My friend sent me a picture today and asked what could possibly cause this phenomena. He said he got roughly 1:100 of this kind of picture. Another one is purple. It's taken by Canon 7D with 580EX flash. (not YN-560, sorry) W/B was set to flash.
Do you have the magenta one that you can post also ... and a regular one. +1 @ RAW data review also.
Looking at the channels, the blue channel looks reasonable, while it kinda looks like the red channel somehow "donated" (or is mia) its share to the green channel. I'm assuming she is wearing some shade of red lipstick. It is showing black in the red channel, but light gray in the green channel which would be too light for a red in the green channel.
Curious to see how this compares with the purple version. It makes me think that the camera somehow assigned the information collected in the individual channels and somehow combined the R&G into just the G. I suspect the purple version will have a similar channel over/under response. I realize I'm outa my league on this one, but that's how it strikes me.
The thing that comes to mind from a lighting perspective is if your ambient was absent of red due to being very cyanic (skylight, north light, stained glass) with sufficient illumination (see green channel @ overexposure??) that your flash contributed very little to the color. @ 1/80, f4, ISO 800 @ 2PM, that could be sufficient illumination that you're flash was very low, or a misfire of some sort wasn't noticed. As to the purple, it would require some other kind of ambient influence (inverse) coupled with a low power/misfire scenario and sufficient ambient exposure.
BTW, what was the position of the flash?
Obviously, I"m grappling and grasping a bit here, but it's all I've got atm ... maybe it'll spark a clue elsewhere.
Here's a strong hue shift of -123 and some desaturation.
This kinda looks like there was incandescent ambient, coupled with a very low power;misfire from the flash, and the WB misalignment @ incandescent wreaked havoc on the shift.
Either way, I'm beginning to think this was a misfire or too much ambient expsoure, resulting in a mismatch of flash, ambient & WB, Any chance your friend "auto" or "click" adjusted the WB, or processed this along with some others (maybe without a low power/misfire)? The strong hue shift bringing things along as this makes me wonder what/when/where an adjustment to the WB was assigned/adjusted.
A plausible scenario ...
Specular flash reflection fools flash to underpower.
Insufficient flash illumination, exposure is from ambient.
Ambient is not balanced with WB, major shift occurs.
Took a stab at a color version salvage from a renewed persepective.
Likely not very close to actual, but I'd like to think the last one is an improvement over the OP.
kenyee wrote:
Bad memory card would be my guess if it's not a bad camera...
visualist wrote:
Yepp, most certainly a memory / conversion problem. Though with both camera or card, a restart / format should solve the problem.
RustyBug wrote:
You say that with a degree of certainty ... based on ... how can you tell? How does a reformat fix it/prevent it in the future?
Unless the shot was made under a green light, it's almost certain that the problem was the red and blue channels not being recorded properly. If all the shots were that way, it could be a sensor issue, but if it's only one in a hundred it's more likely somewhere in the I/O stream. Previous experience suggests to me that it's a card fault, although it could be in the camera.
Reformatting the card sometimes cures recording issues by unscrambling the file allocation structure. (When the card is scrambled, the three channels may not record to the same area, and one or more channel can get "lost" in the mix.) The more photos that have been individually deleted -- as opposed to formatting the whole card -- the more likely this is to happen, although it's still pretty rare in my experience. (I've only had one photo not record properly, and in that case the top half was fine but the bottom was scrambled. My uncle, though, had this exact issue happen to him, and data recovery software was only able to restore a few of them. He reformatted the card, and no more problems.)
The OP mentioned the other one being "purple" ... could it be that the R&B that is absent here (rendering the green dominance), were sent to the other file, giving it the excess necessary to generate the purple? Kind of a "rob Peter to pay Paul" thing with the R&B info being misplaced into the wrong file?
RustyBug wrote:
...The OP mentioned the other one being "purple" ... could it be that the R&B that is absent here (rendering the green dominance), were sent to the other file, giving it the excess necessary to generate the purple?
More likely that on the purple shot shot the green channel got lost, so only the red and blue channels were showing. That's just a guess, though, without actually studying the files.
