best to resample jpegs to 50% or 25% of original?

I read in a post that if a jpeg is to be downsampled then it is better to downsampled by 50% or 25%.

It seems this is something to do with not disturbing 8x8 block DCT boundaries in the jpeg. I don't understand this level of detail myself.

My questions are

(1) How significant is the difference in the final 72 dpi version between downsampling to 55% for example, compared to 50% ?

(2) Does the 50% and 25% refer to a the linear dimension of the image such as the image height? Or does it refer to the total number of pixel points in the image?

Jon D wrote:

I read in a post that if a jpeg is to be downsampled then it is better to downsampled by 50% or 25%.

It seems this is something to do with not disturbing 8x8 block DCT boundaries in the jpeg. I don't understand this level of detail myself.

My questions are

(1) How significant is the difference in the final 72 dpi version between downsampling to 55% for example, compared to 50% ?

Why don't you simply try it yourself? You are the only one whose judgement is important if it comes to your own images.

(2) Does the 50% and 25% refer to a the linear dimension of the image such as the image height? Or does it refer to the total number of pixel points in the image?

Linear, but guess what? If you downsample 50% linear, you will downsample to 25% of the total pixels.

--
Johan W. Elzenga johan<<at>>johanfoto.nl Editor / Photographer http://www.johanfoto.nl/

Jon D wrote:

I read in a post that if a jpeg is to be downsampled then it is better to downsampled by 50% or 25%.

It seems this is something to do with not disturbing 8x8 block DCT boundaries in the jpeg. I don't understand this level of detail myself.

There are MANY ways to downsample an image, so without more context my answers would be purely guesses...

--
Phil Frisbie, Jr.
Hawk Software
http://www.hawksoft.com

Jon wrote:
) I read in a post that if a jpeg is to be downsampled then it is ) better to downsampled by 50% or 25%.
)
) It seems this is something to do with not disturbing 8x8 block DCT ) boundaries in the jpeg. I don't understand this level of detail ) myself.

Downsampling any image is always better done by factors that are 'nice'. In other words, factors that are expressed as rations of small integers.

50% is 1:2, 25% is 1:4. Other good values would be 1:3 or 1:5, and only slightly worse are 2:3 or 2:5

On jpeg, I would guess that the fact that it divides an image into 8x8 blocks means that factors of two are extra good.

) My questions are
)
) (1) How significant is the difference in the final 72 dpi version ) between downsampling to 55% for example, compared to 50% ?

Between 55% and 50%, I would guess a big difference. But that depends on what downsampling method you use, and on the contents of the image.

Photos, for example, are much more flexibly scalable than drawings.

) (2) Does the 50% and 25% refer to a the linear dimension of the ) image such as the image height? Or does it refer to the total number ) of pixel points in the image?

It refers to a linear dimension.

SaSW, Willem
--
Disclaimer: I am in no way responsible for any of the statements made in the above text. For all I know I might be
drugged or something..
No I'm not paranoid. You all think I'm paranoid, don't you ! #EOT

(1) How significant is the difference in the final 72 dpi version between downsampling to 55% for example, compared to 50% ?

Where you would see a big difference is if you tried to use something other than bicubic resampling. With nearest neighbor resampling you'll see aliasing all over the place if you're not using 50% or 25% or something. Actually in certainly situations I've found nearest neighbor resampling to be better than bicubic, because it maintains more sharpness. This can make a big difference when making tiny thumbnail images, as bicubic will tend to just make them too smooth and barely recognizable.

ation for down sampling refers to the linear pixel dimension of the file. It stems from the way that Photoshop handles aliasing when downsampling to a given percentage. It produces the best image when the reduction is 1/2, 1/4, 1/8, 1/16 etc. This also applies when viewing the file on your monitor. Ignore the dpi and pay attention to the pixel dimension as discussed above.
Regards, David "Routemeister" Thompson
http://home.rochester.rr.com/backroads/

"Jon D" wrote in message

I read in a post that if a jpeg is to be downsampled then it is better to downsampled by 50% or 25%.

It seems this is something to do with not disturbing 8x8 block DCT boundaries in the jpeg. I don't understand this level of detail myself.

My questions are

(1) How significant is the difference in the final 72 dpi version between downsampling to 55% for example, compared to 50% ?
(2) Does the 50% and 25% refer to a the linear dimension of the image such as the image height? Or does it refer to the total number of pixel points in the image?

If you want to save the image again as jpeg, then this matters, because in case of 1/2, 1/4 or 1/8 downscale you can indeed preserve the DCT coefficients. This means you don't have to completely decompress to pixel level and recompress (which would add additional loss).

For viewing purposes, it matters mostly speedwise, because in case of 50% you can use a very simple algorithm to downscale, vs a much slower resampling technique at 55%. You can also read in the Jpeg faster, because you need only half of the DCT coefficients.

An example of this are viewers that generate thumbnails of jpeg images. They often read them in at 1/8, which means they only have to read 1 value of each 64 value DCT block. This one "average value" is also stored differently, simpler, than the other 63 ones, so you get a huge speed increase.

(2) Does the 50% and 25% refer to a the linear dimension of the image such as the image height? Or does it refer to the total number of pixel points in the image?

Refers to linear.

Nils