Saturday, April 28, 2012

#1940Census Who Has the Best Images?

In my “1940 Census Image Viewer Comparison” article I noted how different websites took more or less time to display images. Ancestry.com took more than 3 seconds, Archives.gov took about 4, MyHeritage.com took about 17 seconds, and FamilySearch about 34.

The most significant factor affecting download time is the size of the image file. The most significant factor affecting file size is image quality. Thus, there is a tradeoff between download speed and image quality. The faster the download, the worse the quality. The better the quality, the slower the download.

Back on 9 April 2012 in the Monday Mailbox I made a stupid statement. “The Rowdy” asserted that Ancestry.com had the highest quality images. I replied that “NARA did the image scanning so Ancestry.com’s images can’t be better than everybody else’s.” I knew at the time that websites might modify the images prior to publication. But it seemed silly to say something like “Ancestry.com’s images can’t be better than everybody else’s unless everyone else messes up their images worse than Ancestry.” (Thank you to the several of you who kindly wrote pointing out different image qualities of different websites.)

Last time I talked about the quality of the images provided by the National Archives and Record Administration (NARA). As you look at the images provided on the different websites, keep in mind that the focus problems are largely NARA’s fault.

Ancestry.com

Ancestry.com applies an algorithm to its images to increase contrast. Whites become whiter and blacks become blacker. Most people like the resulting effect, as it matches our expectation as to what a black and white record should look like. On the plus side, it makes legible text more legible. On the minus side, it makes illegible text more illegible. The increased contrast also makes it easy to compress the images. Ancestry’s images are half the size of the NARA originals. That in turn allows Ancestry to display images twice as quickly.

image

While not as noticeable, Ancestry also straightened the images; the originals seeming to slope a little down to the right.

Archives.gov

Archives.gov used more compression to decrease the file size by three. You can see the effect if you zoom in close to the image. As shown below, compression causes squares to form in the background and fuzz to grow on the writing. The effect may not be noticeable at normal magnification, so long as the compression isn’t too aggressive.

image

FamilySearch.org

FamilySearch.org did nothing to compress its images. Consequently, FamilySearch has the slowest display time. As Ancestry, they rotated the images slightly to straighten them. FamilySearch also sharpened the images. To some degree, sharpening repairs some of the focus problems. However, sharpening exaggerates errors as much as the real stuff in the image. The original NARA images have weird vertical lines covering the entire image. Sharpening makes these easier to see in the FamilySearch images, even at normal magnification.

image

MyHeritage.com

MyHeritage reduced the size of the images, decreasing the number of pixels by four and increasing the fuzzy appearance of the images.

image

Conclusion

In a side by side comparison, below, it is clear that FamilySearch.org has the sharpest images. As one might expect, the website with the slowest image display has the crispest images.

image

Comparison Table

  Straightened Contrast Resized Compression File Size (MB) Display Speed
Original       1.0 4.712  
Ancestry.com Yes Increased   2.19 2.151 >3
Archives.gov       3.05 (largest) 1.545 4
FamilySearch.org Yes Sharpened   1.07 4.414 34
MyHeritage.com     Smaller 2.60 1.814 17

5 comments:

  1. Thank you. This is very interesting and explains a lot.

    ReplyDelete
  2. Don't you have the display speeds for FS and MH reversed?

    ReplyDelete
    Replies
    1. Dear readers,

      Randy is correct. I accidentally swapped the display speeds for FamilySearch and MyHeritage. I have corrected them. FamilySearch takes 34 seconds on my home Internet connection, while MyHeritage, at 17 seconds, is twice as fast.

      --The Insider

      Delete
  3. Thanks for the great information. I will stick with Ancestry since I am a member but I like the family search images better.

    ReplyDelete
  4. On Ancestry.com you can choose to see the non-enhanced images. On the newer viewer go to the options and uncheck the box that says "Use Enhanced Images." I actually prefer the higher contrast of the enhanced images, but you can switch to the other if that's your preference.

    ReplyDelete