[Award Winner] Bitmapped Images - The technical side of things explained.

[Redrobes]

Would you like to take the middle image from post #11, take the top left small bit and scale it up very high (800%) and cut out the bit of petal like the others so we can compare ?

[RobA]

Here is another freeware application that provides a few other enlarging modes:
http://logicnet.dk/Analyzer/

Here are some results:

Wiener 4:
weiner4_resize.png

Wiener 6:
weiner6_resize.png

Sinc:
sinc_resize.png

XinLi:
xinLifractal_resize.png

Fractal 5 - XinLi:
fractal5xinli_resize.png

-Rob A>

[Redrobes]

Thats brilliant Rob. Are you able to edit the post and label them as to what they mean. I have to say I do not think that they are an improvement over ordinary methods tho. The one Mathuwm suggested is critically acclaimed and supposed to be very good indeed so I am curious. It appears I have repped you enough already

[mathuwm]

resampled with Genuine fractals

[Redrobes]

I am expecting great things from this super resampler but I cant accept this. This is too good. Is this the small image in the top left of the middle image of post 11 - I have attached it again below. How can it reproduce perfect stamen from a single pixel... tell me this was from a different image !

Can you do the image attached at 800% or x8 mag ?


Edit -- I can see that this must have come from top left of post #2 and not post #11. Its still very good but it would be great to have a side by side comparison from the same source.

[mathuwm]

the 120x96 pixal image

No i do not think that would work to well.

the larger the image the better it upscales

this program is used a lot for upscaling images for printing on a large format printer

what i did was take this:
Image size-1.png

and scale it to this

I actually cropped it after I scaled it so the actual differences might be slightly off

Image size-2.png

[Redrobes]

Well the little one is whats being used for all of these samples just so we can compare what a small image resampled would become under different algorithms. Its looks from the dialog images as your using Bicubic in any case - unless it says that when using the Genuine Fractals anyway.

Just for reference, here is my GTS doing it. I will have to look at the source to see what kind its using. I think its Lanczos tho.

Thanks for putting the labels on yours Rob. Interesting ringing on the sinc. That looks a bit over the top for a proper sinc tho. I think there might be a coding issue going on in that one. Sinc in a box window will ring a little but that seems a bit too much to be correct.

[mathuwm]

Temp2.png

ok so here is your tiny image scaled up

[mathuwm]

small detaleTemp3.png

[RobA]

A few more with that freebee Image Analyzer I referred to (quite the swiss army knife, really!), using the sample image, enlarged 800%

Wiener 4:
wiener 4.png

Two of the fractals (no idea what the parameters do just playing). I got the best results by going x2, x2, x2 rather than straight (the image names include the parameters):
resize_Temp x2 region=7 regul=0 x2 region=7 regul=0 x2 region=7 regul=0.png
resize_Temp x2 region=9 regul=0 x2 region=9 regul=0 x2 region=9 regul=0.png

And two more from Gimp - progressive resize, 800% in 20 steps:
gimp_resize_steps.png

and a gimp plugin "smart enlarge" using resyntherizer(?) I enlarged 2x 2x 2x. The first was OK, but after that it certainly added detail, but bolloxed up the image:
gimp_resynth_1.png

-Rob A>