|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
![]() |
![]() |
| Nikon Coolpix 990 | Nikon Coolpix 950 |
![]() |
![]() |
| Canon PowerShot S20 | Casio QV-3000EX |
First thing to note is that three of the cameras managed the white balance almost perfectly: Coolpix 990, Coolpix 950 and PowerShot S20. The Casio QV-3000EX has a slight yellow cast.
We're only measuring colour here. RGB values below were taken from a VGA reduced image (to average colours and eliminate JPEG artifacts) using the Eyedropper tool in Photoshop with a 5 by 5 Average Sample Size.
| Coolpix 950 | Coolpix 990 | PowerShot S20 | QV-3000EX | |
| Patch White | 237,240,237 | 231,231,231 | 232,226,230 | 224,224,216 |
| Middle Gray | 118,118,118 | 108,108,108 | 112,110,112 | 104,104,95 |
| Patch Red | 250,50,68 | 223,52,58 | 200,33,50 | 216,50,55 |
| Patch Green | 0,169,80 | 3,158,70 | 59,144,82 | 32,131,60 |
| Patch Blue | 52,48,139 | 49,39,102 | 51,51,98 | 40,37,103 |
Overall I'd say all cameras did well, the 950 and 990 produced more neutral colours the QV-3000EX suffered from a yellow cast and the S20 suffered from a very slight magenta cast and lower saturation. These samples shouldn't be seen as definitive as they're just one set, taking these samples again in different circumstances may produce slightly different results.
Red Channel Noise

After my 950 review it became apparent that some people had noticed noise in blue skies, upon closer examination I saw this as red channel noise which could be corrected using third party tools. Closer inspection of some of the 990 shots with blue skies also indicated some noise, the samples above are of 50 x 50 blocks cropped from the colour comparison charts above of the Blue and Cyan patches, each block is then broken down into it's red green and blue channels and reflected as such directly below.
As you can see both the 950 and 990 suffer from more noise (speckling) on the Cyan sample (right hand sample) than the S20, the 990 certainly doesn't seem to be any worse than the 950 but then there's no improvement either. Whether this is down to the CCD, electronics systems, algorithms or JPEG compression it's there in the final image and therefore becomes visible in certain circumstances.
Resolution Comparison
Unfortunately the Casio QV-3000EX had to go back before I had the resolution chart setup so no QV-3000EX samples from now onward, though you can find comparisons between the QV-3000EX and Coolpix 990 in the QV-3000EX review.
Shots here are of the PIMA/ISO 12233 standard resolution test chart. Daylight, cameras set to auto, Exposure compensation +0.7 EV for all cameras. How to read the charts: All values are 1/100 th lines/picture height/width. So the "10" value equates to 1000.
![]() |
![]() |
| Nikon Coolpix 950 | |
![]() |
![]() |
| Nikon Coolpix 990 | |
![]() |
![]() |
| Canon PowerShot S20 | |
Measurable findings (three measurements taken for each camera):
| Camera | Measurement | Absolute Res. | Extinction Res. | Moire Res. |
| Nikon Coolpix 950 |
Horiz LPH | 750 | 1100 | 800 (mild) |
| Vert LPH | 800 | 1150 | 850 (weak) | |
| 5o Diagonal | 700 | 1000 | 900 (mild) | |
| Nikon Coolpix 990 |
Horiz LPH | 900 | 1300 | 1000 (mild) |
| Vert LPH | 900 | 1400 | n/v | |
| 5o Diagonal LPH | 900 | n/a | n/v | |
| Canon PowerShot S20 |
Horiz LPH | 850 | 1400 | 1000 (strong) |
| Vert LPH | 900 | 1400 | 1000 (mild) | |
| 5o Diagonal LPH | 900 | n/a | 1000 (weak) |
Definition of terms:
| LPH | Lines per Picture Height (to allow for different aspect ratios the measurement is the same for horizontal and vertical) |
| 5o Diagonal | Lines set at 5o diagonal |
| Absolute Resolution | Still defined detail (below Nyquist frequency*) |
| Extinction Resolution | Detail beyond camera's definition (becomes a solid gray alias) |
| Moiré Resolution | Resolution at which a Moiré (coloured bands) pattern emerges |
| n/a | Not Available (above the capability of the test chart) |
| n/v | Not Visible (not visible on test results) |
image detail. Beyond the Nyquist frequency aliasing occurs.
First off, yes guys (Canon Talk) I can see the magenta cast on the S20 shot, can we have a firmware fix please Canon? For absolute resolution there's little difference between the 990 and S20 (we'll ignore the 2.1 megapixel 950 in these comments). The S20 does however have much more noticeable Moiré (the striped yellow / blue patterns). Alias is a phenomenon which exists for all digital cameras, it's an effect of the definition of detail and the Bayer pattern on the CCD (detail beyond the Nyquist frequency). It's down to the internal algorithms in the camera to reduce their effect on the image, on the 990 it looks like the engineers have done a good job striking the balance between resolution and alias moiré, on the S20 Canon appear to have chosen resolution.
Thanks to Dave Etchells of Imaging-Resource for helping to correct a couple of terminology mistakes on this page.
| Bookmark: | ||||
| Actions: | < Previous | Next > | Print page |
Please wait, fetching latest prices.









