G6 and G12: RAW Noise and Reduction

.

Based on the in-camera JPEGs, it seems to me that the G12 is about two stops better than the G6 with respect to noise. But then I began to wonder if this apparent difference was actually due to noise reduction software in the G12. To test this hypothesis, I took another look at those images but this time from the RAW image files.

The above 100% crops (click to see full size) were processed from the RAW files using Adobe Camera Raw. The G12 profile for ACR is still a “beta” but is probably representative. Processing for both the G6 and G12 was as follows:

... as shot auto white balance
... auto toning
... medium contrast curve
... ACR sharpening at 80%, 1 radius, 30 detail, 60 mask
... 0 noise reduction in ACR
... default noise processing in Noiseware Pro.

This time I skipped the G12 ISO 800 and used ISO 1600 and 3200. The G6 ISO 400 image cleaned up nicely but the G12 still has less noise – I think about two stops improvement but perhaps only 1-1/2 stops.

.

High Dynamic Pumpkin

.

I like High Dynamic Range (HDR) photography although I must admit that my results are still unpredictable. A new book, “Complete Guide to High Dynamic Range Digital Photography” by Ferrell McCollough, is helping me to become more consistent in my HDR processing.

This HDR image was made using four shots from my Canon G9. The G9 was in Av mode, ISO 400, manual focus. ISO 400 helped to avoid some very long exposures and also to dodge the Av “bug” that limits shutter speeds to 1 second. The four shots were combined in Photomatix and also tone mapped in Photomatix before final touchups, including noise reduction with Noiseware, in Photoshop CS4.

As an 12x18 inch on canvas, I like it a lot. A screen sized version is in my G9 gallery on SmugMug.

Why Av mode? A neat trick is to use Av mode and also set auto exposure bracketing. This produces three shots. Next, dial in some negative exposure compensation and fire again to get three more shots. Finally, dial in some positive exposure compensation and get three more shots. Depending on the settings for exposure bracketing and exposure compensation, some of the shots may be at the same exposure but you’ll only touch the camera twice during the process.

.

G9: Sky at ISO 800

.

.
In a previous post, I noted that I was placing some large images in a Smugmug gallery for purposes of comparison. Those first images were taken at ISO 100, 400 and 800. Both the in-camera JPEG and a raw image file were posted. A subtle difference for those first six shots of the sky is that all were slightly underexposed. The last two shots (#7 and #8) were made “just right” by increasing the exposure by +2/3 stop as compared to the evaluative metering mode of the G9. Now to examine an ISO 800 “just right” image and compare the noise to the slightly underexposed variation.

The image above is the ISO 800 in-camera JPEG that was underexposed by 1 stop in the bracketing process. This means it was actually underexposed by 1-2/3 stops in comparison to the "correct" exposure.

The images to be compared are the new Smugmug images #9, 10, 11, 12. These are the bracketed versions (+/- 1 stop) of images #5 and 6. Image #9 and 10 are the in-camera JPEGs and images #11 and 12 are the processed raw versions of #9 and 10 respectively. In particular, notice that image #12 has much less noise that image #11 and even less noise than image #6. The apparent reduction in noise is actually because much less noise is generated when the exposure is correct. In fact, this was the entire purpose of this little exercise: Minimize noise by avoiding underexposure.

.

G9: Large, comparative samples

.
There are many and fierce debates about the relative merits and faults of cameras and processing. Although the debate rages on, I've almost given up because the basis of the debates is so subjective. On the other hand, there is a real difficulty in making comparisons. What is the final output of the "photograph"? A snapshot sized print? An enlargement -- 8x10? Larger? Will the photo only be seen from a computer screen? How large is that screen? Is the screen calibrated for proper color, brightness and contrast? No wonder so many people have such diverging views -- not to mention justification of their expense in obtaining that camera!

For the G9 and many other small sensor digicams, one of the major issues is noise at high ISO. The G9 applies its own noise reduction algorithm to its in-camera JPEG but other noise reduction software is readily available. In particular, some G9'ers, myself included, usually shoot in raw mode and apply noise reduction afterward. Which method is best? Does it matter? Can a G9 high ISO be saved?

