Jump to content

Sean's review - moire


jfgilbert

Recommended Posts

Advertisement (gone after registration)

I just read Sean new comments about moire and the absence of AA filter. First, I will express my thanks and appreciation for Sean's excellent work, his ability to translate equipment capabilities into photographic styles and opportunities, and his willingness to put up with our endless questions. I will also add special thanks for the tip about using the history brush - brilliant and will save me hours of messing with layer masks (I am sure everybody knew that, but I confess I did not).

The point of the thread, however, is Sean's comment that the raw file is not really raw and has been already processed to compensate for vignetting. Surely, Sean, you knew that DNG is not raw. Sensors only read one color, and DNG has three color values per pixel. The two missing values are calculated by a sophisticated interpolation of the values read by neighboring sensors, so there has to be a certain amount of processing. That leads to a couple other thoughts that I would like to submit to the collective wisdom of the forum:

(1) It has been mentionned before and I suspect that Canon may have developed an algorithm to reduce noise at the time the raw data is processed for interpolation. That would make a lot of sense as it seems it would be possible to treat luminance and chromatic noise differently and most effectively by catching it before is gets compounded by the various interpolation schemes. The good news is that, perhaps, someone else - at Leica, for example - will figure out how to do that and the M8's high ISO performance could be improved with a firmware upgrade. Personally, I do not think it needs it that much, but it would be nice :)

(2) It is quite possible that the sensor readings are taken into 16 bit values, that all the interpolation and corrective processing is done with full 16 bit arithmetic, and that only the final results are stored in the 3x8 bit values that we see - which, by the way, has more than an 8 bit dynamic range. That might explain the difference between Leica's specifications and the file format.

Jacques

By way of introduction, I was interested enough in photography as an amateur to set up my darkroom about 30 years ago. Gave up after a few years because it was too much trouble to get good results, then came back with the arrival of digital cameras (I remember when 1 megapixel was high resolution...) Went through a number of Canon and Nikon DSLR's until I figured out that it was the bulk and weight that bothered me (it must have helped when someone "relieved" me of my bag with D60 and the 16-35L). Found a compromise with a D-Lux 2 and learned to live with its limitations, but can't wait for my M8.

(sorry, too insecure to post a link to my pictures)

Link to post
Share on other sites

Actually focussing on those comments about moire...

 

Sean, can the moire be seen on the in-camera display? My understanding is that moving your shooting position even slightly is enough to dampen the interference effect so that you have to work quite hard to get it to happen.

 

The DMR has an option to apply moire correction but I assume this is only for the JPEGs. Does it work? Looks like they have omitted this option from the M8. Do the JPEGs show the same moire?

Link to post
Share on other sites

There are some incorrect statements in the post above. DNG is indeed a standardized package for raw sensor data, but some cameras do modify this data as simply as possible on the way to saving it. Leica's vignetting correction (for coded lenses) and Canon's noise filtering are examples of processing done before the raw file is saved. Each sensor is represented in a raw file by a single value. Interpolation between nearby R, G, and B elements happens in software when the file is developed.

 

Finally, every M8 dng file is 10.2MB in size (I've looked at some, and the instriuctions and the specs say this). There are 10.2 M pixels in the imager, so one Byte is being saved to represent each pixel, not three, which would have resulted in a 30MB file.

 

And now I have to go see what Sean added about getting the benefit of no AA filter (something I have been looking forward to).

 

scott

 

(edit): I see we all jumped on you at once. Sorry about that.

Link to post
Share on other sites

Advertisement (gone after registration)

This is a really muddled, rambling, thread!

 

I think it helps if you are a thread starter to focus on just one topic, otherwise the thread goes off topic before it's even started. Mind you, the woolly thread title should have told me.

 

Jacques, about the only thing you've missed out discussing is the prevailing weather in Timbuktu...

