它是用来补偿CCD/CMOS芯片和相机电路导致的特定像素数值偏移。


    Dark Current 暗电流
    Every CCD camera produces a certain amount of darkcurrent, which accumulates in the pixels during an exposure. The dark currentis produced by heat, and is a source of noise in images. The noise results fromrandomness in the dark current and pixel-to-pixel variations in the amount ofdark current. The pixel-to-pixel variations in dark current can be removed bysubtracting a dark frame, but the associated noise cannot be removed.
    CCD/CMOS相机都会产生一定数量的暗电流,每次曝光期间这些暗电流积聚在像素当中。暗电流是由热产生的,是图像的一种噪声源。这些噪声的结果就是随机性的暗电流和由于暗电流作用导致的像素之间读取差异。由暗电流导致的像素之间读取差异可以通过减去一个暗场来消除,但关联的噪声不能消除。


    Dark Frame 暗帧
    A dark frame is an exposure taken using a CCD camerawith the shutter closed. The duration of the dark frame should match that ofthe images being taken, and it should be made at the same temperature. It isused to compensate for pixel-to-pixel variations in the dark current.
    暗场就是盖住相机镜头盖进行拍摄。暗帧的曝光时间应该与亮场匹配,应当在相同温度下拍摄。暗场主要用来补偿暗电流作用导致的像素之间读取差异。


    Deconvolution 反卷积
    Deconvolution is the opposite of convolution. It is amathematical means of sharpening images, which is used to restore the originalappearance of an image before it was blurred by convolution. Since convolutionis equivalent to multiplication in the frequency domain, deconvolution isequivalent to division in the frequency domain. Deconvolution is more difficultto perform in practice, due to the possibility of dividing by zero, and theproblem of noise amplification.
    反卷积是卷积的逆向操作。是一种用来锐化图像的数学工具,用来重构被卷积模糊的原始图像细节。卷积相当于在频域中的乘法,反卷积相当于频域中的除法。由于噪声放大、除数为零等问题,反卷积在实践中的难度更大。


    Dither 抖动
    In astronomical imaging, dither is a small movement ofthe telescope between images, resulting in a slight offset in the target ineach image. During stacking the images are realigned, resulting in the pixeldefects moving around relative to the image. Statistical combine methods suchas Sigma Clip are then used to eliminate the bad pixels, resulting in a cleanerimage.
    在天文成像中,dither是在两次成像之间的望远镜的一个小的移动,从而在每个图像中目标的位置产生微小偏移量。在叠加图像重新对齐时,缺陷像素在图像中的相对位置也会移动。利用有关统计算法,比如用Sigma Clip来消除坏像素,得到一个干净的图像。(Ditering抖动让每张照片错开一点点,让CCD的底噪随机分布,目的是把CCD底噪随机化,叠加提高信噪比,直观的感觉就是dither叠加后背景噪声很均匀)


    Dynamic Range 动态范围
    For imaging, dynamic range is the range from brightestto darkest pixel value. A 16-bit CCD camera has a maximum possible dynamicrange of 0 to 65535. In practice, the dynamic range is between the noise floorto the saturation level, which is lower than the range of the A/D converter.
    对于图像成像,动态范围是指像素值从最亮到最暗的范围。16位CCD相机最大的动态范围为0至65535。在实践中,动态范围通常在本底噪声和饱和状态之间,并低于A/D转换器的测量范围。

    Flat-Field Frame
    平场
    Aflat-field frame is an image of an evenly-illuminated source. These frames areused to compensate for both pixel-to-pixel variations in sensitivity and theeffects of optical obstructions (vignetting and dust).
    平场是拍摄一个均匀照明光源得到的图像。平场是用来弥补灵敏度导致的像素读取差异和光学障碍物的影响(渐晕光损和灰尘)。

    Full-Width Half Maximum
    半高全宽值
    Alsoknown as FWHM. A measure of the size of a star or other image. Half maximummeans the distance from the center at which the brightness drops to one half ofthe center value. Full width means the measurement is given as a diameterrather then a radius. The FWHM of a point-source in an image, divided by two toget a radius, can be used as an estimate for the radius of the point-spreadfunction model.
    简称FWHM。衡量星点或其他图像的大小。半最大值的意思是亮度下降到一半时距离中心的值。全宽测量的是给定星点的直径而不是半径。图像中的点光源的半峰宽,除以2得到半径,可以作为点扩散函数的半径估计模型。
    (FWHM通常是判断照片质量的重要参数,推荐值在2左右。同等视宁下,通常B通道FWHM比较小,可作为后期处理参考帧)
    无标题.png

