F200EXR versus F31fd -- battle of the giants

Ok ... lots of recent discussion on the Fuji Talk Forum over the detail retention at high ISO of the F200EXR against the F31fd. User PhotoFreak7 has tested the living sh*t out of these cameras and has rendered his opinions thoroughly and completely here. And guess what? The F31fd came out on top.

That started an interesting debate, and MrClick mentioned my name ... suggesting that I or some other noise reduction guru (not claiming the status, just the word he used) should run the test.

Well, seemed like a pretty good idea ... for a third time. You might want to read the first three blog entries I did on this topic here, here and here before you read this ....

So I thought perhaps I could perform a simple test using images that have a pretty good chance of being comparable. Luckily, a set of just such images exists at this review site. Thanks to user rax for linking to the translated site.

So the test goes like this:

1) Grab the 1600 ISO image from the F200EXR shot in SN mode and the F31fd. No need to bother with the F200EXR in any other mode, as the pixel binning mode is the only one that comes close to an F31fd anyway.

2) Load each in ACR5. Open shadows a bit, recover highlights a bit. Try to make them look the same.

3) Crop in three places for each and form a composite image in CS4.

4) Run Topaz Denoise on each image. Same settings for both.

5) Add same crops to composite image to show how each handles sophisticated noise reduction.

And that's it.

Here are the crops (remember to click to open to full size and then use the back button to continue reading):


So what do I see on the six crop pairs from top to bottom?

1) Grain: F31fd is smoothest
Detail: tie

2) Grain: F31fd
Detail: F200EXR by a hair

Third crop is in shadow.

3) Grain: F31fd
Detail: F31fd by a hair (two actually)

Next three have had a strong dose of Topaz Denoise.

4) Grain: tie (a few edge artifacts on smooth surface in F31fd image)
Detail: F31fd by a hair (per pixel sharpness)

5) Grain: tie
Detail: F200EXR by a hair (per pixel sharpness)

Final crop is in shadow.

6) Grain: F31fd (smoother shadow texture)
Detail: F31fd (most decisive difference of all)

The F31fd won 7 and tied 3 tests, losing only 2 of 12 tests ... I guess the answer is rather clear, isn't it?

Now remember that this test is as pixel peeped as you can imagine. Deep, deep into the pixels ... and these two are so close to a tie that worrying the differences seems almost silly. Either one will look great in prints or on the web.

But ... straight from the camera there is an obvious difference in grain. There is very significant grain in F200EXR images versus F31fd. And the F31fd grain cleans up better, especially in shadow. Again ... the F31fd seems to be a little better.

People have been commenting on Fuji being more restrained now and using less luminance NR and cleaning up color noise much better. Well, doesn't look like that to me. I believe that they were forced to turn down the NR because they tend to mangle small details when they use it full strength on the current crop of cams, even after pixel binning. In these images, there is no difference in color noise, despite the obvious darkness of the background.

So if you want straight from the cam perfection, the F31fd is still giving the best output. If you want to clean them up, it's a toss up ... unless you are a low-key kind of photographer, in which case I would choose the F31fd for better shadows even after NR.

But, of course, the F200EXR has the wonderful stabilized lens with better range and it takes SD cards. It's smaller, too. So there are excellent reasons for going that way. But then there's that crappy battery life. *Sigh* ... I've said this many times ... Fuji giveth and Fuji taketh away ...

Good luck on your choice everyone ... I'll stick with my F11 and G10 for now ...

Here are the two final images ... I've worked them over to look a lot like each other ... and frankly, they are near identical. But there's still visible grain in that F200EXR background ... how do you like them apples?



Your mileage may vary :-)

Rogers Data Plans Revisited -- Almost There

I was rather savage last year over the ridiculous cost of Data that characterized the avarice of Rogers and Apple when the initial iPhone gold rush was in full swing. People were complaining about the outrageous multi-thousand dollar bills they were receiving, and the cost of pay as you go data was higher than that of any known street drug ...

So a quick revisit of the data plans shows that they are getting there. The 6GB plan at $30 has long since disappeared ... now you get 1GB for that amount of cash. An increase of 6 times ... not tolerated in many industries, but barely noticed in the Canadian wireless market, known to be the biggest rip off in the western world.

