How to resize an image in Photoshop. : LUSENET : Large format photography : One Thread

I am shamed to say, I have been working with digital images for a few years and it's only today that I realized the way I was downsampling images was wrong. The resampling option in Photoshop is set to "bicubic" by default and is said to be the mode that gives the best results. So I just believed what was said and never looked any further. But when downsampling to a small JPG, the results were awful and whatever method I used to add some sharpness, it never gave a fine faithful image of what I had in the big file. So today for the first time I tried the "bilinear" mode and to my great relief, the results are way better. After adding some unsharp masking with the "Ultra Sharpen2.0" script by Robert Barnett and (kindly suggested in another thread), I could do with the quality although I'm sure it could be better. What are your tricks for fine, detailed JPGs and perhaps your experiences with other softwares (Genuine Fractals..)? Thanks!

-- Paul Schilliger (, February 14, 2001


Paul -

A couple of items that I've come across: the bicubic resampling is the best in Photoshop. And, when you make the images extra sharp, and then save as a JPEG, you're going to create much larger file sizes. Assuming you're intending to use images on the Web, and not for print, then add a bit of gaussian or other blur, and it will result in a much smaller file size.

If, however, you're planning on printing the images, then Genuine Fractals has been reviewed in an extremely favorable light. It's worth the investment.

If you're using the images for anything but print, and are fortunate enough to be working on a Mac, then I would suggest you try the program GraphicConverter. It has a batch processing element (that works in that mode only if you pay the fee) that is absolutely fabulous, to say nothing of the other features that make this the best software investment anywhere, anyhow, at any price.

If you're not using a Mac, well then....


-- Alan Agardi (, February 14, 2001.

Paul, just a question, before you downsize the file, was it sharpened? If so, this can cause poor results...

-- Bill Glickman (, February 14, 2001.

Unless these files are destined for the web, never...repeat NEVER use JPEG! JPEG compression is an interpolating format, and sacrifices clarity for size. When saving in JPEG, the system chooses pixels by shading, and then discards surrounding pixels, thereby making the saved file much smaller. During decompression (the next time you open the file) it "guesses" at what these pixels should be. The obvious results are loss of original quality. The more the file is opened, the more the quality is degraded. For print or any other media, other than web use, TIFF or some other non-compressive format is the only real way to go.

Contrary to what the previous post states, JPEG files will always be smaller than the original format. How much smaller depends on the level of compression used.

...and regarding Mac's... we won't even go there.....:)

-- Matt O. (, February 15, 2001.

OK, I should have mentioned the JPG's are for the web. Not to lead you into extensive explanations, I should precise that downsizing images to normal print sizes has never been a problem whatever the mode used. The Photoshop standard tools are good enough for that purpose. But when making a very small image for the web, regardless of what format it is saved in, using the bicubic interpolation mode produces rather poor results (bicubic is perhaps designed for adding pixels?). Bilinear is much better and in fact should be good enough unless there is an even better method that someone know? Saving JPG's, I use the quality increment 8 in standard mode, a good compromise between quality and size.

-- Paul Schilliger (, February 15, 2001.

Bill, no, I never use software sharpness when scanning and wait until the image is final to give some.

-- Paul Schilliger (, February 15, 2001.

The difference you're seeing is the way that edges and straight lines, especially diagonals, are re-sampled using bi-linear as opposed to bi-cubic sampling. Bi-cubic will always add some 'anti-aliasing' interpolation to straight edges. This makes it far superior for increasing the size of an image, but can give much softer looking edges when downsizing.
Having said that, I've never bothered to change resampling modes from bi-cubic when reducing the size of an image. I find that either the standard sharpen filter followed by some 'fade sharpen' works well, or the use of the unsharp mask tool. Maybe your images contain a lot of straight lines, like architecture pics perhaps?

As for JPEGS degrading every time they're opened, that's simply not true. They do degrade every time they're re-saved, and in fact will actually increase in size on re-saving. This is due to the introduction of false colour artefacts, and the sharpening 'ripple' around outlines being seen as additional detail.
If any afterwork has to be saved part way through, the only format to use is TIFF. JPEGS are small and nasty.

