«

QUESTION 1: What is actually being done to the jpeg image on my
little compact digicam to make it "fine", "normal" or "economy"?

I tried to get some insight by calculating the bits used in a JPEG for
each pixel. The user guide gave me the photo's pixel size and the MB
space needed for each of the three quality levels. From that data I
did some calculations below.

QUESTION 2: What I can't understand is the big difference between
bits needed per pixel between the 6M pixel picture and the 2M pixel
picture.

--------------------

6M pixel picture = 2816 × 2112 pixels = 5,947,392 pixels.

FINE: 3.63MB = 3806331 bytes or 30450647 bits. 5.12 bits/pixel
NORMAL: 2MB = 2097152 bytes = 16777216 bits. 2.82 bits/pixel
ECONOMY: 1.21MB = 1268777 bytes = 10150216 bits. 1.71 bits/pixel

---

2MB pixel picture = 600 × 1200 pixels = 720,000 pixels

FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel


--

Specs say: JPEG (Exif Ver.2.2); DCF 1.0 standard; DPOF compliant.

Sandi <> wrote:
>QUESTION 1: What is actually being done to the jpeg image on my
>little compact digicam to make it "fine", "normal" or "economy"?

JPEG compression can be done on several different levels. "Fine" is the
lowest compression, i.e. it compresses the least but retains the most
details. "Economy" compresses the most resulting in the smallest file
size, but at the cost of loosing the most details.

It is really trading file size for quality of the photo.

http://en.wikipedia.org/wiki/JPEG#JPEG_compression has a pretty good
explanation, see in particular the attached chart.

>I tried to get some insight by calculating the bits used in a JPEG for
>each pixel. The user guide gave me the photo's pixel size and the MB
>space needed for each of the three quality levels. From that data I
>did some calculations below.
>
>QUESTION 2: What I can't understand is the big difference between
>bits needed per pixel between the 6M pixel picture and the 2M pixel
>picture.
[List of numbers snipped]

All those numbers are only estimates or long-term averages. The actual
amount of compression possible for each level varies quite a bit and
depends on the actual content of the picture.

jue

Sandi schrieb:
> QUESTION 1: What is actually being done to the jpeg image on my
> little compact digicam to make it "fine", "normal" or "economy"?

The selection of the quantization parameters. The parameters are whatever the manufacturer
decided them to be for those settings.

> I tried to get some insight by calculating the bits used in a JPEG for
> each pixel.

That's, of course, only an average number.

> The user guide gave me the photo's pixel size and the MB
> space needed for each of the three quality levels.

That's also only an approximation. How large the image will be depends on the image.
(Quite different from JPEG2000, actually, where you can compress to a specific size).

> From that data I did some calculations below.
>
> QUESTION 2: What I can't understand is the big difference between
> bits needed per pixel between the 6M pixel picture and the 2M pixel
> picture.

Why can't you understand that? Seriously, I don't quite get the question. JPEG is a lossy format,
the amount of loss depends on the quantization parameters, and those are selected by the
manufacturer.

>
> --------------------
>
> 6M pixel picture = 2816 × 2112 pixels = 5,947,392 pixels.
>
> FINE: 3.63MB = 3806331 bytes or 30450647 bits. 5.12 bits/pixel
> NORMAL: 2MB = 2097152 bytes = 16777216 bits. 2.82 bits/pixel
> ECONOMY: 1.21MB = 1268777 bytes = 10150216 bits. 1.71 bits/pixel

So well.

>
> ---
>
> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>
> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel

More details, less blurry, worse compression.

The relation between the compression rate (as in bits per pixel) and the
quantization parameters depend on the nature of the image. Try to make a snapshot
of the sky and see *how well* that compresses!

So long,
Thomas

Sandi wrote:
[]
> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>
> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel

Are you sure you have the number of pixels correct? Is it really 600 x
1200, or perhaps 1600 x 1200 => 1920000 pixels?

David

Sandi wrote:
> QUESTION 1: What is actually being done to the jpeg image on my
> little compact digicam to make it "fine", "normal" or "economy"?
>
> I tried to get some insight by calculating the bits used in a JPEG for
> each pixel. The user guide gave me the photo's pixel size and the MB
> space needed for each of the three quality levels. From that data I
> did some calculations below.
>
> QUESTION 2: What I can't understand is the big difference between
> bits needed per pixel between the 6M pixel picture and the 2M pixel
> picture.
>
> --------------------
>
> 6M pixel picture = 2816 × 2112 pixels = 5,947,392 pixels.
>
> FINE: 3.63MB = 3806331 bytes or 30450647 bits. 5.12 bits/pixel
> NORMAL: 2MB = 2097152 bytes = 16777216 bits. 2.82 bits/pixel
> ECONOMY: 1.21MB = 1268777 bytes = 10150216 bits. 1.71 bits/pixel
>
> ---
>
> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>
> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel
>
>
Not sure if that reasoning is valid. The pixel count (or resolution)
doesn't change as result of compression. Compression means the file size
is being reduced with or without information loss.