But the real change is a maximum data fee of $100 per month. Data used to be to Canadians what medical bills are to Americans -- a quick way to eliminate your life savings. We solve the medical issue a long time ago (and I'm shocked that there are so many Americans who think it's a bad idea to solve it there) ... and now we've solved the data problem in Canada, which interestingly I think they solved in the USA quite a while ago ... data is pretty cheap there compared to here. Go figure ...

Anyway ... my only issue now is with the low end plan costs ... I think they need an entry plan that'll get you hooked. Something for occasional users. $25 for 500MB is simply too high. For an extra $5 you get 1GB ... and that's why Rogers does it. They don't expect anyone to pay $25 for half the data you can get for $30 because people don't want to look stupid. It's simple greed on Rogers' part to force people to spend the extra $5 per month.

What they should offer is a plan at $15 for 300MB ... a plan at $20 for 750MB and a plan at $30 for 1.5GB ... that way, people can start cheap (and stay cheap if it works out) and get more data as they go up, instead of this silly step function from $25 to $30.

Whatever ... I'll wait. Hasn't bothered me much ...

Canon G10 Shoots a Hockey Game

Sports is the domain of cameras with excellent high ISO and fast lenses. Thus, there is domination of sports these days by high end full frame dSLRs and huge stabilized lenses on monopods. You can see a sea of them at any major sporting event.

But us lowlifes in the general public who might want to pursue our hobby with much lesser equipment (for example, my D300 with 70-300VR) are simply shut out. Try and bring one into a venue like the Scotiabank Place in Ottawa, and you will run afoul of the "no interchangeable lenses" rule immediately. So I don't ...

Instead, I have always brought in my trusty Fuji F11, an excellent high ISO compact camera, one of the many that Fuji has created over the years. I have shot many concerts with it, and have always been pleased with its capabilities.

However, it is a very difficult camera to work with in real time. The menus are somewhat complex, and there are few external controls. Fuji always creates many limitations in their modes, you can do X in mode Y but not in mode Z, that kind of thing. This phenomenon has reached new heights with the new F200EXR, which is so complex that half the reviews so far have not managed to get decent sample images.

Anyway ... I chose instead to go with Canon's latest entry back in November. I shot a couple of concerts with that camera and, despite its weaker high ISO, I was able to get images with which I was pleased. Its video support is a huge step up from any Fuji camera, which also pleases me no end.

So when I got the chance to see a Toronto versus Ottawa hockey game from a corporate booth (won a lottery for a seat based on a donation to a local hospital foundation), I decided to bring the Canon along. I grabbed a corner seat closest to the home goal (2 out of 3 periods of course) and shot away all game. It was great fun.

I tried a few settings to start with, but quickly settled into a rhythm in manual mode at 400 ISO. I had my best luck at 1/320s at f4.5 (wide open) and 140mm. Since I was at the 100 level, I had a reasonable bit of reach. These settings are underexposed by about one full stop, but I felt that I would retain better detail at the lower ISO in RAW than at the higher ISO. Having shot some at 800 ISO as well, I believe I was right to strike that compromise.

My work flow for these images was rather complex, but then I expect to have to work for it when I shoot without a dSLR. Most of what gets posted to the forums straight from the cam these days looks technically really poor in my opinion. So here's what I did for these:

RAW file imported to ACR5, then:
1) Adjust white balance on the Toronto jerseys
2) Adjust exposure (most were underexposed)
3) Lower contrast
4) Drop sharpening to zero
5) Set luminance noise reduction to zero and color noise reduction to 100
6) Move image to CS4 in 16 bit mode and Adobe RGB

In CS4:
7) Run Neat Image to smooth grain
8) Run Topaz Denoise to enhance detail and eliminate remaining grain
9) Hue/Saturation layer to lower the yellow content on the ice
10) Mask off the yellow strip on the boards, and sometimes the crowd to retain the warmer colors where needed
11) Levels to brighten the ice to around 230
12) Curves to lower the reflections in the glass, masked
13) PKSharpener creative haze reduction to enhance detail
14) Resize to 800px
15) PKSharpener narrow web 800, white edges very low, dark edges very high, reduce opacity to taste

A fairly typical work flow for anyone who wants to get what they can from their camera. What is nice with the G10 is that it retains a surprising amount of detail at 400iso, so these steps actually work.

Here's an early shot of the Ottawa goalie (Elliot) making a save. Remember to click on the image to see the full size. Then come back here with your browser's back arrow.


