This is only a test (but a fun one)…

As many of you know, the forums for the National Association of Photoshop Professionals (NAPP), is a place that I really enjoy spending my spare time in.  It’s an excellent resource for a number of reasons:  first of all, I learn a lot!  Second, some great questions come through there that get my brain going.  Third, and most importantly, when I contribute to discussions in there, it often turns into some meaningful content I’ve created to help explain things.  That content then becomes some very useful tools in creating – yup, blog posts for the readers out here in the non-NAPP world!

For instance, someone was asking in the NAPP forums about methods to ensure all their files are output to a minimum of 50 MB for delivery to a client.  It was a great question with lots of useful contributions.  I decided to throw my two cents in, with a few suggestions that not only supported those made by others, but also include some tangential information.  My thoughts were that a client asking all images to be at least 50 MB suggested that the client doesn’t understand color very well…because different colors have different degrees of data to them.  Translation?  Some colors will produce larger inherent file sizes .

I ran a little test to help demonstrate this by taking some pictures.  Since I really wanted to get a complete illustration I sought out to find scenes that were 100% red, 100% green and 100% blue.  Naturally, I didn’t have a lot of luck in the real world, but I could produce them easily enough in Photoshop.  So, here’s what I did:

1.  Created a new document, 800px square, and filled the background with a pure red, green and blue:



2.  On each new fill, I pointed my camera at the monitor (which is calibrated every two weeks*), and took a picture.  Here’s some screen shots of what the histograms looked like on the back of my camera LCD:




As you can tell, the histograms show that the colors are pretty spot on to where their anticipated locations should be…if you read the details below the histograms, you’ll also notice the amount of data that was in each image, but to make sure I was reading the data correctly…

3.  I then copied the files to the computer, and without any editing on white balance or anything, looked at the file sizes.  The results were pretty interesting:

Color Data

So, the color blue will result in a larger file size than red and green carries more than blue.  I also noticed that each color was successively brighter, which supports my understanding that more light in a scene also produces more data.  Since some colors are inherently “brighter” than others on the color wheel, they will also naturally have more light.

It was an interesting test/exercise to illustrate that different colors carry different amounts of data.  This also ties into the theory of “exposing for highlights and developing for shadows”.  It makes sense…because more data is available from brighter colors and less data is available in darkness.

*Finally, it bears mentioning that I do have a rather envious setup – you see I was given permission to paint the office any color I wanted recently.  I chose an 18% gray.  The office was then lit with a daylight balanced bulb from one desk lamp, and the window is normally covered with a black felt.  This, when combined with a calibrated monitor, profiled paper makes for a pretty good environment in ensuring color accuracy.  This little “color test” for someone else also was a good test for the office environment, and the results there were pleasing indeed.

This post has gone on much longer than I had anticipated, but it was a great chance to expand a little bit on a couple of topics:

1 – What happens with color in your pictures.

2 – How that impacts your file storage needs and requirements for clients

3 – The importance of working in a good color managed environment

If anyone has any questions, comments, or input of their own on the subject, please feel free to share your thoughts with me via the comments section or via email.  Happy shooting, and we’ll see you back here tomorrow!

Please follow and like:

7 comments for “This is only a test (but a fun one)…

  1. May 24, 2010 at 5:28 am

    I had always suspected that might be the case, great report. I know the more detailed the image, the higher the file size will be, but this is interesting info.
    .-= Blake Discher´s last blog ..Flickr Image Stolen in Less Than 24 Hours =-.

  2. May 24, 2010 at 10:30 am

    Hi. I really appreciate this experiment; at some point in the near future I will be pursuing this whole 50MB submission requirement with my image bank in earnest…It just makes NO sense. In the meantime I will do what they want and get my images sent off. When I do question my image bankI will certainly refer them to this blog and to your comments and feedback at NAPP. In case you are interested, during my travels down the rabbit hole on this subject, I found Genuine Fractals to be the best product for interpolation, good website at, in case you are interested.

    • May 26, 2010 at 10:05 pm

      Thanks Matthew for taking the time to stop in and comment here – glad you found the material useful both in NAPP and out here…(you can tell I get a little more wordy out here! 🙂 )

  3. Clay Teague
    May 24, 2010 at 6:39 pm

    I see where you conducted your experiment using Adobe RGB and with the smallest JPG file size.

    I wonder if these results might be different with different compression settings?

    • May 26, 2010 at 1:05 pm

      The file sizes would likely be larger across the board, but I think the relative sizes would be on par to one another…

  4. May 26, 2010 at 12:09 am

    Wow, this terst is really interesting. I’m quite surprised by the result!

    Would be interesting to see that test conducted with images of different lvl of detail. If you have a person standing in front of a plain wall or in front of a fine pattern can make a huge difference in file size while the quality of the subject stays the same. So suggesting file sizes as a means of quality is just plain ignorant to the facts as you already stated above.

    What would you suggest to take as a measure for quality?
    I’d suggest image-with and hight (Megapixels) and compression ratio (“quality” in jpeg-terms).
    .-= kruemi´s last blog ..IMG_2915.JPG =-.

Leave a Reply

Your email address will not be published. Required fields are marked *