Michael Romano
Well-known member
I just finished a 10 day shoot for National Geographic off the coast of Maui using a Phantom HD Gold and an EPIC underwater shooting humpback whales. This EPIC was in a GATES housing. As a side note, I'm impressed with the housing and would recommend it to others. Very solid.
That being said, we had a strange issue with the image coming from the EPIC. Below are before and after corrections done only in RCXPRO. We were not rigged for secondaries as I didn't have the ability to transcode on location so I had to have RMD files which would be good enough to get us through editorial. The post production team did not have extensive (any) experience in digital cinema as most of their jobs are shot on Sony F800 XDCAM systems.
THE PROBLEM:
Essentially, the water becomes magenta and the animals go a teal hue. The whales are actually all shades of black and grey and any color they represent is going to be a reflection of their surroundings. This is a proper pain to fix in primary CC because to correct the water, you remove magenta and the whales then become SUPER teal. The same works in reverse.
My DP, Paul Atkins, has told be he has seen this problem before with a *different* EPIC off the coast of Guadalupe in much cooler seas with great white sharks as the subject. I was unable to get a chance to review this footage but I'm told it looks identical.
Now I know things look odd underwater but this seems to be an abnormal image for a camera. We had a GoPro HERO3 getting the exact same action and it looks as you would expect in regards to underwater color.
We did not use any underwater specialty filtration in the front of the lens as some shots are captured at up to 100ft while others are at the very surface, all on the same dive.
I'll start by detailing the things we tried to fix the issue.
1 - Black Balance. We determined the median temperature of the camera when in the housing, rolling, underwater. In Hawaiian seawater, and with the Gates housing, its about 57º C. We calibrated at that temperature. No perceptible change in the image.
2 - IR contamination. It sure looks like it to me but I have found numerous sources that say IR cannot penetrate water. That seems to make sense as you loose most, if not all, red information the deeper you go. We mostly shoot near the surface, but dive down to get some shots. It seems as if it is evenly effecting the picture at any depth. If it is IR - I would think only the surface shots would be contaminated. I will be doing further testing down this path. Stay tuned for details.
3 - Color correction in RCXPRO. This saved us on this shoot. I think my before and after shots fix it up pretty well but it wouldn't be an easy fix for the uninitiated colorist. Not to mention, it is still a little funky. Not the proper solution.
4 - Although I didn't test it, a red filter for black & white film (At a lower density) might force the sensor to read red channel information. You would then just have an inverter LUT to correct it.
Any thoughts? I'm hoping someone else has come across this thing before!
***THESE PICTURES ARE HUGE***
Let them download and then compare on your desktop. I did TIFF grabs from RCXPRO and then used Photoshop to convert to JPEG @ 12 Quality (BASELINE)
That being said, we had a strange issue with the image coming from the EPIC. Below are before and after corrections done only in RCXPRO. We were not rigged for secondaries as I didn't have the ability to transcode on location so I had to have RMD files which would be good enough to get us through editorial. The post production team did not have extensive (any) experience in digital cinema as most of their jobs are shot on Sony F800 XDCAM systems.
THE PROBLEM:
Essentially, the water becomes magenta and the animals go a teal hue. The whales are actually all shades of black and grey and any color they represent is going to be a reflection of their surroundings. This is a proper pain to fix in primary CC because to correct the water, you remove magenta and the whales then become SUPER teal. The same works in reverse.
My DP, Paul Atkins, has told be he has seen this problem before with a *different* EPIC off the coast of Guadalupe in much cooler seas with great white sharks as the subject. I was unable to get a chance to review this footage but I'm told it looks identical.
Now I know things look odd underwater but this seems to be an abnormal image for a camera. We had a GoPro HERO3 getting the exact same action and it looks as you would expect in regards to underwater color.
We did not use any underwater specialty filtration in the front of the lens as some shots are captured at up to 100ft while others are at the very surface, all on the same dive.
I'll start by detailing the things we tried to fix the issue.
1 - Black Balance. We determined the median temperature of the camera when in the housing, rolling, underwater. In Hawaiian seawater, and with the Gates housing, its about 57º C. We calibrated at that temperature. No perceptible change in the image.
2 - IR contamination. It sure looks like it to me but I have found numerous sources that say IR cannot penetrate water. That seems to make sense as you loose most, if not all, red information the deeper you go. We mostly shoot near the surface, but dive down to get some shots. It seems as if it is evenly effecting the picture at any depth. If it is IR - I would think only the surface shots would be contaminated. I will be doing further testing down this path. Stay tuned for details.
3 - Color correction in RCXPRO. This saved us on this shoot. I think my before and after shots fix it up pretty well but it wouldn't be an easy fix for the uninitiated colorist. Not to mention, it is still a little funky. Not the proper solution.
4 - Although I didn't test it, a red filter for black & white film (At a lower density) might force the sensor to read red channel information. You would then just have an inverter LUT to correct it.
Any thoughts? I'm hoping someone else has come across this thing before!
***THESE PICTURES ARE HUGE***
Let them download and then compare on your desktop. I did TIFF grabs from RCXPRO and then used Photoshop to convert to JPEG @ 12 Quality (BASELINE)
I wish I was a color science expert to explain my process from a scientific perspective. All I can say is I have done the UW Macbeth Chart test you described, and found that the color channels are much easier to recover if the footage is shot with the WB set accurately. I discovered this by accident when I bumped my WB button and shot a days footage with WB set to a wildly inaccurate value. No amount of CC will recover this footage, thereʻs just insufficient information present, especially at depth. I have a theory that the demosaic is more accurate if the WB is in the ballpark, but like I said, Iʻm no expert. The only reason I also record the card is to have a concrete reference in post. Hereʻs a test dive I shot to test my process, no filter involved. The "before" is RG3 with WB set for the scene. The "after" is WB set to the card plus typical CC methods.
