JPEG compression question

[songdog]

I'm wondering how "intelligent" JPEG compression is...

I have a jpeg image that's currently 14 KB in size. The dimensions are 800 wide by 600 high. I'm using the image as a background to a web page, so I'd like to increase it's dimensions to 1600 by 1200 (twice as wide & twice as high) to avoid having to tile the image within the browser window.

Now, I don't want to actually enlarge anything within the image itself. I want to reflect the original image on both its horizontal and vertical axes, as well as on its lower right corner, in order to yield an image with 4x the original surface area. (I hope I explained that so it makes sense.)

My question: Is jpeg compression "smart" enough to know that I'm simply taking the original image, making 3 copies of it, rotating those copies in various ways, and then "stitching" them all together? In other words, will the resulting jpeg image be 56 KB in size, or closer to the original 14 KB?

[kitchin]

It's probably not so smart. It works in 8 x 8 pixel blocks, if I recall. It's not Flash...

Have you tried a GIF? For some images the GIF is smaller.

[doraevon]

Originally posted by songdog:
I'm wondering how "intelligent" JPEG compression is...
[... snip ...]
My question: Is jpeg compression "smart" enough to know that I'm simply taking the original image, making 3 copies of it, rotating those copies in various ways, and then "stitching" them all together? In other words, will the resulting jpeg image be 56 KB in size, or closer to the original 14 KB?

The short, simple answer is no, it is not. However, I just did an experiment with an image to see what would happen if I took and image and did what you described (I think I understood what you're trying to do). Result? An image that had 4x the surface area was only about 2x larger. By adjusting the compression level up to around 70%, I was able to reduce the size even further without excessive image degredation. Unfortunately, a level of compression high enough to approach the original image size would result in too much degredation.

If I remember correctly, JPEG processes images in blocks and does some pretty fancy mathmatics to reduce the amount of information stored. But ultimately, it still has to read the entire image and isn't 'smart' in the way I think you were looking for.

In your case, I suspect that a 14KB original image will likely translate into a 25-30KB image after your transformation. Of course, it also depends on the overall complexity of the image. The only way to find out for sure is to experiment.

Doraevon