    Gain
    增益
    Gainfor an amplifier is the ratio of output / input. In a CCD camera, the gain determinesthe Photoelectrons per ADU; i.e. how many photoelectrons it takes to make theA/D converter output increase by one count.
    增益表示放大器的输出/输入比例。在CCD/CMOS相机中,增益决定每ADU光电子的数量;即多少个光电子才能使A/D转换器的输出增加1个计数。
    建议看椰风关于Gain, ISO与动态范围的讨论
    Gaussian
    高斯
    Themost common type of statistical distribution - the ¡±bell curve¡± distribution. The shape is verycommon because a sum of a large number of random distributions always tendstowards a Gaussian shape (due to the so-called central limit theorem). Theeffects of atmospheric seeing tend to impose a Gaussian shape on star imagesfor this reason. Also known as the ¡±normal¡± distribution.
    最常见类型的统计分布,¡±钟形曲线¡±的分布状态。高斯分布的形状是非常常见的,因为大量的随机分布的总和总是倾向于高斯形状(所谓的中心极限定理)。大气视宁度的影响往往会强加星点的高斯形状特征。高斯分布也通常称为¡±正态¡± 分布。

    Half-Flux Diameter (HFD)
    半流径(HFD)
    Ameasurement that is an improvement on the older Full-Width Half Maximummeasurement, invented by Larry Weber and Steve Brady. The half-flux diameter isthe diameter in pixels that contains half the energy in a star image. In otherwords, if you add up the ADU values (less the background) inside the diameter,and outside the diameter, you will get the same number. This measurement givesa very similar answer to FWHM, but it is much more robust in the presence ofseeing and noise, and can handle images that are so out-of-focus that they looklike "donuts". The Half-Flux Diameter also varies linearly with focusposition, except very close to best focus. This makes reliable autofocuspossible.
    对老的半高全宽测量方式的一种改进,由Larry Weber和Steve Brady发明。半流径是包含星点图像一半能量的像素直径。换句话说,如果你在直径内和直径外增加ADU值(低于背景值),你将得到一个相同的数值。这种测量方式将得到一个非常类似FWHM的结果,但在视宁和噪声面前这种HFD将显得更强大,并可以处理看起来像“甜甜圈”的失焦图像。HFD也将跟随焦点位置线性变化,除了非常接近最佳合焦位置时。这使得可靠的自动对焦成为可能。
    (HFD 的数值会比 FWHM 大1~2像素,另外还有HFR(Half Flux Radius) ,HFD = 2 * HFR)

    Histogram
    直方图
    Ahistogram is a graph plotting intensity on the horizontal axis and number ofpixels on the vertical axis. This shows the distribution of pixels withinvarious intensity ranges. The most common pixel value corresponds to thehighest point on the graph.
    直方图绘制的图形在水平轴上代表亮度(强度)值,垂直轴上代表像素数量。这表明在不同亮度(强度)范围内的像素分布。最大量的像素值对应于图上的最高点。


    ISO
    ISO
    Anumber indicating the camera sensor's sensitivity to light. The higher thesensitivity, the less light is needed to make an exposure. Originally thisreferred to the sensitivity of a film emulsion. In modern DSLRs the ISO settingis really a gain adjustment for the A/D converter.
    表示相机传感器对光的灵敏度的数字。灵敏度越高,曝光量就越少。原来这是指胶片感光乳剂的感光性。现代的数码单反相机的ISO设置实际上是对A / D转换器的增益调整。

    Kernel Filter
    内核滤波器
    Akernel filter works by applying a kernel matrix to every pixel in the image.The kernel contains multiplication factors to be applied to the pixel and itsneighbors. Once all the values have been multiplied, the pixel is replaced withthe sum of the products. By choosing different kernels, different types offiltering can be applied.
    内核滤波器的工作原理是将内核矩阵应用到图像中的每一个像素。内核中包含的乘法因子被应用到目标像素和相邻的像素。当所有值都完成乘法运算后,像素值被运算结果之和替代。通过选择不同的内核,可以应用不同类型的滤镜。


