3
\$\begingroup\$

I just got this Tamron AF 70-300mm 4-5,6 Di LD Macro 1:2 for my Canon 500D.

I tested it yesterday and it works fine. Only one problem: I took pictures of a flower which is dark violet. But when I look at the pictures on screen, the color is kind of light blue, which is a big difference. I use a UV filter and the sun block hood, and there was no sun light at that moment. I changed to my Canon 55-250mm, and it didn't have the same problem.

I don't know if this is a technique problem, or an issue with this lens. Does anyone have the same experience?

This is the one I took with Tamron, I resized it: enter image description here

And this one I took with my Galaxy S7: enter image description here

\$\endgroup\$
8
  • 2
    \$\begingroup\$ Does this persist when the UV filter is removed? Some UV filter transmission curves start dropping rather sharply a bit before 400 nm and that could make a difference if the violet color of this flower is in the 380-410 nm range. \$\endgroup\$
    – 0x5f3759df
    Commented Jul 18, 2016 at 16:20
  • 4
    \$\begingroup\$ It would be very nice to see the sample, small crop is fine. \$\endgroup\$ Commented Jul 18, 2016 at 16:26
  • 3
    \$\begingroup\$ I'd particularly like to see both samples side by side. \$\endgroup\$
    – mattdm
    Commented Jul 18, 2016 at 16:38
  • 3
    \$\begingroup\$ What was the white balance setting on your 500D at the time you took the photo? \$\endgroup\$
    – Michael C
    Commented Jul 19, 2016 at 5:29
  • 2
    \$\begingroup\$ Looks like you have a very different white balance between the two photos you gave as example. You should put a ColorChecker for accurate comparison. \$\endgroup\$
    – Kel Solaar
    Commented Jul 19, 2016 at 8:35

5 Answers 5

7
\$\begingroup\$

The flower is supposed to be dark violet, but it's come out light blue.

That indicates the image is overexposed. If the flower is dark, the image of the flower should be a similar tone. So for a start you need to lower the exposure and darken the flower.

The reason for the color change from violet to blue is that the reds in the image are blown (or clipped) (see also Why are red objects coming out unnaturally in my photos?) It is common in photographing red objects that they come out pink or even orange. When you have a purple object, overexpose somewhat and blow out the red channel, but not the blue channel, then you may end up with a light violet or blue.

Look for the reds being at the extreme right of the histogram like this:

enter image description here

and lower the exposure, or use exposure bracketing, until they are not blown.

Edit: now that you've posted examples, it looks like it may simply be the BLUE channel blown a bit.

enter image description here

Again, just try lowering the exposure a bit.

\$\endgroup\$
5
  • \$\begingroup\$ thank you for the explanation. there was no sunlight at that moment, but maybe like you said, I had a high exposure, so the picture was overexposed. I just edited my question \$\endgroup\$
    – Ragnarsson
    Commented Jul 18, 2016 at 21:25
  • 1
    \$\begingroup\$ Is that a histogram of the picture in question? If I open the posted image in Photoshop, the histogram looks totally different from the one above and reds are not blown at all. Also, in the picture the flowers are blue even in the shadows (actually more in the shadows). \$\endgroup\$
    – MirekE
    Commented Jul 19, 2016 at 15:26
  • \$\begingroup\$ No, that is a histogram from one of the links in my answer. No images had been posted at that point, hence my edit. \$\endgroup\$
    – MikeW
    Commented Jul 19, 2016 at 18:57
  • \$\begingroup\$ Downvote because reds are not clipped. \$\endgroup\$ Commented Jul 19, 2016 at 20:05
  • \$\begingroup\$ Also, exposure is not a problem in this case. \$\endgroup\$ Commented Jul 20, 2016 at 16:12
7
\$\begingroup\$

Despite you saying that this does not happen with 55-250mm objective I claim that this is the problem with camera settings. If you finally upload the photograph made with 500D and 55-250mm I may update the answer.

Cameras do not generally make neutral images because it is impossible to clip while selling cameras which do not produce the image which make your friends astonished. Cameras do bump chromatic saturation and they do bump contrast. It seems that you have set the contrast and saturation too high and this is obvious because of harsh transitions between different colours: enter image description here

I have taken your sample image and made it even worse using the algorythms which cameras are broadly using.

Here is a saturation bump: enter image description here

Here is a contrast increase: enter image description here

You may see that flowers became even more blue than they already were.

Also, here is an illustration of exposure not being the cause of bluish tint: enter image description here

I have further increased exposure and it only made flower look more violet.

