brings us to introduce one of the most controversial topics for the young amateur digital noise.
One might be inclined to think that the dark frame is a form of "noise" to be removed from our raw images, not exactly. The dark frame is a signal: a sign of trouble but still a signal: the signal generated by dark current emitted by the CCD camera (or DSLR) at a given temperature and for a given exposure time. And luckily it's a sign! Luckily because, if we make the sensor work in the same conditions of exposure and temperature that we used to take the ' RAW image, this signal is "almost" perfectly reproducible.
But it is in that "almost" hiding, sneaky, the concept of noise .
provided evidence to back two dark frames in exactly the same conditions (same exposure time and same temperature). Now carefully analyze them pixel by pixel: are identical? Each pixel of coordinates (x, y) in the first dark frame has the same value in the corresponding pixel of the second dark frames? Certainly not. Of this we know it better by doing a simple subtraction: if we subtract the dark frame from the first second (or vice versa) will result in a new image with the pixel value around zero but not exactly all zero as you might expect.
Astroart has the essential tools for numerical analysis of the images as the histogram and the statistics window. To better analyze this phenomenon we can generate a histogram: a histogram, the better we will see later, is a chart with the x-axis values \u200b\u200bof the pixels in the image in ascending order, and ordered the number pixel with that certain value. Well, the histogram of our image in the shape of a bell, with the peak at the value zero. This means that most of the pixels in the image have zero value, but many others have non-zero values, in the slightly greater than zero and some slightly less than zero. The bell shape is due to the fact that the more one deviates from the average pixel value (zero in this case), the lower the number of pixels in the image with that particular value. Ideally, if we had a perfect CCD camera, capable of playing two dark frames exactly alike, stealing from each other, we would get an image composed of all the pixels identically zero. It 's easy to see that the histogram of this image would be composed of a single column at the zero height equal to the number of pixels in the image. It is even more easy to deduce that the amplitude of the bell has a certain relation to the distribution of nonzero values \u200b\u200band thus is directly related to noise present in images taken with our CCD camera. The standard deviation , the statistics of the image shown in the box, it gives a direct measure of the dispersion of pixel values \u200b\u200bin the neighborhood of the average value (zero) and is therefore a quantity related to the "noise" image: the higher the standard deviation, the greater is the dispersion, the greater the "noise" that accompanies these images. Hence the concept of "noise" must be understood as a series of events that never allow the CCD to reproduce an image exactly equal to each other while operating in exactly the same conditions. In this case we used two identical dark frames but the same reasoning applies to any pair of FRAME taken using a CCD camera, two RAW images of any object in the sky, two flat field, two bias frame and so on.: We will never get two FRAME exactly equal to each other.
But what is worse is that we're going to do any math operation with the frames, such as the removal of DARK FRAME RAW image, so we're going to steal the map of thermal noise and BIAS RAW image, but also we're going to aggregate the noise in the two frames. What are the sources of noise and how can we do to minimize (note: we can only minimize and do not remove it entirely) the contribution of noise in the final image we will see in future posts.
0 comments:
Post a Comment