Ask the owner of the site for the HiRez version of the files. (I assume that an internationally reputed client has professionals building their site)
As for the type, you can copy the text from HTML and paste it in InDesign or Quark for editing. If there are no HiRez files, try bilinear at 720ppi (from 72).
The images are supplied by researchers (the research result) to get it published with our client, which is a reputed publishing company.
Most of the authors fail to meet the resolution specification, laid-down by our client. So, there is no question of blame.
Now my question is simple. How can I increase the image quality with minimum information-tamperer. And an advice of the most effective method to do so.
nearest is the way to go…heres why.
nearest neighbour is more preferential over bilinear or bicubic becuase there is limited re-sampling across the pixels….HOWEVER, there is one BIG trick to it….dont increase to a specific DPI, simply keep doubling the DPI until you have enough. If you ONLY double then you will see no difference between your original and your high res except for the extra resolution, you WILL NOT see any distortion of type etc because the doubling enables the extra pixels to be put in equal spacing.
we have done this for about the last decade, and there has never been any problems with the method or the output, despit the pickyness of corporate clients.
Depending on the type of rip that gets used, the nearest neighbor method may or may not be a complete waste of time.
If the rip does not reinterpolate rasters, then all you accomplish with nearest neighbor is to create the exact same image with more data.
So going from 72ppi to 288ppi you end up with a block of identically colored pixels containing 16 pixels (4×4) in place of the original single pixel. The appearance is exactly the same.
If however the rip to be used will reinterpolate rasters (like Scitex or Artwork Systems) then it is wise to do the nearest neighbor bit, or the screen shots will turn to mush.
But the long and short of it is, you can’t really increase the quality of the screen shots, unless you reduce their size without resampling.
I’ve heard some talk about Genuine Fractals for this but I can’t offer an opinion.
The only thing I will add to Prog’s post is that you don’t need to keep doubling the resolution when using nearest neighbour, you need to upres to a resolution that is a multiple of the current one. thus if your file is currently 72 ppi, you can upsample to 144, 216, 288, 360, 532, 604, 676 ppi and so on.
Rene, You stopped short of 720 (my suggestion in post #1 : )
I guess Lince wants advice that will sound like a miracle.
Genuine Fractals is really designed for resampling-DOWN. When I tried it, I found it did a less satisfactory job than Photoshop for resampling-UP.
I havent seen it really surpass itself beyond satisfying people that someting has been upressed "properly" or that their web image will now safely work on their billboard…theres nothing wrong with NN as a solution…i prefer it to nothing because rips can balls up images if they up sample using an unsuitable method.
Rene, You stopped short of 720 (my suggestion in post #1 : )
It’s the multiplication of two two-digit numbers that had me confused 😉
Lince,
Are you sure you need to resample?
What are the pixel dimensions of the images and what pixel dimensions do you need for output? Dpi is only relevant when going to print. If you have a very large (number of pixels) image to start with that has to be printed at a small size you may not have a problem. Little or no resampling might be needed. I see screen shots in books and magazines all the time.
For instance, if I take a shot of my whole 17 inch screen at "72 dpi" as reported by Photoshop mt screenshot is 1280x 1024 pixels. Resize to "500 dpi" without resampling gives me a print size of 2.56 inches, with none of the information lost.
BTW note that Photoshop gets my screen resolution wrong since I don’t really have a 17.77 inch wide screen.
Ed, your screen is closer to 96 dpi than 72.
1280/96=13.33"
1024/96=10.67"
And uncle Pythagoras tell us that the diagonal is about 17.4"
Correct. Photoshop doesn’t know that though and so reports at 72 dpi. Which is after all, meaningless.
What Lince asks makes total sense. He has original image data that will go to hard copy as part of articles he is creating. He must submit in the specified resolution to make sizes work in the publication’s workflow. However, apparently the publication is tolerant of less than perfect image quality (probably dependent on the image) which also makes sense.
So multiply the original ppi times 6 or 7 (whichever necessary to exceed 500 ppi) using various interpolation methods, and test print to see which interpolation performs best with those particular files.
Thanks to one-and-all who have flooded the thread with sensible information. I have got a clear picture of the scope of up-sampling and their scientific aspects.
Thanks
Lince M Lawrence