Wednesday, February 20, 2019

The One Thing That Sometimes Bothered Me Using APS-C And M4/3 Was Digital Noise; That Issue May Seemingly Be No More


Full image of a statue from Chinese Terracotta Soldiers Exhibit; ISO 3200
Olympus E-M1 Mark II, 12-40mm f/2.8 lens @ 31mm; 1/25th sec. @ f/8; ISO 3200
I’ve been an M4/3 format camera user since the introduction of Olympus’ E-M5 in February 2012.  I've been an APS-C format user since Fujifilm introduced the X-T1 in January 2014.  With both systems, I found myself very pleased with my images and frankly amazed with the image quality from small sensored cameras.  That being said, there was one thing that, on occasion, occurred that still somewhat bothered me.  That was digital noise when I rarely had to use a high ISO.

 Since I first adopted these two camera systems and with the introduction of newer sensors and better in-camera processing algorithms, the quality today is even better.  Still, once in a while, I would find a high ISO file that I wasn't able to satisfactorily remove noise and keep detail.  That frustrated me.  No matter what noise reducing function or program I used, I never really was satisfied with the results.  One has to keep in mind that noise was not really an issue when looking at the entire image on my monitor nor in prints, but when looking at an image at 100%, it was visible and bothersome.  I know that doesn’t make sense and I guess it is more of a psychological thing with me than practical.  That is what I get for pixel peeping, I guess.  I now believe that final issue has been largely solved for me.  That solution comes in the form of Topaz Labs recent introduction of a new noise reducing program based upon the incorporation of artificial intelligence or A.I.  It is called AI Clear.

AI Clear has impressed me.  It is not perfect and I suspect there will be continued research and subsequent updates but it is the best I've used so far.  It is simple to use and isn't cost prohibitive.  Does it do everything perfectly?  Not quite, but as I said, it is the best program at noise reduction I've yet tried.  When I say best I've seen, I refer to the trade-off between noise reduction and keeping fine detail.  There is always a trade-off between those two parameters when trying to eliminate digital noise.

 This is not a review of the program.  I just wanted to pass on my initial thoughts on my use of the program.  If you have interest or want to know more about it, then I suggest you visit the Topaz Labs website.
Detail from above; Image on left is from original RAW file; Image on the right is after noise reduction using Topaz
AI Clear using the full automatic setting; Notice the background areas and amount of noise. Also, the detail in the figurine has been retained.
The next image was pretty much a handheld "grab shot" at ISO 3200 at 840mm (FF equiv.) of a hooded merganser.  I spotted this duck while driving, quickly pulled over, rolled down my window and as I did, it started rapidly paddling away to my right.  I quickly snapped two frames of it before it was hidden behind some grass.  The full image, below, has quite a bit of noise when closely examined.  The final composition, which would be heavily cropped, shows some of that noise.  AI Clear did a pretty good job of reducing the noise but keeping the details.  I used the "Automatic" setting on the program for this image.

Hooded Merganser.  This is the overall image shot at ISO 3200.  The noise apparent in the
original image is annoying to me. (click to enlarge)
Fujifilm X-T3; 100-400mm f/4.5-5.6 lens + 1.4x tele-converter @ 560mm (840mm FF FOV); 1/500th sec. @ f/8; Handheld

100% crop from original RAF file, no sharpening nor noise reduction applied in Lightroom Classic CC (click to enlarge)

Image above after processing in Topaz AI Clear using the "Automatic" setting (click to enlarge)
Below is an another example of an Olympus E-M1 Mark II image made at ISO 1000.  The composite below shows left and right 100% crops of an image of snow geese.  It illustrates the differences that can easily be seen in the before image (left) and the AI Clear processed image (right).  Again, the Automatic setting was used.
This is a composite of the same image.  The left side shows the original noise in the image when looking at a 100% crop.
The right side shows how Topaz AI Clear eliminated the noise but kept virtually all the detail. (click to enlarge)
These are just a few of the images I've tried while trying to get to know AI Clear.  I'll be continuing to use it and to better understand all of its nuances and learn how to optimize its settings before pronouncing a final verdict.  

