Albert. ]]>

one remark about the described method to measure row noise.

imagine an image of n rows and m columns,

pixel values being white gaussian noise of an RMS amplitude of X, and the noise being uncorrelated among the pixels.

Applying the method: create a vector of line averages and calculate the std deviation of the vector elements, gives a value which is greater than zero.

However, one will not recognize row noise in this image.

Now, would it not be reasonable to quadratically subtract the X/sqrt(m) contribution upfront and only name the rest “row noise” ?

hs

]]>When trying to work it out analytically the effect of increasing angle on the MTF scales like sinc(0.5*tan(angle)) with sinc(x) as defined by Woodward.

Do you find this relationship back in the measurements?

Best regards,

Bert

]]>While I was a student in the 80’s the revolution occurred thanks to Don Knuth, Xerox Parc scientists, and many others, with Tex, LaTex, and WYSIWYG editing becoming mainstream, and so the author himself produced a copy-ready book with no laborious typesetting necessary.

This should have resulted in the most esoteric technical books being available for $10 or $20… but that did not occur and here we are three decades later, and the situation is WORSE. You could argue that the author now does all the work to provide “camera-ready” books, and reaps NONE of the benefits. Someone is making a KILLING on these things…

There should be a revolt against the greedy publishers…

]]>I do think that the effect you mention is more a measurement/sensor/camera inaccuracy than a blooming effect. Normally blooming occurs first towards direct neighbouring pixels. For figure 2 this would mean from red pixels into green pixels. And the green pixel is reacting pretty linear, even after the red pixel is saturated. On the other hand, the blue pixel is a diagonal neighbour of the red pixel, and blooming is much more unlikely. Even if your suggestion is applied to split the regression line into two parts, you will see that the second part of the regression line gets a smaller slope, but actually with blooming, you would expect a steeper slope.

I do hope that this explanation is clear to you. Otherwise let me know. Albert. ]]>

To me, in figure 2 there actually seems to be a change in sensitivity of blue and green, after red has saturated. The regression line of blue crosses the plot two times. I bet when you split the regression line and leave some space between both of them, there’ll be a difference in slope.

I know it does not matter. Reading on …

]]>