Obviously, you can clearly see the cameraman behind the goal, the faces on the players and in the crowd, and lots of details on their equipment. So what's not to like? I'm quite happy.

The only shot I posted at 800iso is the aftermath of the only Toronto goal. I was distracted and spun around to shoot it, so there is softness due to movement along with the natural softness of the high ISO. Still, it is acceptable. Plenty of detail is still visible.


If you want to see how well this camera retains detail under these conditions, have a look at this crowd shot when the mascot made his appearance at the bench. The hair on these people clumps a bit, as is unavoidable, but there is still the feeling that you are looking at hair and not a helmet :-)


I was lucky enough to catch Alfredsson in the act of scoring the first goal. My shutter speed was slow enough at that time (1/200s) that the puck is simply a blur. But this one ends up in the net and 1 second later the audience is up and screaming.


Despite the fairly slow shutter, note the very fine detail in the audience's faces. This is what I noticed about the G10 when shooting the concerts in late 2008 ... and those were done in jpeg. Here, I find RAW really makes a difference, despite the excellence of Canon's jpeg engine.

Here's another save by Elliot, protecting what is now a 2-1 lead for Ottawa. On this one, you can see the puck as a blur just below his blocker (right hand.) I had raised the shutter speed to 1/250s and I think it makes for a slight improvement in acuity. Of course, the fact is that each shot is subject to the technique used and the slight statistical variance that stabilization will produce. So who can say why some images are extremely sharp while others are just slightly less sharp.


And finally, here's a somewhat cropped image of Daniel Alfredsson taking a spin on the ice after being named 1st star of the game. The detail and smoothness of this image is really something in my opinion. I am very pleased. I shot it at 1/320s, which I eventually found to be the best compromise between exposure and crispness.


If you want to see the whole series, go here.

Fuji F200EXR at 1600 ISO -- Revisited

There's been a bit more hyperbole surrounding this camera. What a shock :-)

The latest comment is that this camera can take images at 1600 ISO that "no compact camera has ever been able to take."

The example that was quoted to me is here. Now, it's a nice shot, but the rails are blown out completely, so clearly the EXR sensor was doing its high ISO trick and not its high dynamic range trick.

I was told that the faces were really good, which presumably means that all other compacts would be unable to duplicate the feat. So to be quite clear on what we are talking about, here are the faces isolated in a 100% crop. Remember to click on the image to see the larger version, then click your back button to return here.


So yes, the faces look decent. Especially when you consider that this image is shot at 1600 ISO.

Now, the first thing I thought about was that my F10 and F11 have shot a lot of concerts and night images ... so perhaps I could find similar images from them. Remember that the F10 was the breakthrough camera that started the whole Fuji low light revolution.

People poo-poo the earlier cameras these days because DPReview said that the F30 was a full stop better with noise than the F10. In fact, this is true when looking at out of camera output. But it is hardly true with competent processing.

For the purposes of this test though, very little processing has been applied to the images ... just enough to tease out some detail. *No* extra noise processing. At all ... on any of these images.

So let's have a look at the F10 ... granddaddy to them all.




Wow ... those 100% crops are pretty smooth for 1600 ISO, aren't they? I think a lot of people have forgotten how smooth and clean Fuji's 6mp sensors looked before Fuji cranked up the NR, contrast and sharpening to create rather water colory images that happen to print well straight from the cam. These are pretty darned good, and remember that there is *no* extra noise reduction here, other than the freebee you get from downsizing for the web.

So ... what about the F11, the one that shipped only to Europe and Asia? I got one on eBay and loved it. That is, until I got the G10 ... but let's wait a moment for that.




That de-icer driver was shot through an airliner window and of course through his window. That's a lot of detail at 1600 ISO through two thick glass elements. The Roger Waters image shows how detailed his hand is ... really amazing when you consider how far I was from the stage :-)

I think the F200EXR is going to struggle to outdo these older cameras. Frankly, these cams are *very* friendly to additional processing, so they can be brought up to a pretty decent standard.

Now ... let's move to the ridiculous. The G10 is not a 1600 ISO machine at all ... that is easily its worst ISO. And yet, when you shoot RAW and process carefully, it's not awful. Remember that you downsize a long way just to get to the EXR's 6mp high ISO images. So what does this all mean?

