Actual imaging output

wildlifephoto · January 17, 2023, 12:41am

I have attached two of my copyright JPEG image files. The first is from the body (Nikon Z9, Sigma 60-600 Sport, manual exposure, AF, VR, auto ISO). The second is a crop of the

bird, PL6E complete DeepPRIME XD with my custom preset followed by a bit of slider adjustment.

The interesting item is the amount of “dark recovery” as well as detail rendition that the current DxO product could produce. Although I would not submit the image to a client, it is suitable for species identification given the additional information as to the geographical location and date. Note that the image was taken after hiking upon muddy trails under less than ideal conditions. There was no time to fully clean the lens front element protector (filter, claimed by the manufacturer to be contamination resistant) as is visible in the open sky. From the same imaging session (the more common USA term, “shoot”, currently is not considered as acceptable due to the use of anti-personnel weapons in mass shootings in the USA), a number of PL6E output images have been accepted by a client.

danielfrimley · January 17, 2023, 6:06am

I do that, a lot talking of sensors, @wildlifephoto how much of the recovery you were able to achieve would you put down to the quality/ performance of the Z9 sensor versus the superpowers of PL?

wildlifephoto · January 17, 2023, 8:00am

Unless the same subject/conditions produced an image on different bodies of the same intrinsic sensor pixel count and format (e.g., 40+ Mpix, 35 mm format) and lens, I have no idea whether it is the Z9 sensor or the capabilities of PL6E to fully use the information from that sensor; only a controlled comparison could settle that. Empirically, the combination produced a useful image.

Stenis · January 18, 2023, 11:58pm

@wildlifephoto
The sky is really not clean at all. That surprises me if you have used DP XD. The thing that use to make images look like this is too much Microcontrast. Thry to reduce that by pulling it almost as far as it gets and meet up with Fine Contrast instead. If you have used used ClearView Plus turn of that too.

wildlifephoto · January 19, 2023, 12:26am

Thank you for the observation. This image was not for submission to clients, merely to illustrate that what would have been a difficult “burn” in printing from a film image (print or chrome) is well handled by PL6E and that the subject (“small and dark” in the original image) can be made quite clear (sufficient for species identification). Both the contrast sliders and Clear View Plus are part of my custom preset for this type of subject, as these do an excellent job on feathers, fur, scales, and flower parts. I was going to use Topaz Gigapixel on the output image from PL6E to possibly make this a “usable” image, but Topaz clearly states that Gigapixel typically should not be used on fur or feathers (too many artefacts as well as other issues). I use a slightly different custom preset for scenics, although the de-haze capabilities of PL6E typically are used. Note that a telephoto lens looking at a distant subject often has “haze” in the image due to the air characteristics between the subject and the lens, and just a bit of “haze” considerably can reduce the detail in feathers, eye rings, fur, nose leather, and the like.

Stenis · January 19, 2023, 12:29am

Z9 has a really good sensor but it is not the lowlight champion of all FF-cameras. I use Sony and it is not one of the new expensive A7r V, A1 och A7s-models that performs the best when it comes to lowlight but the four-year-old A7 III with a modest 24 MP sensor. Despite the last one is the second best of all FF tested by DXO it´s far from super if you look at the images below. The best as you see is the new Canon R3.

The noise handling in the cameras with no exception is a compromise. Cameras are optimized for speed and power economy and that shows when comparing images processed a fraction of a second in a camera with images processed in powerful computers with fast graphics cards and SSD-discs. Despite I have that it use to take my computer 7-8 seconds per image to process and it you have an old one it can often take 20-30 seconds with Deep Prime XD.

Score:
Hasselblad MF 4489
Pentax 4505
Canon R3 4086
Sony A7 III 3722

In that discipline the Z9 2451 is a little below

Uncropped Original

Click to enlarge:
As you can see the sky in the left image is far from clean despite taken with one of the best low light FF sensors there is today (A7 III is second only to Canon R3). It wasn´t really even dark.

Hard cropped detail
Image to the left processed with HQ in PL6
Image to the right processed with Deep Prime XD.

So there is really no competition at all when you compare the image quality right out of the camera and one processed by Deep Prime XD. Please note! Even the left image is processed - with Photolab HQ. It looks even wore right out of the camera!

Deep Prime is just fantastic!

uncoy · April 4, 2023, 9:57am

It doesn’t make much sense to compare the low light performance of 21 to 26 MP cameras with 45 MP to 55 MP cameras. When one downsizes the larger images, the definition goes up and the effective low light performance is usually higher. That said, it’s a lot of pixels to push around for roughly the same performance at high ISO.

Much of the magic of high ISO is from DxO PhotoLab, which has been removing two stops of noise from high ISO Canon images since Prime (lots of chroma noise). Latest DeepPrime XD seems to do the same work for Nikon high resolution bodies (45MP) at very high ISO (12800 and above).

One can even see in this image (no noise reduction) the grid pattern of the D850 sensor (to see the grid pattern, it’s best to download the original and look at the whole image).

With DeepPrime XD, the noise is gone but the image remains detailed and colourful.

For reference, the settings I use for publication are very light (I used heavier settings for a recent test).

In the case of high resolution bodies, DeepPrime and DeepPrime XD give new value for the extra pixels at high ISO. Normally at high ISO it’s hard to both bring down the noise and extract the detail from high resolution bodies. DeepPrime and DeepPrime XD do that.

The autofocus in low light is significantly better on a D5 or D6 so there’s still a place for sports bodies. But in terms of the final image at high ISO, PhotoLab is a huge leveller. I chose the D850 for this shoot as the game started in daylight at ISO 200.

It’s a good feeling to be confident enough in the image quality at high ISO of the high MP bodies to be able to shoot a single body for a mixed light event.

No, I did not notice any image improvement between the D850 and Z9. ~~In effective terms, they are the same sensor.~~ Just checked Photons to Photons. There is no real difference between the D850, Z7 and Z7 II 45 MP sensor, but there are differences between those three and the Z9.

Sensor differences between Z9 and previous Nikon 45 MP bodies in real world photography is a subject to explore in future. For sports photography I didn’t noice any real difference but that’s probably because most of the time for high shutter speed I’m at ISO 1600 and up.