    Master Frame
    主帧
    The combination of all images in a group, that havebeen pre-calibrated using the appropriate bias (and, in the case of a group offlats, dark) groups. Once generated, calibration will no longer need to loadand combine the images in a group whenever that group is needed.
    已使用适当的偏置帧进行预先校准的一组图像的组合(当然,也可能是一组平场、暗场)。一旦产生Master Frame,在进行图像组合时将不再需要进行单独校准。


    Median 中值

    In a list of values, the median is the ¡±middle¡±value. For example, if the list is 5, 10, 15, 20, 105; the median is 15. Theaverage of the same list would have been 31. A median filter rejects any valueswhich are substantially different from the others, and so rejects ¡±impulsive¡± noise like hot pixels or cosmicray hits. When filtering Gaussian noise, however, taking the median value isabout 25% worse at suppressing noise than taking an average.
    在一个列表中的值,中位数是表示 ¡±中间值¡± 的值。例如,如果列表是5,10,15,20,105,中位数则是15。而同一个列表的平均值为31。中值滤波器将滤除任何有显著差异的值,所以能够滤除处于 ¡±脉冲¡± 范围内的脉冲型噪声,比如像热像素或宇宙射线这些。然而在滤除高斯噪声时,取中值比取平均值在抑制噪声方面要差25%。


    Noise 噪声
    Any measurement always contains a certain amount ofrandom variation called noise. There are many sources of noise in a CCD imagingsystem, including photon statistics, dark current variations, and readoutnoise. The ratio of signal to noise in a measurement is a fundamentalindication of accuracy. Since the signal-to-noise ratio is lower for fainterobjects in an image, the practical limit on sensitivity is determined by thenoise.
    任何测量都包含了一定量的随机变化,叫做噪声。CCD/CMOS成像系统的噪声来源有很多,包括光子统计、暗电流变化、读出噪声等。信噪比是衡量噪声水平的一个基本指标。如果图像中的暗弱目标信噪比很低,对灵敏度的限制实际上由噪声决定。


    Nyquist Sampling Theorem Nyquist采样定理
    Sampling a sound (using a microphone) or image (usinga CCD camera) can introduce a distortion into the signal called ¡±aliasing.¡±Aliasing takes many forms; for example the apparent backward motion of wagonwheels in western movies, which occurs because the movie is actually a sequenceof still frames. Nyquist showed that if the sampling rate is twice the highestfrequency component, then the original information can be reconstructed withoutaliasing. In the case of a CCD image, you need to have a pixel density twicethe highest spatial frequency in the image. In practice 2.5X is used tocompensate for the distance across the diagonal of the pixel. Note that properdisplay of a ¡±critically sampled¡± image requires a reconstruction filter - interpolation to a highersample rate - for proper image display. Also oversampling may be desirablebecause interpolation during image alignment can degrade resolution.Oversampling is also required if you intend to use deconvolution to increaseresolution; the final deconvolved image must also meet the Nyquist SamplingCriterion.
    采样声音(用话筒)或图像(使用一个CCD/CMOS相机)会引入失真的信号称为¡±失真¡±,失真有多种形式,例如电影中车轮明显的反转运动,因为这部电影实际上是一个静止的帧序列。奈奎斯特定理表明,如果采样率是最高频率的两倍,则原始的信息可以不失真的重建。对CCD成像来说,你需要像素密度2倍于图像的最高空间频率。在实践中通常使用2.5倍率来补偿像素对角线距离。注意正确的显示一个¡±临界采样¡± 的图像需要重建滤波器来插值到一个更高的采样率。也许过采样更值得去做,因为图像对齐时的插值会降低解析力。过采样是很有必要的如果你打算用反卷积来增强解析力;最终的反卷积图像也必须满足Nyquist采样准则。


    Pattern Noise 模式噪声
    Unlike CCD sensors, CMOS sensors are subject to fixedpattern noise. Each pixel in a CMOS camera has its own separate readoutamplifier, each of which can have slightly different characteristics. CCDcameras do not suffer from this effect because every pixel is read out by asingle amplifier (or for large high-speed sensors, a small number ofamplifiers).
    不同于CCD传感器,CMOS传感器存在固定模式噪声。一个CMOS中的每个像素都有自己单独的读出放大器,这些放大器可能会有一定差异。而CCD不受这种影响,因为CCD每个像素均由一个单一的放大器读出(对于大的高速传感器,也只有少量的放大器)。


