Slashdot Mirror
Reuters Bans RAW Photo Format (petapixel.com)
grcumb writes: Reuters is the latest agency to join the ranks of the technically clueless who think that ethical problems can be solved using technical means. They recently issued a circular to their contributors, stating in part: "In future, please don't send photos to Reuters that were processed from RAW or CR2 files. If you want to shoot raw images that's fine, just take JPEGs at the same time. Only send us the photos that were originally JPEGs, with minimal processing...." The problem they claim to be addressing is doctored images, but they don't explain how they plan to ensure that the JPEGs weren't simply exported from RAW files with their EXIF data altered, or heck, just altered as JPEG. They also assert that getting JPEG files straight from the camera is quicker, which is fair enough. Lots of professionals shoot with RAW+JPEG at newsworthy events. They can send the JPEGs off quickly to meet the first deadline, then process the RAW files at leisure for higher quality publications.
A lot of executive decisions boil down to demands to solve a problem (e.g. photos may be doctored) and an executive deciding he has to "do something", else when it does blow up he did NOT do "something". For example, if an unknown terrorist might strike, it doesn't matter whether the action (ban refugees at a state level) actually matters, it's insurance that when something did happen that you demonstrated precaution. CYA
Gently reply
As a former photojournalist, I can saw that you simply blacklist them and/or fire them from being a contributor/stringer/staffer at that image bureau. There are ethical standards in the professional photography world, and it is nothing bad to those of us who upheld our high ethical standards to see someone get fired for unethically altering images and cheating and breaking the rules. I doubt this is as much a problem from a "who altered their photos?" problem as it is the photographers are submitting larger files (even if lossy down converted into JPG from RAW) and Reuters is having problems handling so many large files in their infrastructure and pushing photos out in distribution to their newsroom client "on the wire" servers. I know in my past when dealing with AP, if you uploaded a file that was too large they either rejected it, or WORSE, applied their lossy compression using whatever software they saw fit. When what your image looks like is everything to a shooter, and when a perfect images is ruined by crap third party compression due to file size, the lesson is hard learned and PJ folks are pretty savvy getting the best bang per MB.
What's a JPEG at -90% compression? Whoa. Mind blown.
They're not trying to prevent "doctored" images.
The original memo reads:
I’d like to pass on a note of request to our freelance contributors due to a worldwide policy change.. In future, please don’t send photos to Reuters that were processed from RAW or CR2 files. If you want to shoot raw images that’s fine, just take JPEGs at the same time. Only send us the photos that were originally JPEGs, with minimal processing (cropping, correcting levels, etc).
And a follow-up quote reads
While we aim for photography of the highest aesthetic quality, our goal is not to artistically interpret the news. [...] Speed is also very important to us. We have therefore asked our photographers to skip labour and time consuming processes to get our pictures to our clients faster.
Which doesn't mean they're trying to prevent people from faking photos; as that line is clearly referring to the "minimal editing" part of the above guidelines, and the "JPG not RAW" is just for workflow-related reasons.
No. Processing raw files involves more than just compression, it includes things like demosaicing and setting white balance.
I am pretty sure the real issue is file size and standards, not doctoring. As manufacturers keep ridiculously upping sensor MP size, photo sizes continue to balloon to larger and larger sizes. RAW files are notoriously huge and non-standard. The extra processing they are referring to is probably just the need to convert those various RAW files back to JPEG, which takes/wastes time/energy by their staff.
You would have to be a pretty big idiot to think that JPEG files are harder to doctor than RAW files. Any photo format can be used when exporting a doctored image... has nothing to do with how it is saved.
Is a JPEG at 0% compression a RAW image?
It would be close but not exact. The way you would get close is to set the 8x8 quantization matrix to all 1's. In JPEG compression, the image is divided into 8x8 blocks, discrete cosine transformed, elementwise divided by an 8x8 quantization matrix, rounded to the nearest integer, and then (usually) Huffman encoded. The primary problem with being perfectly lossless is that the DCT produces a fractional result. So even if you set the quantization matrix to all 1's, the rounding step would lose information.
Care to enlighten me as to how one sets jpeg compression to 0%?
It's not easy to do in most image editors; even the highest (12) quality setting in Photoshop has quantization. You can do it in ImageMagick, however.
Also, no, RAW formats are not simply uncompressed, but largely unprocessed data as well (certainly less processed than what you get from an out of camera tif or jpf.)
Raw formats are indeed compressed; they're just losslessly compressed.
Finally, there is a true lossless JPEG format, though it is distinct from the usual JPEGs.
Make them think it was doctored by a child.