Link to post
Share on other sites

Guest guy_mancuso

I can only speak of the DMR at this point and the moire issue if you want to call it a issue. leica has chosen thankfully not to use a Anti Alaising filter like Canon and like any medium Format digital back they also are CCD sensors with no AA filter. At certain high frequencies I guess they call them Nyquist frequencies you will get the effect of Moire , however it is pretty rare and not a big issue as it was on older MF backs in the beginning. On the DMR it comes and goes , somtimes i may see it in certain clothing material that has a certain pattern like a tweed suit or a tie. Now in C1 Pro there is a PS plug-in call Demoirize that after raw processing in PS you can use this filter on a affect area . I use the lasso tool and feather it than apply the filter and you have control of this also to eliminate the Moire'. Sounds complicated but it is real simple and also not a big a issue as it sounds to be. i have not read Seans update yet but from my reading of the M8 leica seemed to find a way to have a cut off on the sensor of the Nyquist limit or process this out with the Raw file and the Jpeg , reason your not seeing a moire filter on the camera like the DMR which only works with jpeg BTW. I guess this is something new to the M8 but i would think there still maybe some coming though at times, again i have not read Seans update but any case it can be removed fairly simple in Post. There are also some other PS techniques to elminate this in PS also, the filter is just one example.

Link to post
Share on other sites

This is a really muddled, rambling, thread!

 

Jacques, about the only thing you've missed out discussing is the prevailing weather in Timbuktu...

 

I couldn't find the current weather, but the average conditions for each month in Timbuktu are available at Timbuktu Weather Forecasts on Yahoo! Weather .

 

HTH (;<))

 

scott

Link to post
Share on other sites

They usually use chroma-smoothing to fight with moire, and it will kill fine color details and desaturate some areas on image. And sometimes large high-frequency areas of clothing texture is so affected to aliasing artefacts (mostly due to specific interpolation algorithm) so you need to cut-off too much details.

From my observations ACR algorithm has good antialiasing properties, but it has a lot of disadvantages such as low sharpness. Also in case of DMR i remember strange patterns on asphalt for example, they are not colored, so its hard to notice them, and I think they are results of specific interpolation.

Link to post
Share on other sites

OK....for those of you who do subscribe to the site, please take a moment to read the revised section as background. Those who don't read the site will no doubt pick up the gist of what we're discussing as the thread progresses.

 

1. RAW - One can think of a "pure" RAW file as one that has had the very least possible processing prior to being written to the card. Various makers do different kinds of additional processing to the file before writing it to the card. The M8 applies vignetting correction (if one is using a 6-bit coded lens and the option to read that data is switched on). Canon clearly does some noise filtering before writing the RAW file. Etc...

 

Canon's high-ISO performance gains somewhat from this pre-RAW processing but I believe that the signal to noise ratio of high ISO files from the 5D is also quite good from the start. In other words, there's more to the 5D's high ISO abilities than just the pre-filtering of the RAW file.

 

I would not want to see Leica pre-filtering their high ISO RAW files because it costs detail (which we can see in the Canon files in the M8 vs. 5D high ISO crops). I would prefer to have the camera give me all of the data and let me decide what trade-offs I'm willing to make to reduce noise at the expense of detail. I think of the M8 high ISO files as being "honest" - one gets the detail and the noise and can decide how to proceed from there.

 

In fact, one could say that there seems to be a guiding philosophy in the M8 that the resolution provided by the lens not be intefered with at any point in the digital process. Hence, no artificial filtering via the AA filter and no (apparent) pre-RAW noise filtering. It is, in a sense, a "high fidelity" camera.

 

Leica did pretty well with high ISO performance through ISO 1600. The future direction, I think, would not be pre-RAW noise filtering but, instead, continuing to improve the S/N ratio of the sensor/engine itself. They're already way ahead of what they had from the DMR in this respect.

 

So there's Part One of this multi-topic thread.

 

Cheers,

 

Sean

Link to post
Share on other sites

Actually focussing on those comments about moire...

 

Sean, can the moire be seen on the in-camera display? My understanding is that moving your shooting position even slightly is enough to dampen the interference effect so that you have to work quite hard to get it to happen.

