Return to Digital Photography Articles

Comparing JPEG Quality

This page shows a comparison of the JPEG compression quality used in the [Photoshop] (Save As 06) vs the [Photoshop] (Save As 10).


For more information, see Photoshop Save As vs Save For Web

Source #1: Photoshop (Quality: Save As 06)
Source #2: Photoshop (Quality: Save As 10)
Differences
Positive numbers indicate that Source #1 is better quality than Source #2, while Negative numbers indicate that Source #2 is better quality than Source #1.
Quantization Table: Luminance
-6-4-6-10-12-15-20-6
-4-4-6-9-13-16-3-1
-6-6-8-12-16-3-10
-10-9-12-16-3-100
-12-13-16-3-1000
-15-16-3-10000
-20-3-100000
-6-1000000
Quantization Table: Chrominance
-6-6-4-5-5-5-2-2
-6-6-4-10000
-4-4-100000
-5-1000000
-50000000
-50000000
-20000000
-20000000
Result: Different, Range:[-20...0]

Across all frequency components, Photoshop - (Save As 10) has far better compressed image quality in the luminance channel than Photoshop - (Save As 06)
Result: Different, Range: [-6...0]

Across all frequency components, Photoshop - (Save As 10) has far better compressed image quality in the chrominance channels than Photoshop - (Save As 06)

Chroma Sampling:
Photoshop - (Save As 06): 2x2 (Poor)
Photoshop - (Save As 10): 1x1 (Excellent)

Choose another comparison of JPEG Quality:

Select Source #1Select Source #2
Compare the selected sources:

Related background material

 


Reader's Comments:

Please leave your comments or suggestions below!
2010-04-17Joen
 IrFanView has chroma subsampling enabled as the default, which the program defines as 1x1 sampling. Checking the "Disable Subsampling" box causes IrfanView to use 2x2 subsampling, resulting in a much bigger file size and an image "poorer" in quality.

Which brings me to my question. There is an editor called, "Photo-Brush," that gives the user a choice between Standard sampling and "Enhanced Sampling." Their definition seems to be the reverse of IrfanView. "Enhanced Sampling" uses 2x2 chroma subsampling, and "Standard sampling is 1x1.

The company, Media Chance, has a program called, Jpeg Quality estimator, or JpgQ, that measures average quality and recompression error. They provide two values: one for the "Enhanced sampling" as used by Photo-Brush, and "Standard sampling," which they have as 2-1-1, the typical digital camera setting.

What are your thoughts on their use of the term, "Enhanced sampling," and the worth of the quality estimator program, "JpgQ?"

Thanks, Cal!
 It certainly looks like quite a bit of variability in how people define the chroma subsampling. Probably the most commonly-accepted terminology in technical circles is based on the 3-part ratio (eg. 4:4:4, 4:2:2, etc.), but this is quite confusing for the average consumer. For the purposes of my discussions, I elected to standardize on a "HxV" nomenclature that defines the number of 8x8 blocks in each direction that are covered by a single MCU.

With no subsampling (what I call 1x1), a single MCU is 8x8 pixels in size and encodes Y,Cb,Cr. With the common 2x1 subsampling (often seen in digicams, also known as 4:2:2), the MCU size is 2 blocks horizontally by one block vertically (16x8 pixels). In this arrangement, a single chrominance block spans 2 luminance blocks in the horizontal direction (effectively reducing the sampling rate in the horizontal direction by two).

Now, addressing your observations: If Irfanview describes a mode where it can "disable subsampling", then this should theoretically be the highest quality mode (greatest file size), where full chroma resolution is encoded (ie. 1x1 in my terminology). This seems to be opposite from what you have reported.

You point out that "Photo-Brush" has "standard" and "enhanced" sampling. Again, I would presume that "enhanced" implies greater quality (ie. no chroma subsampling, 1x1). Running JPEGsnoop on both of these tools should allow you to clearly compare the terminology in use versus the mode definitions.

The Jpeg Quality estimator (JpqQ) probably attempts to perform a very similar calculation as that displayed in JPEGsnoop. The value is only really useful if the program used to generate the JPEG file encoded the image using a quantization table based upon the table provided in the JPEG (ITU-T) standard annex. If the JPEG encoder used this common table and then scaled by the "standard" quality factor formula, then the quality estimator can attempt to work backwards to derive the quality factor. Unfortunately, most digicams do not base their tables on scaling this "standard" table, and therefore the approximate quality value can be somewhat meaningless.
2008-04-21T10
 Perhaps you should retest this. Save For Web has ALWAYS produced better results at the same file size. If by "060" you mean the "60" setting on the quality slider, then it's far bettr than Save As 06. Try it yourself and examine closely.

 


Leave a comment or suggestion for this page:

(Never Shown - Optional)