Hi guys, sorry for the late reply. After I talked to my friend last night, he confirmed that it was 580EX for that particular session, not YN-560 which I stated before. (post updated)
cgardner,
Sample image has been resized with higher compression ratio. It should be faster to load. And, yes, he was shooting RAW. (not full RAW, but mRAW or sRAW) I will post the download link when I got it.
RustyBug,
Thanks for the analysis. I did the same thing on PS, and realized it was red channel's problem. At first, I was expecting that the red channel was completely gone (H/W error), but it turns out there are a few on the image. So, I don't know how to explain this.
Unfortunately, the purple one was deleted immediately. The green one is the second one that catches his eye, so he decides to keep it. According to him, the ambient light is sun light only, and it's very dim at that moment. 580EX was placed on the upper left triggered by YN-622C E-TTL remote trigger.
Keenyee,
He got one Sandisk 16G only. I wouldn't say it's impossible to see a faulty Sandisk, but the RAW file is intact and can be opened without problem. He never had this problem with this particular memory card though.
visualist & BrianO,
I will tell him to reformat the card, thanks. It also appears that his 7D has an older firmware. I already told him to do the firmware update.
As soon as I got the RAW files, I will update the thread. Again, thanks for the input.
phuang3 wrote. According to him, the ambient light is sun light only, and it's very dim at that moment. 580EX was placed on the upper left triggered by YN-622C E-TTL remote trigger.
That's interesting.
Any chance the flash was aimed in such a way that the lens saw the flash directly (akin to a specular reflection) or is reflecting off the bodice and it yielded a very low/underpowered flash due to the ambient @ high ISO. This would leave a disproportionate need for ambient to provide exposure. If it was a dim cool light (cyanic/absent of red) that could account for the extreme lack of red, coupled with just a touch of contribution from the flash (ever so slightly warm). Curious if any of the RAW data can tell us at what power the flash was delivered?
Are there other pics from the same lighting setup?
You can see how cyanic the ambient is in the shadows. This link is to one that is in a pretty bright ambient setting, but with a dimmer ambient, the cyanic cast could be even stronger ... just wondering if cyan + (barely) yellow = green that we are seeing. I know I'm "odd man out" on this, but underpowered flash allows ambient cast. We sometimes see this in rooms lit with tungsten and a weak flash (snaps usually) to give a magenta/purple cast, so it is plausible that we have a weak flash and cyanic ambient to give us a green.
The thing that's kinda got me stuck on the "underpowered flash" is the fact that with the (global) hue adjustment, it readily turned into what looks like an underexposed image in a room with tungsten lighting. If the data was truly absent, I don't think it would have been able to look as it does after the hue adjustment, and subsequent tweaks.
Even though the "purple" one wasn't retained, could you post up a normal image so we can see the proper colors in her dress/flowers/skin?
Unless the shot was made under a green light, it's almost certain that the problem was the red and blue channels not being recorded properly. If all the shots were that way, it could be a sensor issue, but if it's only one in a hundred it's more likely somewhere in the I/O stream. Previous experience suggests to me that it's a card fault, although it could be in the camera.
Reformatting the card sometimes cures recording issues by unscrambling the file allocation structure. (When the card is scrambled, the three channels may not record to the same area, and one or more channel can get "lost" in the mix.) The more photos that have been individually deleted -- as opposed to formatting the whore card -- the more likely this is to happen, although it's still pretty rare in my experience. (I've only had one photo not record properly, and in that case the top half was fine but the bottom was scrambled. My uncle, though, had this exact issue happen to him, and data recovery software was only able to restore a few of them. He reformatted the card, and no more problems.)...Show more →
The card isn't going to kow anything about color channels. All it sees is a file and the file is already built when the camera goes to write the file.
And reformatting can fix the problem because the card controller writes bad block information into the card file table so that bad blocks don't get used. It's not about scrambling anything. The controller autmatically randomizes where it writes on the card to avoid using the same blocks over and over again.
... how can you tell? How does a reformat fix it/prevent it in the future?