They have no idea how real photography works. JPG is a 'final' format. You capture an image on an SLR as RAW so you get all of the information the sensor can give you, and then you process it to pull the JPGs you want to give to the user of your shots. In journalism, many photographs are taken under marginal conditions, such as four stops below optimum in a sandstorm. Shooting RAW gives you the most latitude to recover usable images that might give us the ability to identify a terrorist. You can apply high dynamic range processing to a single RAW frame to show detail not recoverable any other way, and given a bracket of five RAW frames one stop apart, even handheld, you can postprocess them into a great picture.
Yes, today's journalism photography is being done with many devices that shoot JPG as their native mode, and as any photographer will tell you, the best camera in the world is the one you have with you. But anyone who prohibits high-detail RAW imagery is a person who does not deserve to be in journalism. Manufacturers have responded to the phone-photography challenge with formats like Micro Four Thirds, which gives you SLR versatility in a compact body and lens format that you can take to wherever the news is being made.
A camera sensor is arranged like this. The sensing elements aren't paired up with 1 red, 1 green, and 1 blue, and the values stored by the elements have a larger range than the values in a JPEG (0-4096, say, instead of 0-255). A RAW image is essentially a direct dump from the image sensor, minimally processed and stored in a known format (CR2 is an example). A rasterized image is rendered from the raw data (includes white balancing and other processing), and then it's encoded into a standard image format (usually JPEG).
It is pitch black. You are likely to be eaten by a grue.
The whole purpose of shooting raw images is to do advanced processing later. However, any such processing involves creative choice which alters the image to the taste of the person doing the processing. It's easy to alter the white point and have some journalistically important details lost in the shadows.
Also in a high stakes case suspected forgery, it may be possible to detect forged images by looking at minute noise and encoding choices made by a particular camera model. Faking these details well enough to fool the experts would be beyond the expertise of most would-be forgers.
Of course, Reuters could ask for RAW files themselves and have even more fidelity/authentication potential. But those files are huge, many journalists do not have a fast internet connection where they work, and the publisher would need expertise on RAW workflows.
All in all, I think it's a reasonable decision and will be successful against unintentional/unconscious alterations and causual forgery.
I would suspect that "often" is really "always". A typical Canon RAW file, for example, has 14 bits per pixel. Because of the extra precision, the effective dynamic range of a RAW file is dramatically wider than the dynamic range of a JPEG image. For example, if you have the following samples in the RAW image:
65500, 65532, 65515, 65533, 65473, 65535
And you convert that to JPEG, you'll get:
255, 255, 255, 255, 255, 255
Iff you later need to pull the highs down to make them less blown out, if you're starting with the RAW image, you'll get a fairly accurate rendition of those values (up to the limits of the sensor), whereas if you start with the JPEG image, you'll get white blotches, because there's no detail there to recover. For recovering highlights and/or shadows, JPEG doesn't even come close to RAW, and can't. There's just too much data lost.
Check out my sci-fi/humor trilogy at PatriotsBooks.
To answer in a more technical way (than "use ImageMagick").
JPEG encoding inherently can be completely lossless. 8x8 pixel squares of pixel values are converted to 8x8 matrices of frequency components - transforming the representation of data as a superposition of specific sine waves of fixed set of frequencies and parametrized amplitudes. Due to small area and range of values being covered, this mapping is lossless - the data is sufficient to recreate the exact image, errors of the "floating point nature" of sine waves being less than 1 bit of value representation.
Then, depending on the settings of the software - the "compression rate", the parameters of lowest values and of highest frequencies are replaced by zeros. The "quality" parameter decides how many, and how significant ones. Unlike with direct value function which would leave black pixels, with sine waves this leaves the characteristic "artifacts" of JPEG, a kind of wavy imprecision along any sharp edges, some colors being misrepresented etc. This is not very visible to human eye, and you can get away with zeroing half and more of the parameters without significantly altering the perceived image.
And then this is compressed using a standard lossless data compression.
The gain comes from the fact that strings of zeros, repeating zeros and such compress very well - much better than "random" data of standard image.
Of course if you don't strip any zeros and keep all the values, JPEG will be lossless and you won't gain anything size-wise, the compression being equivalent to standard lossless ones. And still you can lose relative to RAW, because the original (input) data uses 8-bit color channels, so 3 8x8 matrices of bytes per one "block". RAW can keep much more bits per pixel.
45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
You can think of it, very basically, as Reuters insisting on a 4x6 physical print, rather than wanting the negative.
In this analogy, it's easier to work with the print than the negative, but if you want to scan it back to digital, blow it up, whatever, you're losing quality. With the negative, you have more work to do, but you have a higher-quality starting point, and you can do all sorts of work with it and get far better results than by working from the 4x6 print.
Vintage computer games and RPG books available. Email me if you're interested.