traciyosh wrote in post #18560225
Thank you - finally have a term for it! I've googled so many variations of descriptors, haha.
So the moire happens in dropbox as well (or possibly just the preview once downloaded) as the client is the one who said "it looks almost over sharpened on her hair" for some photos. Is that something that would show up on a website (I shoot for a local boutique and I think she uses Shopify)? I will do some additional digging to see if it's something I can eliminate altogether because I'd much rather deliver her hassle-free images than have to assure her it'll look great on her website, but it'll still be good to know.
Also, since you mention the resolution: I'm going into senior and family photography soon and will be displaying images on an iPad (I think, haven't decided) for them to choose which ones to purchase. Is there a way to know which images that'll happen with before it's the middle of a sale and they're uncomfortable with it? I have a lot of faith in my reassurance skills but just another situation I'd rather avoid altogether if possible.
Also, thank you for the dropbox input. I'm honestly not a huge fan at all for using it for image delivery, but I also edit from it after failed externals so it's just been the only thing I can find that works well. Any other recommendations that has a Download All option?
What you often have to worry about with Dropbox is that when you view the image on their website they substitute showing the original version of the image with a quite highly compressed, and certainly on mobile devices, a much smaller image. Since they tend to use very high levels of JPEG compression for this it means that your image can suffer from high levels of JPEG compression artifacts. To the point that they are even visible to a client. Clients these days also seem to be very good at the right click save of online images, and will often do this in preference to using the correct download feature. Let's face it they will always go with what they know if possible. Thus you get hit with this as a problem.
The moire issue you got with particularly file IMG_6706 is completely unrelated to the above situation. The full resolution image maybe has a little colour moire on the shirt, but it's really hard to decide to be honest. As I said attempting to fix it IMO made the shirt look a little flat, and so I would have left it alone myself.
The problem really was with the resampling of the image. Not having the original CR2 file to try it's not possible to see what you get from exporting the image directly at 1200px on the long edge. My expectation is that with that particular shot the pattern of the shirt is perfectly matched to the resolution to produce moire. In the digital world moire is almost always the result of exceeding the Nyquist-Shannon sampling limit for digital sampling.
The biggest issue with Nyquist is that if your input data exceeds the cutoff frequency of the digital sample, in this case it is how fine a repeating pattern is in space, you cannot fix the issue after the conversion from analogue (continuous) space to digital. In this case the original sampling frequency was much higher than that of the pattern of the material, so no problems. For no problems you have to have two samples for every repeat of the pattern. Unfortunately when you reduce the resolution of the image, so that you only have 1200 px on the long edge you reduce that sampling frequency. So now you have three or four repeats of the pattern for every sampling point. What happens in that case is the higher frequency pattern is interpreted as being a much lower frequency pattern instead.
The only way to prevent this happening is to put an analogue filter in place before digitisation takes place. That is what the Optical Low Pass Filter in front of a sensor does. It literally filters out the highest frequencies of patterns that are above the Nyquist limit that might be passed by the lens to the sensor. Sensors have much lower colour resolution than for grayscale contrast, hence the slight colour effect being visible if you look really closely at the large image. Once you have recorded the artifacts from the undersampling of the image it is impossible to remove them with a generally applicable filter, as you can in the analogue space. There are some things you can apply manually, such as the moire tool on the local brush in ACR, but that mostly only works for small amounts of colour moire, and not the levels seen in this image.
So this is a problem when you resize a digital image and make it smaller. You are effectively re-digitising it, but you don't have an analogue image to filter out the higher frequency parts of the image. You could do some clever processing and use an FFT (Fast Fourier Transform) to convert the existing image data to the frequency domain, then filter out any elements that are going to be above the cut off frequency of the resized image. You would then need to use the inverse FFT to convert back to the spatial domain, where you can now do the resize operation, knowing that you are not going to be undersampling the image. Doing this on what is a large amount of two dimensional data though is going to be very compute intensive indeed. Another objection to this is that you will need to calculate a new digital frequency cutoff for every single possible set of pixel dimensions that are smaller than the original image, based on the relative values of the size before and after resizing.
For that image you simply can't use 1200px long edge as a viable image size. You need to make the image a little larger, even using 1280 px reduces the visible moire by quite a bit, and the frequency of the moire is less intrusive too. I also tried 1500 and 2048 px too. At 2048px the moire was pretty much removed.
You ask if uploading the image to another website will improve it, and the answer is simply no. Once you have the moire present you can't get rid of it, not even if you resize the image again in either direction. So nothing is going to enable you to be able to use that image at that size without the moire. Sometimes life just kicks us swiftly between the thighs and laughs at us. This is one such occasion.
With the iPad thing you should be OK if you are putting the JPEG files directly on the device. That way you can use the full size image, and if you do get some moire at full screen view, you can zoom into 100% view and show them that the issue is with the display panel, not the image file itself. The print in such a situation will always be OK.
Finally when you get told by the client that you have issues with the images that are being downloaded from Dropbox, or anywhere else for that matter, you shouldn't really just be going to those images directly to fix them. What you need to do is go back to the originals, and QC check them at every stage to find out what the problem really is. When you are in a situation where moire is possible in downsampled images I would always do a QC check, even if that means I have to do it on images uploaded to the website that have been automatically resized, before setting them to be visible by the client.