To systematically investigate the various high ISO and processing options, I thought to post some full size images for comparison. These will be posted on a special SmugMug gallery. The first set of images is a simple set of shots at the sky. These images were taken at ISO 100, 400 and 800. The G9 was set for raw+jpeg so the images are in pairs. When the G9 is set for raw+jpeg, the jpeg is full size but a little more compression is used. I know this confuses the issue somewhat but I made additional comparisons (unposted) and the differences between the G9 "fine" and "superfine" images are small. As it is, this set of comparisons includes eight images and there's no real point to adding more.

My intent was to simply point the G9 at the sky and fire away on Av mode. Then I decided to throw in exposure bracketing so I added a polarizing filter to keep exposures within range. Even so, the G9 neutral density filter was necessary for the ISO 400 and 800 shots. So the first two shots (ISO 100) are without the ND filter, the ISO 400 and 800 are with the ND filter. All of these shots are slightly underexposed.

The exposure was increased by +2/3 stop for a final set of shots. Of course, I forgot to remove the ND filter for these so the exposure is much longer than expected.

The in-camera jpegs were all done at "normal" settings. The raw files were converted in ACR and followed with noise reduction using Noiseware. Noiseware was used at its default settings. The raw files were all done semi-automatically using the ACR auto toning (exposure) and a Photoshop Action.

In comparing these samples, the general idea is not so much which exposure, contrast, saturation, etc is preferable to you personally as it is to see that there are, or can be, differences. Don't even bother to tell me that you can process the in-camera jpeg to have the same saturation and contrast as the processed raw image. Also, don't bother to tell me that the in-camera settings can be adjusted for more or less saturation, sharpness and contrast. But do notice that the first three conversions from raw have quite different look from the first three jpegs. This difference is largely due to the initial underexposure. These are comparisons of the in-camera auto processing of underexposure to the ACR auto processing of underexposure.

In the last two images, the G9 exposure setting was 'corrected' by adding 2/3 stop compensation. The in-camera jpeg and the auto process raw are much more similar.

Examine the files, pixel peek, print your own, order from SmugMug -- whatever you prefer. There's something or other to be learned here and it is different for all of us.
.

G9 and ISO 400

.
Much is made of the fact that the G9 is a bit noisy and that noise increases at high ISO settings. In general, I agree. But I disagree with the oft-stated wisdom that the G9 is horrible or, at the very least, unusable above ISO 200. Although I’ll grant that the condemnation or acceptance of noise is largely a matter of taste, I find the G9 to be very usable at ISO 400 – especially if shot in raw mode, properly exposed and processed with noise reduction software. This shot of balloons is from the G9 in raw mode, ISO 400, 1/60 second, f5 using bounce flash. It was converted in Adobe Camera Raw, slightly cropped and tweaked in Photoshop CS3. Noise reduction was accomplished with Noiseware immediately after conversion from raw mode. The Noiseware setting was simply the default.

In my opinion, the G9 produces reasonable 8x10 prints at ISO 400 when shot in raw mode, converted in ACR and treated with Noiseware. In this digital age, many people do not print their photos and, if printed, most photos are printed as snapshots. I’ve posted a 1024x768 pixel version (full screen for many) of the balloons on my Smugmug gallery.

Of course, your mileage – and tastes -- may vary and I have no idea of the noise characteristics of the new G10. But if you have a G9, don’t be afraid to use ISO 400. At the very least, try ISO 400; you may be pleasantly surprised. (In fact, sometimes I use ISO 800 but that's another story ...)
.

Watching the Storm

.
Here’s my favorite photo from Hurricane Gustav. My granddaughter wasn’t really watching the hurricane but the wind and rain continued for a while.

This shot was taken with my G9 in raw mode, ISO 400, max zoom, f4.8, 1/60 second. The small Flexizone frame was used to focus. My granddaughter is quite active and quick but was captivated by the scene and stayed at the door long enough for me to check exposure in Av mode, see that the histogram wasn’t right, shift to manual exposure and get a good histogram. One shot and she was gone!

Processing in ACR from the raw file, I decided to let the back of her head and dress go a bit dark. At ISO 400, noise begins to be noticeable with the G9, especially for enlargements, so I used Noiseware at about half strength to reduce the noise.

The final 8x10 print was cropped only slightly from the original capture. I like it.

.