On Mon, 15 Dec 2003 09:48:09 GMT, "Richard Peterson" reverently intoned up the aether:
Although the actual diagonal pattern is not a function of the CCD noise, I don’t see what you mean here. BTW, Sean is right in stating that it is not CCD noise from low light. Low light does lead to CCD noise – especially in the preamplifier.
Everything I have read references dark current noise as a function of temperature and not as a function of incident light. Hence CCD noise will be more prevalent at a gig due to the high ambient temperature created by all those closely packed bodies, but the low light will not cause.
I do agree that dark current noise (CCD noise) will be more notable in low light as it is relatively larger in magnitude in comparison to the desired signal.
The thing is, Whatever the noise source in a CCD image,
it is still noise. Noise, by definition, does not have a pattern to it.
But Sean, I don’t get how dragging the shutter has anything to do with that pattern in the image.
The only "diagonal lines" I can see in the posted example are causes by musical equipment and by the ghosting effect caused by dragging the shutter (sharp image plus a ghostlike image added to it from the longer exposure). There is a highly visible noise pattern in the distant out of focus (OoF) areas but
It seems to me to be some non-noise function of the
electronics, not something optical. I mean, I have never seen something like that on film.
Could you explain?
I will try. The starting point for this is to begin with an understanding of image formation on the film plane (CCD) and how focus and optical defects affect the blur circles of points of light. These ideas then need to be extended to more complex images.
When we focus upon a point of light we expect the camera to record this as a point of light on the film plane. Pretend you are 100 miles from the nearest electric light on a moonless night with no wind and you are using a 50 mm lens to focus on a small candle flame 20 meters away to get a point of light. When this point is no longer on the plane of critical focus we expect the point to be rendered as a blur circle. The size of this blur circle corresponds to how out of focus the point of light is. In an idealized optical system the shape of a blur circle should be a uniformly distributed disk of light.
Because of the physical limitations that must be overcome in engineering lenses this idealized blur circle is never achieved. Take a look at figure 6 in the following document to see some reasonable examples of what occurs with most real world lenses.
http://luminous-landscape.com/essays/bokeh.shtml The term bokeh (boke with a sharp e in Roman alphabet Japanese spelling) is used to describe the character of the blur circles in an image. If you wish to read more on this subject then I would recommend taking at look at the following link.
http://www.trenholm.org/hmmerk/ATVB.pdf The section where he spouts idiocy and tries to explain convolution should be skipped over as it makes no sense and shows that the author may not completely understand it. Sadly, the only clear explanations of what convolution is are in introductory mathematical (not engineering) texts on Fourier Analysis and that is not the most accessible material.
Taking a look at the specs for the Toshiba PDR-M25 I see that it is a point and shoot consumer digicam so I would expect to have the most common optical characteristic of having good bokeh for objects nearer than the plane of focus an bad bokeh for object beyond the plane of focus. This indicates that the blur circles for out of focus points of light beyond the plane of focus are highly likely to be bright circles with diffusely lit center like the two leftmost blur circle in figure 6 at
http://luminous-landscape.com/essays/bokeh.shtml. It turns out that in this case the blur circles actually appear to be a bright ring with a diffuse fill and a bright center. This type of optical distortion of the blur circle will render fairly similarly whether the OoF object is closer or further from the lens than the plane of critical focus. The determination of this is based upon the shape of out of focus specs of airborne dust near the lens that were struck by the on camera flash. Take a look at
http://www.tearnet.com/sean/tmp/BlurCircle.jpg and you can see how there are actually several of these dust induced lens flares that can be seen from if the contrast of shadow detail is strongly enhanced.
Taking these ideas and expanding them to cover image formation we get the idea that in focus points of light in the image should be rendered as points of light. This occurs on the two heads of hair near the lens and on all objects as close as or closer than the bassist is to the lens. In complete honestly the bassist is only acceptable and not critical focus.
Next we begin looking at the portions of the image which are further away from the lens than the bassist. These areas are out of focus. So each point of light becomes rendered as bright circle of light and bright centered point of light connected by a diffuse and dimly lit circle of light. Now this can create interference patterns. A nice extreme example of this can be seen by looking at the background of the image on the first page of
http://www.trenholm.org/hmmerk/ATVB.pdf where a single spire of temple is rendered as two overlapping diffuse spires.
Now if we begin considering how each point in the background there is rendered this way we get a large collection of overlapping circles and points. These interference patterns are often moderately noisy and somewhat aesthetically lacking. But being out of focus such noisy blurs are usually moderately soft at the same time.
This leaves us with the OoF areas being a bumpy but smooth blur (not terribly aesthetically pleasing, but blurred nonetheless). This type of noise can also be found in film shots. If you would like I can hunt some down and scan a few to explore this topic further if you would so like.
Having taken a look at the specs of the Toshiba PDR-M25 I have noted that this camera also only records jpegs. Now jpegs tend to do a very poor job of rendering smooth textures (i.e. blurs) as it is normally high frequency detail lost in the quantizer that makes it smooth. This tends to add a significant amount of noise to image regions where there are smooth textures unless very little compression is used. This specific image was compressed by a factor of 6 so the introduction of such noise is highly likely. Also, by looking at the EXIF data I can see that this was a 1/4 second exposure at f/3. A 1/4 second exposure is not long enough for dark current noise to become significant. With my digicam dark current noise is not visible to the naked eye in an 8 second exposure at 68 degrees fahrenheit. I have found dark current noise to become significant in 30 second and longer exposures at lower temperatures than that.
Hence in a final analysis of this image I would conclude that the noise seen is mostly compression induced noise on noisy blurs. By overlaying this compression induced noise over dark regions it becomes relatively stronger. Zooming in to view the OoF areas at 400% plus zoom levels you can see jpeg compression artifacts introducing small amounts of color noise on top of the noisy induced by defocus.
Whereas, if this were CCD induce noise, then I would expect the amount of noise to be roughly uniform across in focus and out of focus elements in this image. But this is not so.
It is also very common for people to malign digital capture methods while claiming that phenomena they are seeing never happened on film when it was actually the wider exposure latitude of film compressing artifacts into a smaller percentage of the total visualization space that prevented such issues from becoming quite as prominently.
If any of this is unclear, please feel free to ask for clarification as this is Usenet and you are only getting rough draft quality writing from me here.
enjoy your day,
Sean
"In the End, we will remember not the words of our enemies, but the silence of our friends."
– Martin Luther King Jr. (1929-1968)
Redesigned – Last Updated 15 October 2003
Photo Archive @
http://www.tearnet.com/Sean