Normal jpeg compression is lossy. It seems to work quite well for
reasonable compression values. The easiest way to see this is to take a
high quality image and save it in an editor using varying compression
factors and judge for yourself.

Most cameras also provide for shooting at different resolutions. Most
people prefer to shoot at maximum file resolution to allow for cropping,
resolution can always be adjusted later in an editor.
My canon defaults to fine and I can't distinguish between this and extra
fine. Others may not agree.
Dave Cohen

On Dec 5, 9:03*am, Jürgen Exner <jurge...@hotmail.com> wrote:
> Sandi <n...@email.com> wrote:
> >QUESTION 1: *What is actually being done to the jpeg image on my
> >little compact digicam to make it "fine", "normal" or "economy"? *
>
> JPEG compression can be done on several different levels. "Fine" is the
> lowest compression, i.e. it compresses the least but retains the most
> details. "Economy" compresses the most resulting in the smallest file
> size, but at the cost of loosing the most details.
>
> It is really trading file size for quality of the photo.
>
> http://en.wikipedia.org/wiki/JPEG#JPEG_compressionhas a pretty good
> explanation, see in particular the attached chart.
>
> >I tried to get some insight by calculating the bits used in a JPEG for
> >each pixel. *The user guide gave me the photo's pixel size and the MB
> >space needed for each of the three quality levels. From that data I
> >did some calculations below. *
>
> >QUESTION 2: *What I can't understand is the big difference between
> >bits needed per pixel between the 6M pixel picture and the 2M pixel
> >picture.
>
>[List of numbers snipped]
>
> All those numbers are only estimates or long-term averages. The actual
> amount of compression possible for each level varies quite a bit and
> depends on the actual content of the picture.
>
> jue

Unless it's simply academic curiosity, what's the point? Shoot at the
highest resolution, using the largest card your camera will support.
You'll get better pictures; simple as that. There's no such thing as
too much information in the file to work with during post processing.

Back in the day, when a 64MB card cost a hundred dollars, worrying
about shooting resolution and card capacity made sense. But with the
price and capacity of cards today, even if you shoot in RAW and
produce files up to 20MB each, it's a non-issue.

"David J Taylor" <david--this-
bit.co.uk> wrote:

> Sandi wrote:
> []
>> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>>
>> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
>> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
>> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel
>
> Are you sure you have the number of pixels correct? Is it really 600 x
> 1200, or perhaps 1600 x 1200 => 1920000 pixels?

I'm sure that's the answer, since the bits per pixel then come out almost
identical to the 6MB photos.

Having said that, it would actually make sense to increase the bits per
pixel for smaller photos, as they're likely to be printed or viewed at a
lower resolution which would make any artefacts more visible, and since the
file size is going to be inherently smaller you can afford to be more
generous with the settings.

(I can't imagine ever going as high as 13.63 bpp though, since you may as
well use lossless compression, in fact I doubt any jpeg would ever give
such low compression: the highest size I've ever got using photoshop's
maximum jpeg setting was around 10 bpp, and that was for 35mm scans where
you have film grain to contend with.)

Marty Fremen wrote:
[]
> Having said that, it would actually make sense to increase the bits
> per pixel for smaller photos, as they're likely to be printed or
> viewed at a lower resolution which would make any artefacts more
> visible, and since the file size is going to be inherently smaller
> you can afford to be more generous with the settings.
>
> (I can't imagine ever going as high as 13.63 bpp though, since you
> may as well use lossless compression, in fact I doubt any jpeg would
> ever give such low compression: the highest size I've ever got using
> photoshop's maximum jpeg setting was around 10 bpp, and that was for
> 35mm scans where you have film grain to contend with.)

You need to be careful here. Some cameras - Kodak digitals - seem to
produce very poor JPEG images. Now I haven't checked carefully, but I
think that there may be an art in selecting the parameters for JPEG
compression to best match the sensor and other characteristics of the
camera, and that perhaps Kodak didn't choose the best values. On the
other hand, the Nikon cameras I've owned have always done well even with
more compressed JPEG images.

