Ian reynolds writes: >> I have been messing around using the bicubic interpolation photoshop on files from my d100. I process them as 16bit raw files from bibble. I am interpolating at 125% in four steps. So the file I end up with is roughly a 205mb 16 bit file at at 12 bit is 102mb. <<
I have not looked into high bit tests when performing interpolation, although I have a sneaking suspicion that things may look very subtly slightly softer in high bit. >> I am using unsharp masking at 100% four times with 1 radius and no threshold. << As bicubic interpolation uses a form of sharpening as a native part of the process this could be a bit aggressive depending on content. Then there are the multiple filtering to take into account. I guess you have done tests with no USM or only one or two USM steps or other variables. You may like to explore the 133% or 166% increment resample, then 10% steps or single step to bump up to the final size to compare things with as well. There are the other obvious suggestions like Genuine Fractals, or the lesser known ones such as the free Panorama Tools 'sync' interpolation: http://home.no.net/dmaurer/~dersch/interpolator/interpolator.html http://home.no.net/dmaurer/~dersch/Index.htm >> I am resizing my image to 72dpi then using the view print size function in the zoom tool to approximate the print size on screen. I interpolate in 16 bit to retain as much tonal info as possible. I only tried this as I remembered someone on the list saying that anything that looked good at 100% on screen would print well. << I would qualify that with the statement that if an image is at the appropriate physical canvas size for the desired output and that the resolution was sufficient for the image content and output/viewing conditions...and you are viewing at 100%, 50% or 25% then you have a better idea of pixel quality vs. print quality. >> I also created a duplicate image as an original 18mb file and to compare the two. The other one is risized to the same physical dimensions as the interpolated one. Then using the view print size compared the two. My observation is that you get the jaggies of the pixels visible, but the interpolated one looks better. << Have you tried an interpolation, small gaussian blur then high pass sharpen to the other workflow? Would the result be similar in visual feel? >> My question is, is this method a good way of going around it. The print indicated it would now be 260cm high at 72dpi. Now I now normally you would submit most files at 300dpi. Would I have to change this figure???. I am only asking this as a customer was asking about blowing up really big and I said it would be fine to about A0. Um well being the viewing distance is going to be way back, is this an issue. I know that those with an s2 can go just that bit farther, and I suppose it depends on what kind of media its going to be printed on. With the on screen approximation I would say it looks fine from about 2 meters away. I did here someone on the list saying that they had an s1 file printed onto a 72 sheet poster with no worries. << Let the output and viewing distance dictate things. For a large format inkjet plot, you can usually supply 100 ppi files or lower and get great results - which means that your image is now three times larger than the one that is being used at 300 ppi at same size. So less or no upsampling is required...in some cases you may even downsample or have the ability to send through slightly more data. For a press - the halftone screen will be the guide. Then factor in viewing distance and other variables. Large format inkjet output is _very_ forgiving of the originals when viewed in the correct conditions...or even when up close in some cases. Always work back from output where you can. How large is the output and what resolution is required? Rather than assuming that a device needs higher resolution, often you can get away with far less - which in the case of oversize output can be a big deal. Stephen Marsh. =============================================================== GO TO http://www.prodig.org for ~ GUIDELINES ~ un/SUBSCRIBING ~ ITEMS for SALE
