To mask, or not to mask? (Non-covid related!)

I was shooting a very well-camouflaged spider last night and as I was choosing a postable shot this morning I thought, boy–it would sure be easier to see this if I masked out the background.

But I haven’t noticed (yet) anyone else using masking in their pics. Lots of use of solid coloured backing cards and such, but not digital background removal.

So, what’s the background on background masking?

BTW, I also saw this piece this morning on NPR.

Exquisitely delightful work by pro nature photographer David Liittschwager, and all masked. But that’s aquatic (not to mention super technical) and I think that his intent to use masking was fundamental to the results. But for everyday amateurs?

Is it too risky in terms of results and research info potential to tamper with the background area?

1 Like

I don’t see a need for this for your photo, it won’t look good, just edit it, it lacks cotrast that would make spider visible, highlights are too high and maybe blacks could be moved down too to help with contrast.
p.s. something like

4 Likes

I just ID’d a bunch of plants yesterday where the observer had digitally blurred the background out to focus attention on the plant they wanted identified. It was quite obvious but worked well for iNat purposes. Too often I run across pictures in Plantae with multiple species growing together and into each other and no clue which one is the target.

That said, I think it’s clear on your picture that the spider is the target and not the plant it’s sitting on. The background behind it is already blurred from the way the picture was taken so I agree with marina_gorbunova that further masking would not be necessary.

Another thing to consider is that background can provide additional data on habitat, communities etc. that can in some cases help with IDs or add valuable info to an observation. So maybe if you mask out the background for easier ID, also add the original unmasked picture to the observation.

7 Likes

This is pretty much my approach too.

I tend not to like masking things (unless it happens naturally as with this Crimson Sunbird or this Pied Kingfisher) as context is often an important, or at least useful, aspect of the observation.

5 Likes

Great photos! I don’t think these are masked, so much as being taken in a portable studio, so the background is already white.

4 Likes

Don’t mask. Crop. Make your subject relatively bigger in the frame.

7 Likes

Many thanks. It’s another question I’ve been struggling with too—do I tamper with the natural lighting to aid in isolating the organism from the background or could that interfere with identification? A lot of grey moths come to mind.

1 Like

Excellent solution. Thank you.

4 Likes

Camera rarely captures real light, so you need to edit to get to it, and using flash would darken background, in my experience it only helped, just make sure with moths you need less light as they’re very reflective.

2 Likes

One thing I’ve done (not for iNat, for other purposes) was to select the background and completely desaturate it. This left only the subject in color, making it easier for my gardener friend to see which plant I was discussing.

1 Like

I agree with others who have noted that background often provides important information that I wouldn’t want to lose with masking. Automated masking can also cut off some parts of organisms that algorithms might not detect well (antennae, really camouflaged parts, etc).

I think cropping works best, especially for small subjects. Since iNat will reduce image size of large photos on upload, cropping also often increases resolution of key characters.

You can also describe in words where to find your focal organism (eg, “perched on the left side of twig”) or I’ve seen some people add a little arrow or line in pointing to a focal organism as well.

7 Likes

Yes, there is no reason not to mask (personally, I don’t know how to do that) to make an observation more visible, but also include the same photo in its natural state. Personally I would put the natural one first and then leave a note about the masked one.
I do mess around with brightness, as my photos may be too dark or too light. I’ve found it hard to get a good light setting against a grey sky or snow. I crop for detail, adjust the brightness/contrast, but don’t mess around with colour. Somehow, in my mind, that is a step too far!

A good way to prevent colour (love the ‘u’, are you Canadian too?) is to edit in LAB mode’s ‘L’ (Luminance) channel for detail with the preview on for all channels. Works well for sharpening, tonal adjustments. But sometimes, the colour truly IS washed out by a poor guess somewhere along the line (silicon or biological software), and you need to step in.

But how much do you step in is always the big question. Colour is exceptionally subjective (not just the spelling) because it is so processed by the human observer. We tend to ignore colour light castings/reflections and replace them with what the optical system ‘thinks’ should be the right match. And we haven’t even touched on the technical rabbit holes of colour calibration between systems and in particular digital screens.