So I wouldn't regard bpp alone as a measure of compression quality. I'm
sure you're aware of that, and when comparing on the same camera it's
probably a fair thing to measure.

On my 6MP Nikon DSLR, "normal" compression is good enough for me. But on
my 7MP compact camera, which can produce noisier images at higher ISOs, I
find that "fine" is better. Probably about what one would expect....

Cheers,
David

Sandi wrote:
> QUESTION 1: What is actually being done to the jpeg image on my
> little compact digicam to make it "fine", "normal" or "economy"?
>
> I tried to get some insight by calculating the bits used in a JPEG for
> each pixel. The user guide gave me the photo's pixel size and the MB
> space needed for each of the three quality levels. From that data I
> did some calculations below.
>
> QUESTION 2: What I can't understand is the big difference between
> bits needed per pixel between the 6M pixel picture and the 2M pixel
> picture.
>
> --------------------
>
> 6M pixel picture = 2816 × 2112 pixels = 5,947,392 pixels.
>
> FINE: 3.63MB = 3806331 bytes or 30450647 bits. 5.12 bits/pixel
> NORMAL: 2MB = 2097152 bytes = 16777216 bits. 2.82 bits/pixel
> ECONOMY: 1.21MB = 1268777 bytes = 10150216 bits. 1.71 bits/pixel
>
> ---
>
> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>
> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63 bits/pixel
> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel
>
>
The compression is not done on a "per pixel" basis. The number of bits
per pixel remains the same for each pixel. Rather, the compression
scheme essentially eliminates pixels in the stored file, then
"recreates" them when the algorithm re-expands them. For instance, if
there is a large blob with the same "color" the algorithm only stores
markers for the region, and one value for the color, but then recreates
all the pixels when expanded. So file size depends highly on the actual
image. Files with lots of fine detail do not compress as far as, say, a
sunset picture.

Don Stauffer wrote:
> Sandi wrote:
>> QUESTION 1: What is actually being done to the jpeg image on
>> my
>> little compact digicam to make it "fine", "normal" or
>> "economy"?
>>
>> I tried to get some insight by calculating the bits used in a
>> JPEG
>> for each pixel. The user guide gave me the photo's pixel size
>> and
>> the MB space needed for each of the three quality levels. From
>> that
>> data I did some calculations below.
>>
>> QUESTION 2: What I can't understand is the big difference
>> between
>> bits needed per pixel between the 6M pixel picture and the 2M
>> pixel
>> picture.
>>
>> --------------------
>>
>> 6M pixel picture = 2816 × 2112 pixels = 5,947,392 pixels.
>>
>> FINE: 3.63MB = 3806331 bytes or 30450647 bits. 5.12
>> bits/pixel
>> NORMAL: 2MB = 2097152 bytes = 16777216 bits. 2.82
>> bits/pixel
>> ECONOMY: 1.21MB = 1268777 bytes = 10150216 bits. 1.71
>> bits/pixel
>>
>> ---
>>
>> 2MB pixel picture = 600 × 1200 pixels = 720,000 pixels
>>
>> FINE: 1.17MB = 1226834 bytes = 9814671 bits. 13.63
>> bits/pixel
>> NORMAL: 780KB = 798720 bytes = 6389760 bits. 8.87 bits/pixel
>> ECONOMY: 390KB = 399360 bytes = 3194880 bits. 4.44 bits/pixel
>>
>>
> The compression is not done on a "per pixel" basis. The number
> of bits
> per pixel remains the same for each pixel. Rather, the
> compression
> scheme essentially eliminates pixels in the stored file, then
> "recreates" them when the algorithm re-expands them. For
> instance, if
> there is a large blob with the same "color" the algorithm only
> stores
> markers for the region, and one value for the color, but then
> recreates all the pixels when expanded. So file size depends
> highly
> on the actual image. Files with lots of fine detail do not
> compress
> as far as, say, a sunset picture.

Yes. This is well demonstrated by comparing a night shot of the
moon with a daylight scene with lots of detail. At the highest
jpeg quality setting in my camera, the daylight shots come out at
3 to 4 MB, whereas a moon shot takes up around 0.35 MB - a ratio
of ten to one. The dramatic reduction in file size for the moon
shot happens because the actual image of the moon takes up only a
small fraction of the whole scene and there is little variation
of color and brightness over the rest of the picture.