Jpeg image sizes?

[Valvebounce]

Hi Folks.
I was just spending some time getting to grips with DxO. I exported some images to file as Jpegs, they were in the order of 20Mb. I thought I could get away with half size for what I wanted as I have no intention of discarding the raw file!
I see a scale in DxO export settings, Quality 0-100, must be percent right? That is the first assumption that will prove a point! ;D
It obviously isn't percentage, but what is it, from trial and error I find that somewhere in the 96-97 range gives half size files! Please see attached list.


Dxo jpeg sizes by Valvebounce25, on Flickr

The blue highlighted set is from one image, I realised it was a crop when I went to compare the raw file so I repeated the 95-100 set and added the raw for comparison. Can anyone tell me how this scale works, it just seems completely random that a quality of 50/100ths (%? ;D) can give a file size of 1/16th the full size file.
Someone must know the answer to this, please! It's making my brain ache! (How do you draw a smiley with smoke coming out the ears?) ;D
Thanks in advance for help with this, I know it is not necessary to know the answer but enquiring minds need (like) to know.

Cheers, Graham.

 

[Mt Spokane Photography]

I can give you some lightroom values for comparison.

With my D800, a raw nef image was 51.2MB

Exported to jpeg at 100 it became 21.8mb

Exported to jpeg At 75 it became 5.48MB

Exported to jpeg At 50 it became 2.58MB

 

[mackguyver]

As you've both posted, it is a percentage, but it's related to the amount of compression, not the resulting file size. At 100%, it's nearly uncompressed and quite large. As you increase compression (lower percent), the algorithm throws away more and more data and in essence becomes more efficient leading to a psuedo-logarithmic reduction of file size. Here's plenty more if you want to geek out:
http://en.wikipedia.org/wiki/JPEG#JPEG_compression

 

[Valvebounce]

Hi Mt Spokane.
Thanks for that, I guess that all software will use a similar scale, I believe even the gimp which I used for the attached image has the same sort of scale, I just never wanted to reduce the image size below maximum before as I only ever fiddled with out of camera Jpegs which are small enough anyway!
I was interested in the logic behind the scale as much as the sizes, though I will say I'm glad my files are not as big as yours!

Edit.
Hi Mackguyver.
I think our replies crossed in the post as it were! Thanks for the response, I shall have a look at the link, not sure I want to geek out too much though!

Cheers, Graham.

 

[Marsu42]

Thanks for that, I guess that all software will use a similar scale

Definitely not, and there are even some really weird scales like LR/ACR that isn't linear at all. See: http://regex.info/blog/lightroom-goodies/jpeg-quality

Also unknown to some, there is another quality factor that has a noticeable impact on the filesize - subsampling. Some software changes it from medium quality 2x1 to hq 1x1 on very low jpeg compression like 90+, that's when you see a strange filesize bumb. See: http://www.impulseadventure.com/photo/chroma-subsampling.html

 

[Mt Spokane Photography]

With Lightroom, you set the DPI of the jpeg image, so its another variable. I sold the D800, at high ISO, those 50MB nef files uncompressed to over 150MB and took a long time to post process. I used a huge amount of NR, and then any resolution advantage was lost. Since I tend to use mostly high iISO like 12800, the camera was not the right choice. My Nikon 24-70 2.8G had a ton of CA. I thought it was defective until I read the fine print in the reviews. I was using it at 24mm to take group photos, but the edges had so much CA that Lightroom could not correct it. There is more to a camera system than a sensor.

 

[StudentOfLight]

I think the content of the image plays a role in the final file size. If there is a lot of repitition of data, the smaller a resulting compressed file can be. So a file with for example lots of gradiated blue sky will result in a more data efficient compressed file. A noisy image with lots of fine textures will result in a larger compressed file.

 

[privatebydesign]

With Lightroom, you set the DPI of the jpeg image, so its another variable. I sold the D800, at high ISO, those 50MB nef files uncompressed to over 150MB and took a long time to post process. I used a huge amount of NR, and then any resolution advantage was lost. Since I tend to use mostly high iISO like 12800, the camera was not the right choice. My Nikon 24-70 2.8G had a ton of CA. I thought it was defective until I read the fine print in the reviews. I was using it at 24mm to take group photos, but the edges had so much CA that Lightroom could not correct it. There is more to a camera system than a sensor.

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

 

[Mt Spokane Photography]

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

Makes sense, I have no excuse for my senility

 

[privatebydesign]

And neither do I

 

[Marsu42]

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

Makes sense, I have no excuse for my senility

But I can understand the notion because pixel processing software like LR seems to make makes a large fuss about dpi settings, but it only matters for import into desktop publishing and is irrelevant for other print/screen purposes.

 

[Valvebounce]

Hi privatebydesign.
I had read that in previous posts, makes you wonder why the products all have DPI so prominent in the menu?
DxO has the DPI setting in an expanded box above a collapsed re-sample box, took me ages to find the settings to resize an image!
Do as many people as the manufacturers think actually use these softwares to output to print?

Cheers, Graham.

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

 

[privatebydesign]

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

Makes sense, I have no excuse for my senility

But I can understand the notion because pixel processing software like LR seems to make makes a large fuss about dpi settings, but it only matters for import into desktop publishing and is irrelevant for other print/screen purposes.

I don't think LR makes a large fuss over dpi, it is just a tag. But it is what gives you your print size unless you deliberately specify something different, and if you still have a version of PS that has the "Print View" option in the zoom settings and you have set your monitor up correctly then it gives you a pretty accurate WYSIWYG idea of your print. I use Print View all the time and that 100% is driven by dpi.

 

[privatebydesign]

Hi privatebydesign.
I had read that in previous posts, makes you wonder why the products all have DPI so prominent in the menu?
DxO has the DPI setting in an expanded box above a collapsed re-sample box, took me ages to find the settings to resize an image!
Do as many people as the manufacturers think actually use these softwares to output to print?

Cheers, Graham.

DPI does not affect jpeg file size. Number of pixels does, as does content (more detail bigger files), noise, iso (because of the noise), quality setting etc.

DPI doesn't just relate to print, it relates to what size you see your image on screen in some modes. It might be a legacy concept much of the time, but that is down to our misunderstanding of the idea rather than manufacturers being stuck in a previous time.

I use DPI as a key component to handling files non distructively on a daily basis.

 

[Valvebounce]

Hi Folks.
Well I have read some from most of the links you all sent me, Mackguyver, I got lost by the time I got to the maths on the Wikipedia link you sent, I couldn't geek out that much! Marsu, I had a look at the links you sent, incredible how good the blind looks at low quality that messes the sun blind, interesting about the chroma sub-sampling.
I think I am none the wiser about the answer, though I do now understand a bit more about the quality level not needing to be 100 to get acceptable images, in some instances it would appear that a lot lower (50?) might be perfectly acceptable.

Thanks for trying to educate me folks much appreciated.

Cheers, Graham.