-- Pete Andrews (, February 15, 2001.

Pete, your explanation on the two different methods makes sense. May I suggest you try the bilinear mode however next time you make a small JPEG ? Although not visible on larger images, you may notice the difference very easily on small images. In fact the difference is such to me that I will have to redo all the small images I have done before! As you explain, the bicubic mode does mix up some pixels to avoid the aliazing or moire effect, but small details greatly suffer from that intervention when the image has just a few pixels to render those tiny details. For me, bilinear is now the way to go for down sizing and bicubic is to be used only when bringing up a file size- which I never do in practice. In fact I just realized I could change the default mode in Photoshop preferences and I did.

-- Paul Schilliger (, February 15, 2001.

I'll certainly give it a try Paul. I've only had call to make a few thumbnails for the web, and I must admit that they did look very soft, and needed a good deal of sharpening. Bi-linear is probably the way to go for thumbs. Cheers.

-- Pete Andrews (, February 16, 2001.

I just saw this today and will add a few thoughts. Before you play with your images in Photoshop, you scanned them and saved that file to somewhere. I scan files for the ultimate purpose of producing prints of those images at sometime in the future and if you plan on printing out these images then you've got to talk about how you scanned them in. Output in line pairs and the percentage scale are two quantities that you set to a certain optimum and they go a long way to contributing to a good looking print.

Remember that regardless of how you scan your images in, they will look great on your monitor but that doesn't mean that saving what you're looking at on the monitor will be the same way it prints. If you scan in a 35mm neg and you want a good looking 8x10 print(say a fujix or lambda), you've got to start thinking of 120 or 150 lpi and scaling to 800% at least IMHO. It's how you scan the image in and THAT resolution is what is ultimately going to affect the quality of your print. Oh sure you can play with image size in Photoshop but that is not the same thing as the amount of resolution you dialed in when you first scanned the image in.

You know you're going to ultimately print an image you have to 8x10, then you're fighting yourself when you set your output and scale so that it is optimized for 4x5. My rule of thumb which some people might consider overkill, is to set the resolution way bigger than I intend to print and that way I can always go down in print size without the print looking like mush. If I change my mind and decide to print larger than I had anticipated, I don't have to go back and re-scan. If you've got a scan set up for 4x5 and you dial in the image size in Photoshop in an attempt to come up with a 8x10 print of that same file, it's not going to be the same quality as if you had re-scanned the image and set you output and scale for a 8x10 print.

Hope this helps

-- Jonathan Brewer (, February 18, 2001.

Never admit you have learned something or people will take you by the hand to elementary school. Oh well, I may have done it for others. Sorry!

-- Paul Schilliger (, February 18, 2001.

My answer wasn't just for you. You weren't speific about what you do or don't know about scanning, so I told you what works for me. There was no intent to sound paternalistic and if I came off that way then I'm sorry. I was trying share what I know with everybody, and even though you might even know more about scanning than I do, someone else who doesn't know the basics might be able to use the info.

-- Jonathan Brewer (, February 19, 2001.

Jon, sorry for my impatience. My remark wasn't just for you either, please don't be offended. I sometimes place an answer after reading a thread hastily and when going through the content and previous answers again, realize I was a bit off topic or was just repeating what others had already said. In this thread, I was pointing out the difference of quality between the bicubic and bilinear mode in Photoshop when preparing small images for the web. Some of the factors explained in other answers, as yours, are very important but are on a different level (yours concerns printing). If I shared this finding, it's because I wanted to spare some other photographers the false maneuver of using the default bicubic mode for the purpose of making small images and getting rather coarse images that way. But of course all photographers already knew what I thought was a major discovery! Cheers.

-- Paul Schilliger (, February 19, 2001.

Paul, though I thoughtI know a lot about Photoshop your discovery was new to me and of elementar relevance. I've already forwarded this to many friends and companions. Thank's!

-- Jan Eerala (, February 19, 2001.

The kind words and clarification are understood and accepted. It is a pleasure to converse intelligent and mature grown-ups. Thanks

-- Jonathan Brewer (, February 19, 2001.

Moderation questions? read the FAQ