New at this! resizing images for pdf?

[KarenFPL]

I'm working in InDesign and creating a long document with close to 100 images. I've been told that I can resize my images in InDesign (rather than doing so in Photoshop) and then export to PDF for press. Is this correct? The images I'm working with are all 2592 by 1944 at 96 ppi (27 inches wide by 20.25 high) and they will all be reduced to a final size of about 3.5 inches high at 300 ppi. Right now, they are 16-bit RGB from a microscope camera. If I resize in InDesign, will I lose quality and/or cause problems for a printer when providing a PDF that I'm assuming will be quite large? Or, am I better off resizing to 100% in Photoshop before placing in InDesign? I believe this is what a printer told me to do. Also, do I need to convert the files from 16-bit down to 8-bit for press? Thank you....I really need some concrete advise from print professionals! Karen

[Colorblind]

It really depends on what your printer can handle. Personally, I would import and scale the original images in InDesign, this way you'll preserve the images integrity (color space, bit depth, resolution) until PDF creation or ripping time. Once the document is ready, you can export it to the PDF flavor your printer prefers and choose to downsample image resolution so it doesn't exceed 300dpi which will help keeping your document to a decent size.

[kyle]

If you are supplying a PDF and not the native files, definitely do it in Indesign. Then you can resample images as the PDF is made using the settings in the PDF export dialog. You can also have RGB converted to CMYK automatically without having to edit each image (if all CMYK is what you want to send). I believe 16 bit images will always be quantized to 8 bit also.

If you are sending the native files, then the size of what you send would obviously be lower if all of your image links were resampled and changed to 8 bit (and as ZIP compressed TIFF to preserve all detail, or as JPEG if you don't mind the lossy compression). This would reduce file transfer time and probably some processing time at the printer, but it probably wouldn't be worth all the extra time it would take you, unless perhaps you ran a batch automation with a Photoshop action.

[gordo]

How does the quality of the image differ by having it resampled in a PDF compared with having it resampled in PShop? Do you get the same final result either way?

thx, gordon p

[kyle]

How does the quality of the image differ by having it resampled in a PDF compared with having it resampled in PShop? Do you get the same final result either way?

The result from exporting is not the same as any of the options in Photoshop. It is better for most images. If you have an image that is 500 ppi resampled to 400 ppi, then you are reducing every five columns to four and every five rows to four, or a bunch of 5x5 matrices to 4x4 matrices. For every 4x4 matrix of final pixels, the upper-left 3x3 will be kept exactly as-is from the original pixels, and the last row/column pixels will be a 50/50 averaging of the last two rows/columns from the original pixels. The lower-right pixel ends up being a 25/25/25/25 averaging of four source pixels.

This preserves sharpness almost as well as the nearest-neighbor method, without completely ignoring entire rows/columns like the nearest-neighbor method does. It expect that it would usually produce less moire than the nearest-neighbor or bilinear methods for patterned images. It is also probably the fastest algorithm second only to nearest-neighbor.

Bilinear interpolation would logically be a more space-preserving method, but consider a corner-case scenario of reducing a 1 x 1 inch image from 301 ppi to 300 ppi. If you used bilinear, the outermost pixels of the final image would be nearly identical to the outermost pixels of the source image, but the pixels in the exact center would be 180 degrees out of phase with each other, meaning the center pixels are going to be an equally weighted average of four pixels. Averaging out four pixels effectively blurs the image, so the image would be preserved on the outer edges and blurred in the center. If the source image were 302 ppi, then it would phase in and out twice instead of once. The export resampling method, however, would exactly preserve all of the pixels except for the last two rows/columns, squashing them down to a single row and column. The blurring is then limited to a single row and column instead of a large area, and the blurring is half as strong because the averaging is of two pixels instead of four. For a 302 ppi source image, there would be two rows and two columns of averaging.