and were made much more apparent in the Super Resolution version, probably because the algorithm thought it was real detail. The artifacts are there in the original, very subtly. Here we see artifacts enhanced by ACR's Super Resolution. It does not have to connect to the net to work. What's interesting is that somehow it is using what the machine has learned in an algorithm that runs locally on your desktop computer. I'll be doing some more tests and posting the results here. The interesting question to me: Is the algorithm sharpening detail that is present, like in a planetary image, or is it creating new detail from what it expects to be there? Probably not smart enough to know much more than "push here" and it's supposed to make it better. Ok, at this point, I'm starting to talk out of my hoo-ha. But looking at other daytime examples, I see all point sources of light reduced in size, so that must be part of algorithm to increased the resolution. I wonder if this learning model has examined millions of high-resolution astrophotos, learned to recognize the field, and then put it in at a resolution twice as good as what you start with. What I can say is that there is a lot of fine detail that can be brought out in planetary photography that is really there, but you can't see in the original. But I'd guess that a "convolutional" neural network is not the same thing. The one word I have heard of is "convolution." As in deconvolution in astrophotography. There is some sophisticated math going on here.
![open jpeg in camera raw open jpeg in camera raw](https://images.squarespace-cdn.com/content/v1/5487c777e4b01bcc8b835e5e/1473877357554-K1K17HV38TG99N7IADT8/image-asset.jpeg)
Sensei uses all forms of artificial intelligence using machine "deep learning."Īll very interesting. He said the algorithm used a "deep convolutional neural network, a fancy way of saying that what happens to a pixel depends on the pixels immediately around it."Īll of this is built on Adobe Sensei. It was the core technology behind Super Resolution.Įnhance details used "an advanced machine learning model trained on millions of photos" as Eric Chen described Super Resolution in a post on the Adobe Blog. Artificial IntelligenceĪ couple of years ago, Adobe's Michaël Gharbi and Richard Zhang developed Enhance Details. The D5300 is known for producing green stars, and Super Resolution made them much worse.Ĭomparison images show here have been enlarged another 200 percent to more clearly show the differences. In this case, ACR's Super Resolution did not do such a good job. This shows the results of Super Resolution with a Nikon D5300 with smaller pixels, and the Sigma 105mm f/1.4. To open JPEGs or TIFFs, select Edit > Preferences > Camera Raw and in the File Handling tab, set JPEG or TIFF to Automatically open all supported JPEGs or TIFFs. Raw files open in the Camera Raw dialog box by default. You have to open the image by selecting File > Open As and selecting Camera Raw from the drop-down menu. Note, you can't access this Enhance command for Super Resolution if the file is already open in Photoshop and you select Filter > Camera Raw Filter. Then right click on the image and select Enhance > Super Resolution, or use the keyboard shortcut CTRL SHIFT D. You access Super Resolution in the ACR window when you open a raw file from Bridge into Adobe Camera Raw. The tests were done on Nikon Z6 NEF raw files that were opened in Adobe Camera Raw (ACR) with the adjustments below. It did not work well with that same lens on my Nikon D5300 which has smaller pixels and a notorious green pixel problem. It worked well with a single frame from my Nikon Z6 shot with my Sigma Art 105mm at f/1.4. I did some quick tests with some long-exposure deep-sky astronomical images and found that it works. The latest version of Photoshop 21.2.6 offers a new "Super Resolution" image re-sampling method that will double the pixel size of your image using artificial intelligence (AI) with machine learning in Adobe's Camera Raw module. The second image has been resized to 200 percent using ACR Super Resolution.Ĭomparison images shown here have been enlarged another 200 percent to more clearly show the differences. The original image has been resized in Photoshop with Bilinear interpolation to 200 percent of its original size. Click or tap again to go back to the previous image.
![open jpeg in camera raw open jpeg in camera raw](https://camerajabber.com/wp-content/uploads/2017/12/photoshop_cc_2018_review-05.jpg)
Photoshop's Super Resolution does a pretty remarkable job of improving the apparent resolution of this image, a single 30-second exposure of a star field taken with a Nikon Z6 and Sigma Art 105mm at f/1.4.Ĭlick your mouse cursor in the image, or tap the image if you have a touchscreen, to see the comparison.