    Photoelectrons Per ADU 光电子每ADU
    ADU stands for Analog-to-Digital Unit. For CCDcameras, the number of Photoelectrons per ADU tells you how many photoelectronsare required to increase the pixel value by one count. A photoelectron isproduced whenever a photon (light packet) is detected by the pixel. Also knownas Gain.
    ADU是指模数转换单元。对于CCD/CMOS相机,光电子/ADU值告诉你需要多少光电子才能让像素值增加1。每当像素检测到一个入射光子(光的最小组成)便产生1个光电子,通常也称为增益。(ADU代表星点亮度值,减去BackG背景天光值后,通常最小ADU:2000左右,最佳ADU:30000左右)


    Photon 光子
    Light is radiated in small energy packets calledphotons. Although photons are often visualized as particles, they also have thecharacteristics of a wave. Photons from a light source are emitted in a randomfashion approximating a Poisson distribution. This is a fundamental source ofnoise in any measurement of light.
    光辐射出的最小能量包成为光子。虽然光子通常是一种可视化粒子,但它们也波的特性。从光源发出的光子以随机发散的形式接近泊松分布。这是任何光学观测当中的一个基本噪声源。


    Pixel Math 像素运算
    The Pixel Math command allows you to perform apixel-by-pixel mathematical operation on either a single image or two images.
    像素运算命令可以让你在一个单一的图像或两个图像之间执行像素级的数学运算。


    Pixel Scale 像素分辨率
    Pixel scale is a measurement of how much of the sky iscovered by one pixel. Image Scale (arc-seconds) = 206 * Pixel Size (microns) /Focal Length (mm).
    像素分辨率用来测量一个像素能覆盖多大的天区。图像分辨率(角秒)= 206*像素大小(微米)/焦距(毫米)。


    Point-Spread Function 点扩散函数
    Also known as PSF. A function which describes the blurthat was applied to an image. The ideal point-spread function for anastronomical image is a noise-free image of a star taken through the sameoptical system and observing conditions. Often this is modeled as an idealizedGaussian or Exponential distribution.
    也被称为PSF。一个描述图像模糊的函数。天文图像的理想点扩散函数是在一个光学系统和观测条件下拍摄的无噪声恒星图像。通常情况下建模成为一个理想化的高斯或指数分布。


    Poisson 泊松
    An important type of statistical distribution. APoisson distribution results when the arrival time of an infrequent event ismeasured. When measuring the amount of light from a source, a Poissondistribution describes the arrival rate of the photons of light.
    一种统计分布的重要类型。当偶发事件的到达时间被测量到时,会产生泊松分布的结果。当测量某个光源的光线时,泊松分布描述了光子的到达率。


    Quantization 量化
    Quantization occurs when an analog signal is convertedto a digital representation. Quantization adds a small amount of noise into ameasurement; for example, if a sample value is 7.3, it will be represented as7. That difference is referred to as quantization noise.
    当模拟信号转换为数字表示时,将发生量化。量化将会为测量增加少量噪音;例如,如果一个样本的值是7.3,量化后表示为7。这种差异被称为量化噪声。


    Typical Analog-to-Digital converters can producenumbers with 8 to 16 bits resolution. An 8-bit converter can represent valuesfrom 0 to 255, whereas a 16-bit converter can represent values between 0 and65,535.
    典型的模拟-数字转换器可以达到8至16位的分辨率。一个8位的A/D转换器可以表示0到255的值,而16位转换器可以表示0到65535的值。


    In order to adequately represent image sensor data,the converter must have enough bits of resolution to record both the brightestvalue possible (saturation) while simultaneously recording the noisebackground. In order to avoid adding extra quantization noise into the image,the noise should typically be sampled by 2-3 bits of resolution.
    为了充分反映图像传感器的原始数据,A/D转换器必须有足够的分辨率来记录最大范围的亮度值(饱和度),同时记录背景噪声。为了避免在图像中添加额外的量化噪声,噪声通常利用2-3位的分辨率进行采样。


