----- Original Message ----- From: "Roger Eichhorn" : According to Scott Kelby, in PS CS (re: ISBN 0-7357-1411-8), in : "downsizing" an image, one should turn off the "resample image" : option in the Image Size dialog box. In "upsizing" an image, one : should leave the "resample image" on (bicubic smoother, say), but : increase the size by no more than 10 percent at a time, applying that : formula repeatedly until the desired size is reached. He offers no : explanation for this magical formula but asserts that it works, : attributing the magic to the gurus at PS. the explanation is that a gross adjustment will result in a gross (!) picture ;) Given that the interpolators in PS are very poor quality, his advice is very solid. it's effectively creating a 'stair' interpolation algorithm manually ..i=unless you want to buy the $15 plugin that does this 10% increments for you :-/ : Presumably, the final : result can be saved for whatever purpose one desires. He asserts : that this way, one can create a "poster size" image from a 3.2 : megapixel digital image with little loss in definition. In the book : referenced, he shows the result of taking a 4.75" X 4.75" image to : 20" X 20" with little loss of quality. It's one of the major bones of contention I have with the product.. B-spline, FastLinear, Hermite, Lanczos3, Linear, Mitchell, Nearest, Triangle, Bell, Bicubic, Bilinear can be found in free programs, but PS is restricted to a very lowly set of interpolators indeed.. I've posted these before: http://www.americaswonderlands.com/digital_photo_interpolation.htm different algorithms are better for different jobs, so it's well worth knowing which one is best for whatever you are doing.. seriously, this stuff SHOULD be important, fundamental knowledge to photographers in digital. a basic primer can be found here: http://www.interpolatethis.com/interp.html The b-spline and lanczos alg's are a massive improvement over the pure junk photoshop includes, and these can be free to anyone who wants them. "The nearest neighbour method calculates new pixels according to the value of the nearest pixel in the original. It can produce a blocky result when upsampling and a grainy effect when downsampling. Bilinear resampling operates on a 2 x 2 cell of pixels surrounding each new pixel. The resulting images are smoother but not usually as sharp. The bicubic method looks at the closest 16 pixels (a 4 x 4 pixel cell), computing new pixel values using cubic splines. This method usually yields good results, which can be improved by additional sharpening (not more than 20%). Other interpolation methods are also available. One popular option is Lanczos resampling, which operates on a 4 x 4, 6 x 6 or 8 x 8 cell of pixels surrounding each new pixel location. The 4 x 4 method is so similar to bicubic interpolation that most photographers prefer the other options, which produce better results - but take longer to compute. Lanczos resampling is provided in the freeware IrfanView Image Editor. An alternative is Genuine Fractals, a Photoshop plug-in developed byAltamira .. Genuine Fractals can work well, provided it's applied to large enough files. Files are first saved in the application's native format, after which they can be enlarged as if they were vector images. Any further editing should be done after resizing to avoid potential artefacts. Fred Miranda's Stair Interpolation (SI) also works as a Photoshop plug-in. It uses Photoshop's bicubic method, one small step at a time, but in a rapid automated succession. The Dutch Shortcut Software company's S-Spline 2 applies bicubic interpolation on the smoother parts of the pictures, but uses proprietary interpolation techniques in sharper areas. Like IrfanView, S-Spline is a stand-alone program. " http://www.photoreview.com.au/Articlexasp/a9c228b2-25b9-49f8-81f0-51423c994 b43/Default.htm nice page of photography links.. software at the bottom: http://infohost.nmt.edu/~armiller/cameras/cameras.htm k
