JPEG File Conversion Size

[bobsofpa]

I converted the same TIFF file to JPEG in PWP and DXO with both conversions set at 100% quality. The DXO JPEG file is 50% larger than the PWP file. Any comments. Will it make any real difference to save the files using DXO other than chewing up more hard drive space?

[jsachs]

The degree of compression for a given quality setting depends on the particular JPEG implementation so there is no consistency from one program to the next. The DxO version is probably a slightly more faithful rendition of the original image, but I doubt you would ever notice the difference. You can always save both ways and compare the two images.

[doug]

When testing whether the "lossy" nature of multiple saves of a jpeg (without any intervening transformation activity) actually had a detrimental effect, a technique I used was to use the "absolute difference" setting in PWP's Composite transformation to compare an original image with one resulting from multiple, iterative saves.

Any comments about the relevance of using that technique in assessing the question in this thread? Would also appreciate any comment about things to keep in mind when using the technique to measure the degradation from multiple saves of a jpeg as I did?

[jsachs]

The absolute difference technique is a good way to see where the differences are, but ultimately just comparing the images side by side with the original is the only test that matters since JPEG is based on compressing images so as to minimize perceptual differences rather than absolute differences. For example, the human eye can resolve much more detail in luminance than in color (chrominance). This means that chrominance information can be compressed more than luminance.

[Dieter Mayr]

When using the absolute difference method I like to use Window - Count Colors on the result image as kind of a quality indicator.
The less colors the resulted difference image has, the less the difference.
A Levels and Color - Full Range Transformation on the difference image can also help to indicate where the areas with the most differences are.
Then one can focus on this area in a lets say 1:1 view on the original images and check if the difference is siginificant or not.