State-of-the-art?

This image is not for artistic critique but please comment on the technical approach which is described in full detail.
This Merlin was flying around our local wetlands. Of course, it would not come close enough to get a reasonable shot or was it? Certainly, there is nothing like a full frame image with a bird filling at least 50% of the frame. This is a processing exercise. I’m not trying to show off my skills; I’m just trying to present what is currently possible and I am no expert.

In the last few years, Topaz labs has been fine tuning their products and a lot of us are using them with good results. Perhaps, the noise reduction AI and then sharpening AI get the most accolades from our members. I don’t disagree. (Topaz didn’t pay me to say this; in fact I had to pay them to renew my upgrades for a year.)

This image starts out at 9504 x 6336 pixels which is the size of the full frame Sony’s A7r4 sensor. This was taken at 800 mm equivalent so the bird is quite far away.


And what I envision is approximately 2000 x 1400 pixels crop. The original image non cropped is 61 megapixels. Cropped as above is 2.8 megapixels or 4 1/2% of full frame. That is a fairly large crop. On top of that the exposure is not great; some of the whites are blown and the shadows are really too dark to see a lot of detail.
First step was to open the image and make initial corrections with DxO photo lab 4, which was just released last week. They have a new AI based algorithm on noise reduction called Deep Prime. The next step was to use Topaz Gigapixel AI which interpolates the image and increases its size between 0.5 and six times. This is a 2.0 increase.
I then ran that image through Topaz De noise AI using the AI clear setting. And then I ran the whole image through Topaz AI sharpen. Finally, I did some cloning to remove a few light areas that were blown out and I did some local Photoshop sharpening with opacity set to 50%.
Then the final image was converted to JPEG format at 72 dots per inch using the TK web sharpening algorithm, set to 1%. Final image size is a 2000 x 1280 pixel JPEG. This is just over 4% of the original full frame.
My own critique: I’m still seeing some noise. Some areas are over sharpened and others could be sharper. Yes this is a long process and some of the Topaz programs take a very long time to process because of the size of the original image. Gigapixel process takes about 10 minutes or longer and converts the original image to 19008 x 12672 which I think is 240 megapixels. Running it through the Topaz de noise program adds another 10 minutes. Sharpening, once the image is reduced in size to approximately 4.5 percent is a lot faster.

Please feel free to message me if you have any questions.

What artistic feedback would you like if any?

Pertinent technical details or techniques: iso-160, 400 mm plus 2X extender, F5 .6, 2500th, handheld, A7r4

(If backgrounds have been removed, etc. please be honest with your techniques to help others learn)

If you would like your image to be eligible for a feature on the NPN Instagram (@NaturePhotoNet), add the tag ‘ig’ and leave your Instagram username below.

Certainly the processing has brought a tiny bird in a frame to a much larger bird in the frame. Exposure looks good and photo is sharp. It seems the high contrast areas on the bird have been well preserved. These areas would be the wing and tail. The head area does not seem to do as well.

I think you’ve done an impressive job of enlarging the bird without losing detail and even maintaining (and improving) on the colors, particularly the tail feathers. One difference I notice is the color of the sky, which seems more natural in the original. I think what you’ve done works well, but seems quite time consuming, but probably worth it for selected images. I think this an American Kestrel.

Hi David
The end product looks vary nice. I my self have gone through this process. I started with DxO PhotoLab 3.2 (4 is a better product). From DxO I saved the file as a 8 bit tiff and then bumped it up to 1.5 and 2 times in Topaz Gigapixel. In both cases I did most of the cleaning up in DxO before importing it into Topaz. The Gigapixel output was a jpeg. As you can see this all works and the 2000 by1280 looks good.
Next question, what would it look like in Sony’s A7r4, C form witch is 6240 by 4160. That is larger than my 7Dll frame of 5472 by 3648. Plus you bump the 400mm to 600mm without the extender?
Nice work David.
Peter

Thanks for everybody’s comments. I thought it was a Kestrel but another photographer standing close by insisted it was Merlin. There are some similarities but the colors look more Kestrel like to me unless it’s my processing. Yes the processing is time consuming. I generally don’t think that birds against blue sky are that artistic. Yes they have their place. In this exercise, I think it had its place although something like Kestrel environment and the background would be considerably better.

I’m going to try this to compare to my simpler method, but I’m not sure of one thing – it sounds like you’re doing Gigapixel on the entire original frame. Wouldn’t it work to crop first to save time?