 

The DMR has an option to apply moire correction but I assume this is only for the JPEGs. Does it work? Looks like they have omitted this option from the M8. Do the JPEGs show the same moire?

 

Hi Mark,

 

I haven't checked yet but it probably can be seen on the display. I'll check when I get a chance. My own approach when I'm working usually doesn't involve the LCD beyond a quick check of the histogram.

 

An AA filter softens a files somewhat. Removing it allows the resolution of the lens to travel (more or less intact) all the way to the sensor. As I discussed in the article, removing the AA filter from the path means that there will be moire found in *some* files (it's rare in my work) but there is stronger resolution in *all* files. So all files gain something from this and only some of them show moire. The DMR owners are the ones to listen to about how this has worked out for them in practice but my sense is that it is not a frequent problem for most. Thinking about work I've done with the DMR and M8, it's rare that I see sufficient moire that it would need to be corrected. As such, I prefer the "no AA filter" approach.

 

Again, for those who read the site it would be worthwhile to check the new section before proceeding in the thread. That's not meant as a promotional comment, I just don't want to have re-explain everything I said already in the article.

 

I think that the successor to the M8 should continue this design path of having no AA filter. It's not just resolution that's gained, the M8 also shows subtle variations in color detail that are remarkable.

 

Cheers,

 

Sean

Link to post
Share on other sites

The DMR owners are the ones to listen to about how this has worked out for them in practice but my sense is that it is not a frequent problem for most.

 

I exclusively shoot weddings with two DMRs and in over a year with it, I have yet to see a single case of moire in my images...

Link to post
Share on other sites

I wouldn't rely on the LCD to evaluate moire as the LCD is going to contribute it's own effect from yet a different pixel density. (Aside: I tried a magnifier once on the digilux 2 LCD, thinking I could see deeper into that excellent display; the result ended up being the equivalent of looking through the EV!)

Link to post
Share on other sites

Guest guy_mancuso
I exclusively shoot weddings with two DMRs and in over a year with it, I have yet to see a single case of moire in my images...

 

 

Not surprised to hear that JR , really the only time I get and it is rare is studio lights and with certain cloths. Give me no AA filter any time of day over a camera with one. I want nothing between my glass and my sensor .

Link to post
Share on other sites

Guest stevenrk

I think that the successor to the M8 should continue this design path of having no AA filter. It's not just resolution that's gained, the M8 also shows subtle variations in color detail that are remarkable.

 

Cheers,

 

Sean

 

Sean, that's very interesting. I hadn't made the connection between your discussion of the subtlety of color variaition (as you point out, also very important for BW) and the AA filter. Are you saying that the quality of the color rendition in the M8 is because there is no AA filter?

 

Best, Steven

Link to post
Share on other sites

Sean, that's very interesting. I hadn't made the connection between your discussion of the subtlety of color variaition (as you point out, also very important for BW) and the AA filter. Are you saying that the quality of the color rendition in the M8 is because there is no AA filter?

 

Best, Steven

 

Hi Steven,

 

I didn't make the connection myself until this morning. I can't answer your question in any certain way but I believe the lack of AA filter is one of the reasons why we see such subtle variations in color detail in the M8 files. It's a working hypothesis for me right now.

 

Cheers,

 

Sean

Link to post
Share on other sites

Hi Steven,

 

I didn't make the connection myself until this morning. I can't answer your question in any certain way but I believe the lack of AA filter is one of the reasons why we see such subtle variations in color detail in the M8 files. It's a working hypothesis for me right now.

 

Cheers,

 

Sean

 

I was thinking the same thing when I first got my DMR (and still do) -- the extremely fine details and subtle differences in tones is the first thing that jumped to me. As a matter of fact, the details is so fine (in the skin tones for example) that it is very difficult to create a noise profile for the DMR for an image (not a target) as the noise reduction software can't differentiate between "noise" and "detail".

Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...