Lossy Compression & Image Properties (Or how many people talk out of their ass)
There is something that’s driven my nuts for a long time now. It’s rampant at all levels of the photography, print, advertising, web world, and that’s the lack of technical knowledge of the people in all of these businesses. Most of this has to do with image formats and settings and I’m fed up, so I’m writing a primer about it.
Film, since it is an analog medium, has always been something of a black art. A photographer will use film X with developer Y and paper Z to get his particular look. And that’s a best case scenario when they did the work themselves. Think of all the photographers out there who just handed their film over to the high priestess at the lab who did whatever it is she does to make your pictures look fabulous. This is the world before scanners and file formats and technical requirements.
Now it’s different, photographers know what a JPEG is, and they may have heard of a TIFF. However most people don’t really know anything about the differences between the two. They have heard that JPEG is ‘lossy’ compression, and that TIF is ‘lossless’. Well who in their right mind would want to lose data? I guess I’ll use TIFF’s then. Well it’s not that simple. First off, TIFF files are generally about an order of magnitude larger than JPEG which makes them take a lot longer to transfer over the internet for example. And while lossy formats are theoretically inferior to lossless formats, but would anyone know the difference? I have a smaller magazine I do work for who requires TIFF files, but big magazines just want jpegs. I know from my own experience that high quality JPEGs look pretty good compared to the original, but I’ve never taken the time to test it until now.
I took a 100% crop of part of the Golden Gate Bridge in SF straight from a RAW file that came out of my 5D. I then exported it as a JPEG with a quality of 80 out of 100, in some software it’s 1 to 10, so in that case it would be an 8. High-quality, but not ‘best’. I then loaded this saved jpeg back in Photoshop, copy and pasted it on top of the original pre-export and switched the layer blend mode to ‘difference’. This would show me exactly where any differences lie between the two files. You know what I get? A black image, or at least visibly black. I was prepared to show the difference image as an example, but I won’t bother showing a black image. Now, just to test to make sure it was working, I added a levels adjustment layer and had to pull the white end almost all the way down before I saw anything show up, and even then, it was faint dark pixels that were hard to make out.
The point is that hi-quality jpegs are perfectly fine to make prints from. Do I save my final master images as JPEGs? No, I’m the kind of guy who rarely even merges layers in his 3.5GB 16 bit PSD files. However I do export a full-res quality 8 JPEG of my final work and have it uploaded to an off-site backup service. Those are my worst-case-scenario files in the event that my house gets burned down or the world markets collapse and I have to leave NY without my stuff. And if you took that JPEG and edited it and saved as a jpeg, and then edited again, and saved as a JPEG, eventually you would see visible image degradation and nasty jpeg artifacts, but first generation? No way. Maybe graphic designers want TIFF’s because they were the preferred image format for Quark back in the day, but that’s not true anymore. JPEG’s have other limitations such as color bit depth, lack of an alpha channel, and a technically archaic compression scheme, but that’s a whole other article, and those things rarely impact what people are using them for.
Another pet peeve is people’s concept of resolution. I got an email the other day where the woman said, “We would need the photo in no less than 400dpi” and I just stared at the sentence. First off, they were planning to use the portrait at a very small size in a book, from what I understand, probably less than a couple inches. So 400 dots per inch? If so, how many inches? Maybe she meant 400 pixels.
People throw dpi around all the time and few people know how it relates to computer files. DPI or dots per inch is a print specification, it’s related to the physical world, of which digital files and data most certainly are not. Digital image files are a certain amount of pixels wide and a certain amount of pixels tall, as far as dimensions go that’s the only absolute value. Print height and width and resolution are all relative. For our example, let’s say that an image file is 1200px by 1200px. Our image could, for example be printed 4″x4″ at 300dpi, or 6″x6″ at 200dpi, or 12″x12″ at 100dpi. Or even 1″x1″ at 1200dpi.
Telling me that you need a 400dpi image is meaningless. It’s like telling me that you need a place to visit to that’s 6 hours away. Well, it all depends on how fast you’re going. In 6 hours, I could walk from one end of Manhattan to the other, or I could drive to Portland, Maine, or I could fly to London, or I could go around the earth 4 times in the space shuttle. It’s only if you say, I’m driving a car on the highway and want to go somewhere that’s 6 hours away that I can help. Then I can say ok, you’re going to be traveling around 60mph for 6 hours, so that means you want to go somewhere that’s about 350 miles away.
Lightroom for some reason, defaults to exporting images to Photoshop at 240dpi, and sometimes when I reinstall I forget to reset it, or if I do reset it, I make a mistake. Last year I sent an image to an Art Director at one of the major book publishing companies. They had requested a file with a resolution of 300dpi as most people do for print work, but I had set the Lightroom export to 300 dots per centimeter accidentally without noticing (bumped the inches/cm drop down without noticing). I sent off the file to the guy and got an email back about 5 minutes later that said that the file I had sent them was unacceptable. Too low-resolution and that they need a better one. Now the file I sent them was full-res from a 21MP 1Ds Mark III, it was 5616 pixels by 3744 pixels. It was huge, and definitely NOT low-res, in fact, if anything it was too high-res as 300dpc is about 850dpi. I opened the file, looked at the image size in photoshop, saw the error, fixed it, and sent it back to a now happy AD. The thing is that he should have known better. Was the mistake mine? Yes. But his email shows that he doesn’t understand the relationship between resolution and pixel dimensions. It was a 2 second fix, and he had everything he needed to change it himself, but he didn’t. And he’s an Art Director, he should have known better.
So, here are the things to take away from this rant.
1) For final images that you’re done editing, high-quality (quality at 8 or better) JPG files are fine and will likely be indistinguishable from the lossless file in print.
2) Pixel dimensions of a digital file are the only things that really matters. Physical print dimensions and resolution are relative and can be easily changed. So if someone asks for a 300dpi file without specifying dimensions, then they probably don’t really know what they’re talking about.
[I’ve decided to cull through some old essays of mine and repost the relevant ones over here on Medium. I hope you enjoy]