DP2m: first images, first problems

mclaudio wrote:

yes, I have seen the problem only in clouds (expecially in high contrasted zones), in fog conditions, and in some snow conditions

All this hints at it being a conversion problem, due to non linearities in the brighter areas. It is very hard to convert the Foveon image when it starts to go non linear in any of the channels.

For RGB cameras, like Bayer, this is not a big problem. If one of the channels start hitting the non linear region, then the image only gets more neutral, i.e. no so much of the clipping channel.

But for foveon, all three resulting RGB chanels are depending on all three foveon RAW channels. If one start clipping, all three are affected, some with positive sign and some with negative. Ops, the blue sky is green or purple. The same for clouds.

digi2ap wrote:

Two more thoughts:

One is that clouds may need more careful exposure than I thought and maybe in this regard don't respond well to ETTR. Looking at the X3F in SPP again the problem relates to areas of cloud that I had exposed a little too far to the right - the clouds needed -0.3 on the Exposure slider to bring them under the highlight warning (I was at the time more concerned about the shadows).

OK

ETTR is probably a bad idea for clouds then, maybe always.

Now although the problem isn't visible in SPP I just tried the Highlight Control slider moving it fully to the left (Neutralise) and then I exported the TIF into ACR and from there exported a JPG. The colour issues in the JPG are no more and so maybe this is the solution? When I next get a moment I can try and post the findings.

That it is more visible in JPEG is not strange at all. When converting to JPEG, there are some compression of tonality made in bright areas.

Roland Karlsson wrote:

mclaudio wrote:

yes, I have seen the problem only in clouds (expecially in high contrasted zones), in fog conditions, and in some snow conditions

All this hints at it being a conversion problem, due to non linearities in the brighter areas. It is very hard to convert the Foveon image when it starts to go non linear in any of the channels.

For RGB cameras, like Bayer, this is not a big problem. If one of the channels start hitting the non linear region, then the image only gets more neutral, i.e. no so much of the clipping channel.

But for foveon, all three resulting RGB chanels are depending on all three foveon RAW channels. If one start clipping, all three are affected, some with positive sign and some with negative. Ops, the blue sky is green or purple. The same for clouds.

Makes a lot of sense to the problem under discussion.  Do you think the Highlight Control works to neutralise these colour issues?

Updated JPG using 0.5 Highlight Control.

Roland Karlsson wrote:

digi2ap wrote:

Two more thoughts:

One is that clouds may need more careful exposure than I thought and maybe in this regard don't respond well to ETTR. Looking at the X3F in SPP again the problem relates to areas of cloud that I had exposed a little too far to the right - the clouds needed -0.3 on the Exposure slider to bring them under the highlight warning (I was at the time more concerned about the shadows).

OK

ETTR is probably a bad idea for clouds then, maybe always.

Now although the problem isn't visible in SPP I just tried the Highlight Control slider moving it fully to the left (Neutralise) and then I exported the TIF into ACR and from there exported a JPG. The colour issues in the JPG are no more and so maybe this is the solution? When I next get a moment I can try and post the findings.

That it is more visible in JPEG is not strange at all. When converting to JPEG, there are some compression of tonality made in bright areas.

Yes, with RawTherapee's X3F default Merrill decode (SD1M only), negative EC was also needed to bring the highlights into view on their excellent histogram - even in their native ProPhoto working space.

And raw ETTR for the sky in this image would certainly blow some of the building/construction highlights, since the highest sky values ("blue") are only ~3200 out of 4095 in RawDigger.

In the discolored areas, there is significantly high saturation in the RawTherapee review image - as high as about 26%. Which, of course, is totally incorrect for a daylight sky. It takes but a few clicks upward on the saturation to up the greens and to force the reds all the way to zero (100% saturation) - just playing, not trying to fix it.

I've read that Merrill conversion colors are dependent on captured luminance and that lower levels can mess with the colors. The discoloration at left of the image is -2EV (green) here's a histogram from the left; I've set EV0 to 4095 for this one.

Left Part of the Sky.

Right now, I'm betting on the matrices used by the various possible converters - with SPP remaining "best" but not perfect for the above area.

PS. My experience of the SD1M is that a raw area with green about 1/3EV below blue and red about 1EV below blue will get converted to more or less neutral (with correct WB). That implies that the above raw values will not convert to neutral (red -2EV to blue, green -2/3EV to blue) but still . . . . that can't be the whole story because similar raw ratios are in the sky areas that show as close to neutral.

Perhaps less than perfect lens correction is a factor, combined with raw -> RGB conversion matrices. Speculation, though.

In ACR, is the conversion to JPEG not done by PS or LR? When I had PS Elements 6, the conversion was done by Elements, is why I ask.

So with a more considered SPP conversion I was happy with the end result after all.

I don't post images here other than for technical interest so I'll make the point that all the images in the above carousel are made with DPM's of one variety or other.