 The program has several controls including an automatic function which works quite well.  It has low/medium/high reduction, a low/high sharpening function, a masking function, clarity slider, details recovery slider, opacity slider as well as other controls.  In other words, you can tailor your noise reducing needs very specifically to each image, or part of an image, if necessary.

 I'm still experimenting with this program.  From my initial results, I am quite pleased with the noise I can eliminate but keep detail.  As I wrote, it is not perfect but it seems to be the best I've found to date.  This post is just a couple of examples of my use of the program on my images to share with you.  YMMV.  If my experiments with the program turn out how I expect, I think my issue with noise in smaller sensored cameras will finally be solved.  Hopefully, anyway.  Lol.

 I've been a fan of Topaz products for a long time and use their plug-ins.  I am not sponsored by Topaz Labs nor have I ever receive any money or anything free or specially discounted from them.  I purchased all of my Topaz Products at the same cost as everyone else.

 Join me over at Instagram @dennisamook or my website, www.dennismook.com. 

 Thanks for looking. Enjoy! 

 Dennis A. Mook 

 All content on this blog is © 2013-2019 Dennis A. Mook. All Rights Reserved. Feel free to point to this blog from your website with full attribution. Permission may be granted for commercial use. Please contact Mr. Mook to discuss permission to reproduce the blog posts and/or images.
at

6 comments:

  1. Jim RestleFebruary 21, 2019 at 4:11 PM

    Thanks for showing this Dennis. As you know, Olympus said there is A.I. involved in the focusing system of their new body, and I wonder if A.I will eventually solve all of the complaints about m4/3: noise at high ISO, shallow depth of field, and maybe even dynamic range. Interesting times indeed.

    ReplyDelete
    Replies
    1. The Wandering LensmanFebruary 21, 2019 at 5:15 PM

      Jim, thanks for your comment. I agree with your premise. I’m not sure how much real improvement can come from just sensor developments, but I think more and more improvements will come from computational features. A good example, as you said, is the focusing in the ne E-M1X. I would bet Olympus will add birds and other objects to their focus tracking algorithms. Several years ago, Fujifilm developed a dual sensor. I don’t remember what it’s name was, but the sensor was bifurcated into pixels that were low ISO and pixels that were high ISO, for lack of a better way to state it. The combination captured bright and low light on the same sensor. I think it was ahead of its time but that type of thing will be computational in the future. It’s all good and getting better each year.

      Delete
  2. DoylFebruary 23, 2019 at 1:17 AM

    Děkuji za pěkný článek. Zkusím tento program na odstranění šumu a vylepšení mých fotografií vyfocenych s M1mkII.

    ReplyDelete
    Replies
    1. The Wandering LensmanFebruary 23, 2019 at 3:22 PM

      Google translation from Czech: Thank you for a nice article. I'll try this program to remove the noise and improve my photos taken with M1mkII.

      Doyl, thank you for your comment.

      Delete
  3. sdufordFebruary 23, 2019 at 6:43 AM

    Thanks for the article. I recently wrote a blog post where I tried some of the top image processing apps on a noisy poorly exposed 6400 ISO image and then a properly exposed 12,800 ISO image from my old E-M1. While Topaz AI Clear is very good, I found that it leaves some colour artifacts, especially in metallic or silver-colored parts of images. According to my tests, DXO Photolab Elite with its Prime noise reduction algorithm, was by far the best, virtually eliminated noise while preserving colour and detail. Like Topaz it's a one-button operation and takes some time to process. You can download a free trial and give it a go. Here's a link to my article: https://www.duford.com/2019/02/a-completely-unscientific-test-of-noise-reduction-software/

    ReplyDelete
    Replies
    1. The Wandering LensmanFebruary 23, 2019 at 3:27 PM

      sduford, thanks for the comment. I think it is important when we can consider multiple opinions and points of view. I downloaded and tried the DXO Prime noise reduction for 30 days last summer. I wrote about it in my July 18, 2018 blog post. At that time, I wrote that it was the best noise reduction program I had yet tried. It is excellent. However, with the images I used for testing, I find the Topaz program gives me a bit better results with a bit better control and fewer artifacts. It may all depend upon which images are used in the processing for noise. No doubt they are both excellent. Again, thank you for your comment.

      Delete