Of course the other danger is that if we become too dependent on digital correction we risk shifting ‘normal’ itself. And in some ways, this is more of a risk with wildlife because everyone really wants the most brilliant, vivid, ‘natural’ look in their shots. But as we all know, natural – and in particular biological – surfaces can vary tremendously dependent on natural lighting conditions, surface optics (iridescence of butterfly wings, bird feathers, for example) and even cognitive viewing biases (colour contrast, as an example).

So if we start to push colours into what we would ‘like to think is more believable’ range we risk encouraging more data distortion. I’m think of all the impossibly vivid shots of flowers and birds I’ve already come across.

Plus, it effects identification and identification skill-building. A lot of organisms utiilzie drab colours very effectively for survivability. Does that mean that we should digitally remove that drabness?

This discussion reminds me so much of much of my career work as a corporate designer and all the times we were provided with very poor, unflattering photos to work with, or when a client (or manager!) would ask us to subtly improve VIP shots. I called it ‘Virtual Realtor’ editing—so named because of the phenomenon that realtors never seem to age at all in their published images over time. But it’s true with almost all public people imaging these days. (With an accompanying negative effect on young women, most sadly.)

Technically, when it comes to identification photography, the professional guides seem to be a little split on their approach to realistic colour. It’s more understandable with illustrated or painted works, for sure. (And yet, ironically, a good painting or illustration actually looks MORE realistic than a lot of photography!)

But how many times have you found yourself thinking, ‘Gee, the specimen I saw didn’t look nearly so vibrant’?

Deep fake iNat observations? Hmm… I sense another topic brewing.

4 Likes

Yes I am Canadian. I live in Winnipeg (for my sins).
I work with very rudimentary editing software, so cropping and brightness is all I want to do. I also identify a lot of Noctuid moths, and have often found that the images that look ‘purple’ (natural lighting) are the best to identify.
I have seen images - on another page - that look convincingly like a red Blue Jay. That is partly why I stay away from colouring images. Natural light, even if it’s not great, renders an image more recognizable than one with altered colouring.
I suspect ‘deep fake’ images would be very rare on iNat!

1 Like

I wouldn’t call purple a natural colour, tint should be at least close to 0, but usually in greens, green enough to not look neither green or purple, to make photo look natural, purple photos look very artificial. Ideally you need to edit more than once, as your eye should get used back to natural light, or you will look at photo a week later and notice all the faults you left.) I talked irl with a person once who was against editing, said it’s not natural, then showed me his photo, it was purple as if it was made on another planet, nothing as what human sees, so I recognised that as “I’m too old to learn this”.

2 Likes

I guess what I meant was ‘natural’ light. I find they show more detail than an image taken with a flash.

1 Like

If you mean automatic flash, yes, it will wash out details.

1 Like

I don’t take a formal approach but I have come across observations of my own where I found it hard to locate the target species despite having taken the photo. One option comes to mind. Use Paint or other software and copy the selected species like cropping a photo but without cropping the background. Then, paste the species cropped photo back into the picture in paint software. The attached photo didn’t need this technique, but in this case, would have saved posting two photos, one showing more habitat and the other showing the species close up (another method used here with Paint software).

Available at: https://www.inaturalist.org/observations/16247813

If the observer has a reasonable number of observations (10-20 or more?) and it’s a good quality photo with multiple plants to ID, I found that @ing the observer asking what they want identified helps, referring to the top, bottom, left, right, color or shape.

I also like @Sedgehead’s approach to boxing photos. I think it’s a bit better than masking because it leaves the background intact which may offer context to help with identification.

I personally go with @efell’s approach of cropping. My first observation photo is usually a zoomed in version of something that I think is most relevant for ID and my last photo is the whole subject. Photos in between are other helpful parts/features and measurements as needed.

Example: Black Nightshade (Solanum nigrum)