    The number of bits required to adequately record asensor's image depends on the well depth, or maximum number of photoelectronsit can record. The well depth is larger for physically larger pixels, but alsodepends on the exact details of the particular sensor technology used.
    传感器能够充分记录图像所需的位数取决于阱深,或者说是能够记录的最大数量的光电子数。阱深越大就需要更大的像素尺寸,同样也取决于传感器的结构和制造技术。


    For typical sensors, 8-bit resolution is insufficientfor high quality images; whereas 16-bit converters are overkill in most cases,except for large pixel sensors.
    典型的传感器,8位的分辨率不能提供高质量的图像;而16位转换器,在大多数情况下矫枉过正,除了大像素尺寸的传感器。


    Reconstruction Filter 重构滤波器
    An image or signal that has been digitized accordingto the Nyquist Sampling Theorem contains all the information necessary toreconstruct the original. It is not however an exact duplicate of the originaluntil it has been passed through a reconstruction filter. In the case of astereo system, the reconstruction filter is a low-pass filter on the output ofthe digital-to-analog converter. In the case of a digital image, the onlypractical way to reconstruct the image is to upsample it to a higher samplingrate and then low-pass filter it digitally. This produces a very goodapproximation to the original image.
    一个图像或信号被数字化时,根据Nyquist采样定理,必须包含所有必要的信息以确保能重构还原。在运用重建滤波器之前,它并不能实现和原始信号一模一样。对于一个立体声系统,重建滤波器是一个A/D转换器输出端的低通滤波器。对于数字图像,重建图像的唯一可行方法是放大采样到一个更高的采样率,然后通过低通滤波器进行数字化。这将得到一个很好的近似原始图像的结果。


    Roundness/Rness星点圆度,判断照片质量的参数


    RMS Wander RMS徘徊
    The average deviation of the guide star from thecenter position, and is measured in pixels. 偏离导星中心位置的平均值,用像素表示。


    Root Mean Square 均方根
    RMS is calculated on a set of data points (pixels) bytaking the square of each pixel, calculating the sum of those squares, andtaking the square root. The result is scaled by the number of pixels. RMS givesan accurate measurement of the amount of noise present.
    均方根值是对一组数据点(像素)进行计算,取每个像素的平方,计算这些像素平方的和,然后取平方根。其结果随像素数量的多少而改变。RMS给出了一个噪声测量的精确描述。



    !reply!

    !thread_magic! !report!

    该用户从未签到

    10

    viewthread_userinfo_threads

    43

    viewthread_userinfo_posts

    446

    viewthread_userinfo_credits

    中级会员

    Rank: 3Rank: 3

    viewthread_userinfo_credits
    446
    !poston!: 2017-2-5 23:18:09
    | !thread_show_all!
    前排学习
    !reply!

    !thread_magic! !report!

    该用户从未签到

    0

    viewthread_userinfo_threads

    11

    viewthread_userinfo_posts

    8951

    viewthread_userinfo_credits

    论坛元老

    Rank: 8Rank: 8

    viewthread_userinfo_credits
    8951

    最佳新人论坛元老

    !poston!: 2017-2-10 11:20:52
    | !thread_show_all!
    后排入座
    !reply!

    !thread_magic! !report!

  • TA的每日心情
    难过
    2018-2-10 22:36
  • 签到天数: 1 天

    [LV.1]初来乍到

    36

    viewthread_userinfo_threads

    140

    viewthread_userinfo_posts

    610k

    viewthread_userinfo_credits

    论坛贵宾

    Rank: 6Rank: 6

    viewthread_userinfo_credits
    67974

    最佳新人活跃会员热心会员推广达人宣传达人灌水之王突出贡献优秀版主荣誉管理论坛元老

    !poston!: 2019-1-31 22:19:29
    | !thread_show_all!
    催更
    !reply!

    !thread_magic! !report!

    !post_credits_rule!

    QQ|手机版|小黑屋|开源天文论坛     开源天文论坛 TinyAstro 讨论组

    !time_now! , !processed_in! 0.092369 !seconds!, 37 !queries! .

    Powered by Discuz! X3.3

    Release 20170120, © 2001-2019 Comsenz Inc.

    MultiLingual version, Rev. 555, © 2009-2019 codersclub.org

    !fastreply! !scrolltop! !return_list!