Your solution is to:

  1. drive contrast and saturation back
  2. decrease the exposure because the blue channel seems to be clipped
  3. if that is not enough try altering the white balance because it seems to be off compared to photo from smartphone (too "cold" what worsens the bluish tint cause by saturation and contrast bump).

Here is the blue (not red) channel of your image which does not contain details of flowers: enter image description here

Interesting thing is that this clipping well may be caused by the bump of contrast and saturation because of how limited sRGB is: a ProPhoto triplet (50,10,200) does not have an sRGB equivalent and will be something like (20,0,255) after absolute colorimetric conversion, and increasing saturation and contrast makes it only worse.

\$\endgroup\$
8
  • \$\begingroup\$ Good analysis of the image. +1 \$\endgroup\$
    – Michael C
    Commented Jul 20, 2016 at 6:20
  • \$\begingroup\$ What does, "Cameras do not generally make neutral images because it is impossible to chop bucks selling cameras which do not produce the image which make your peeps astonished." mean? \$\endgroup\$
    – Stan
    Commented Jul 20, 2016 at 16:22
  • \$\begingroup\$ @stan: it means that OEMs should follow suspicious trends in image processing to earn money. I am not a native speaker, feel free to suggest an edit. Is it better now? \$\endgroup\$ Commented Jul 20, 2016 at 17:02
  • \$\begingroup\$ @EuriPinhollow Do you mean that realistic images are not as attractive to us as fully saturated ones are? Are you saying that "realistic" and "average" isn't good enough for us now? \$\endgroup\$
    – Stan
    Commented Jul 20, 2016 at 17:05
  • \$\begingroup\$ @stan: roughly speaking: yes. I am not pointing at users but pointing out that cropping money is much more difficult without that processing. It is better to continue talking in chat. \$\endgroup\$ Commented Jul 20, 2016 at 17:18
1
\$\begingroup\$
  1. This is not a problem with the lens
  2. There are several possible reasons why this can happen, some trickier than others. But most likely problem is the color balance. You probably used automatic color balance settings on your camera and since there are only limited colors in the picture, the scene was not evaluated correctly. You should be able to fix that in a photo editor.
\$\endgroup\$
1
  • \$\begingroup\$ Color balance certainly affects the way the final image looks, but it doesn't (usually) affect the exposure unless the camera in question is one of the newer types that has an RGB+ir meter or is used with Live View. The 500D has a standard monochrome light meter. Even if the white balance had been set properly the blue channel would have likely been just as blown out in any auto/semi-auto mode unless the photographer used an RGB histogram, noticed the blue channel was blown and then took the shot again with enough negative EC applied. \$\endgroup\$
    – Michael C
    Commented Jul 20, 2016 at 6:24
1
\$\begingroup\$

It is neither a lens nor a technique problem.

There are instances where trying to capture true renditions of naturally occurring hues is not possible. The morning glory bloom is one specific example.

Violet is one hue with wavelengths that cannot be artificially reproduced at the present. Your subject pictured above appears to be in that category. The closest we can come with pigments and/or other colourants is purple (a mix of red and blue). As soon as you compare a reproduction with the original, the difference is evident.

The difficulty is the hue is not within the spectral sensitivity of the sensor. It out of the sensor's "gamut," if you will. This also happens with light-sensitive film emulsions or any system which only approximate human vision.

\$\endgroup\$
1
  • \$\begingroup\$ This is the only correct answer. I've run into the same problem multiple times myself. This being the reason becomes much more obvious if you try to take a picture of a UV led light. Your eyes will see violet, but the camera will only see blue. \$\endgroup\$ Commented Nov 16, 2020 at 20:23
1
\$\begingroup\$

This happens with film also. Violet can be approximated with blue and red mixed up, but in reality, violet can be a single wavelength. The camera's sensor (or three colour film emulsion) measures the quantity of light energy in three bins, the blue 'bin' (475nm) being the one for the shortest wavelengths, and so violet (400 mm - a shorter wavelength than blue which is at 475 mm) is in that same blue bin. The other 'bins' are green (500 nm) and red (650 nm). The red sensor (a longer wavelength at about 650 mm) does not pick up any of the violet, so you get blue (475) where violet (400) used to be. To accommodate this, a fourth sensor could be used to grab the violet at the expense of resolution, but I think in film the extra layer would be an issue in terms of a complementary dye coupler. Adding an additional sensitizing dye to the red layer on film is probably not possible because the red layer is underneath the yellow filter layer. If colour reproduction of violet was important, one green cell in the mask pattern of a digital sensor could possibly be swapped out for violet, in theory.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.