Well, here's an interesting image ... shot in jpeg only because it was my first night with the cam and I was unfamiliar with the menus and controls. I included two 100% crops here ... one of the original 15mp image, and one of a 6mp version of the image. The latter is the only fair comparison, since it is the same size (exactly) as the EXR and F11 images.


I actually find the detail to be pretty decent for 1.7 micron photosites shot at 1600 ISO. The web image looks pretty decent to me.

But I don't tend to shoot this cam at 1600 ISO. I prefer to shoot at 200, 400 or 800 and to underexpose a bit in RAW, then pull up in post. After doing this for a couple of concerts, I am pretty sure that it works better than shooting at 1600 all the time. The easy manual control of the G10 makes it trivial to experiment with the technique ... and I am pleased enough with the results.

Two more example ... 400 ISO then 200 ISO.



So there you have it ... I think this shows that there is plenty of detail available in current cameras in low light. I am sure that the F200EXR will be pretty good at it, providing some of the strange noise patterns we've seen do not get out of hand under certain conditions. But I don't consider head and shoulders above the rest of the Fxx series at high ISO.

In fact, I don't find its images particularly smooth ... Fuji could have done better with a simple 6mp sensor with current technology. But then they'd have no way to get the high dynamic range mode ... and there would be no way to advertise it as a 12mp camera.

The Apple iPhone after 90 days or thereabouts ...

Just get one ...

Is the F200EXR that much better than the F31fd?

Today, the Fuji Talk Forum on DPReview is a buzz with interest in this review. It's been pointed to before, but it seems that there might be some new comparisons on there. Specifically, the F200EXR is directly compared to the F31fd.

Why does this matter? Because there is a cult of people who believe that the F31fd was the culminations of Fujifilm's great Fxx series of cameras, the low-light masterpieces that were able to shoot decent small prints at 1600iso ... a first for a compact camera.

Personally, I like the F11 just fine ... it used a softer touch on the noise reduction and sharpening, and could therefore handle quite a bit of processing. I compare that camera to the Canon G10 (my new favorite) in this blog entry.

The buzz is made up of a lot of superlatives. The F200EXR is "a hundred times better" ... "VASTLY better" ... and so on ...

I compared the F200EXR to itself at 1600ISO previously in this blog entry ... and found that the new 6mp "binning" mode seems to work. You get a better result than with the 12mp mode, which means that this camera should get back the high ISO performance that was lost when jumping to 12mp. A bit ironic, though, that it was achieved by dropping back to 6mp in this "special" mode :-)

The real benefit though, is that this camera increases the effective pixel size while taking advantage of the really nice lens from the F100fd. This lens shows less tendency to CA and PF, and it seems sharper corner to corner than the F31fd's lens ever was.

So there should be an improvement, and indeed there is obvious improvement at 100%. That is, you can see less smearing from noise reduction and you can see less blur from the lens. You also see much less CA and PF.

But what does this mean to j6p? Remember that the average person does not print and tends these days to post to facebook or equivalent.

Once downsized, a lot of these differences simply disappear. And for those enthusiasts who process their images (I don't see any point in just taking what the camera gives you), the differences becomes very marginal.

Here is a pair of 800 ISO images (the highest useful ISO on almost any small sensor camera) where I downsized for the web after performing essentially identical processing. Basically, for each image, I:

• Loaded into CS4 directly as a copy / paste
  
• Removed distortions using PTLens
• Adjusted saturation and vibrancy (it was terrible on both cams)
• Adjusted contrast, opening the shadows a bit
• Adjusted local contrast with the DSLR Tools clarify filter
• Added more local contrast to the subject (middle forground) using PKSharpener's narrow edge local contrast filter. Also worked over the background branches and foliage on both images.
  
• Resized to 800px

I performed one special step to each of the images:

• For the F200EXR, I ran Neat Image 6 to slightly reduce the obvious grain. I added luminance sharpening to preserve detail. The end result was still more grainy than the F31fd.
  
• For the F31fd, I added a slight amount of output sharpening to equalize to that received by the 200 in its Neat Image step. Both images appear to me to have the same amount of sharpening now.

Now remember, I used the same amounts in each of the steps ... the two images were similar enough that no variance was needed to achieve the same effect.

So ... drum roll please. Which is which? In my opinion ... all these superlatives get washed away once properly processed.

Remember ... click to get the 800px version, and use the back button to come back.