Enlarge Images Without Pixelation

Why?

Usually, we want to reduce the size of images, not enlarge them. However, there are times where we have a reasonably small source image that we need to make bigger for one reason or another.

I actually routinely enlarge images because I have a variety of cameras that I create images with and they’re all different resolutions and bit depths, so I normalize the keepers to one larger master resolution and bit depth.

It should be noted that my method described below is not a miracle worker. You won’t be making 10x or 20x enlargements that look good, but you can easily make 3-4x enlargements that will look totally passable.

The best part? You don’t have to pay for any software beyond just having Adobe Photoshop, so no plug-ins or extra software to buy (Perfect Resize, I’m looking right at you).

How?

The how is actually pretty simple, and before I explain the specifics of what I do in my version of Photoshop, I’ll explain the generic version so that you can convert it to your software, which may not be the same thing as what I’m running. Ok? Let’s get started.

Before doing this, start with your source image and save it as a 16 bit tiff file at its original resolution.

  1. Figure out what your resulting resolution is to be and multiply the longest edge of the image by four. For example, I generally normalize to 7200×4800 pixels, so I’d end up with 28,800×19,200 pixels. Resize your source image to that new, really huge resolution.

  2. Add some noise to the newly resized really huge image. The better way to do this is with monochromatic noise so it looks more organic, but if your software doesn’t do that, any noise is better than no noise. I generally add between 5% and 10%. It should be Gaussian noise.

  3. Add a Gaussian Blur to the really huge image. It’s radius should be 2 pixels.

  4. Resize the really huge image down to your resulting image size. Save it as a 16 bit tiff file and do the rest of whatever post processing you’re going to do.

That’s it! This method also works like a charm for upgrading 8 bit images to nice smooth 16 bit images that you can really push around in Lightroom/Photoshop without any ugly banding or posterization popping out at you.

Why it works

What?!?! You’re adding noise and blur to the image! Doesn’t that destroy image quality and detail and make the image blurry?

If we were to do that to an image at its native size, then yes, all we would be doing is adding noise and making it blurry. The thing to keep in mind is that we’re doing this to an image that is 8 times the resolution of our final output resolution.

When we scale the image back down, that noise and blur has the effect of filling in and smoothing out what would otherwise be pixelation in the image.

The key to this working so well is the bit depth and ratios that we use. The small amount of noise that gets added when we’re at 8x resolution gets reduced down and visually provides a smoothing effect to the new image size without making it look soft or blurry.

Likewise, the Gaussian blur we added was two pixels to an image that is 8x the resolution of what it ultimately will be, meaning when we scale the image back down, a 16×16 block of pixels gets turned into a 4×4 block of pixels, which means that we’re scaling down more than we blurred. When we scale down more than we blur, the blur that was applied starts to do interesting things for us. It visually provides the effect of filling in and smoothing out what would otherwise be pixelation in the image.

Combined with the added noise in step 2, it’s a very dramatic one-two punch to an image that would otherwise look pixelated and awful.

How I do it

We all use different software. I happen to use the latest version Adobe Lightroom CC and Adobe Photoshop CC. This isn’t a tutorial for how to use Lightroom or Photoshop, it’s just a basic walk-through of what I do. You can and should modify it to suit your needs.

All of my images start off in Lightroom at their original resolution and bit depth. I have a Lightroom catalog that I use for staging these images for processing that all of my keepers make their first stop in. In this catalog, I add all my metadata to each image (it makes tracking it later easier) and the only image adjustment I make here is to remove image noise that was introduced by the camera in the form of color and/or luminance noise. I’m very conservative with this and look at the image at a 1:1 or 2:1 ratio in the area where noise is most prominent in the image. I only do just barely enough noise removal to tone down the noise since the more noise removal you do, the more the image fine detail gets muddled. This is done on an image by image basis and how much noise reduction is applied varies greatly between what camera took the image and at what ISO the image was shot at.

From there, I export the image out as a 16 bit tiff file at the “super-sized” resolution (28,800×19,200 pixel or there abouts depending on the image aspect ratio).

I then open that tiff file in Photoshop and add a layer over the background layer that is the image. I change the layer’s blending mode to “overlay” and fill it with 50% gray. From there I convert the layer to a smart object.

With the smart object selected, I go to the ‘Filter’ menu, and select ‘Noise’->’Add Noise’. In the dialog box that pops up, select ‘Gaussian’, and check the ‘Monochromatic’ check-box. Change the amount to a value between 5% and 10%. I’ve found that less than 5% tends to be not enough, and more than 10% is too much. I generally set it to 7.5% as a start then tweak it up or down as needed for best results. You should experiment around for what values work best for you based on what resolution you are working at.

From there, I go to the ‘Filter’ menu again and select ‘Blur’->’Gaussian Blur’. In the pop up dialog box I select a radius of 2 pixels.

From there, tweak the amount of noise up or down for best results (you can do this because it’s a smart object).

When you’re happy with the image, go to the ‘Layer’ menu and select ‘Flatten Image’.

Now resize the image to your final enlarged size (in my case 7200×4800 pixels) and save it as a 16 bit tiff that you’ll pull into your real Lightroom catalog.

Import the new enlarged tiff file into your Lightroom catalog that you use for managing your media, convert it to a DNG file, then finish the rest of your post processing on the file.

Isn’t this a lot of work?

Yes and no. We only do this on our keeper images, which for most photographers is only a fraction of the images that they take. That and the only thing we’re doing that is any different is the Photoshop bit. You should still be doing noise reduction, adding meta-data, and post processing. The only real difference is a quick middle step in Photoshop that literally only takes a couple minutes per image, if that.

Everybody should do what works for them, and what works for me isn’t necessarily for everybody, however, the process I outlined above allows me to shoot everything from an iPhone jpeg, to a DSLR raw or jpeg image, to frame grabs from video as jpegs, and end up with a reasonably sized standard image resolution that is actually very usable.

As proof in the pudding, most all of the images I’ve recently shot digitally have had this treatment, and if I hadn’t told you that those images got this treatment, while looking at them, you’d be none the wiser. It makes differentiating between jpeg and raw, and lower iPhone/video resolution and DSLR resolution extremely difficult, which is the point.

Caveat Emptor

This works best with an image that already has reasonably good resolution content to begin with. This does not magically add detail where there is none, nor does it really add resolution or rescue images that already look terrible.

What it does do is add very high frequency or broadband information to an image in a way that our brains find very pleasant, which allows our brains to do the heavy lifting of ‘seeing’ what detail is there in the enlarged image without seeing the unpleasant visual effects of scaling up that image. In short, we’re playing a very effective visual trick on our brains in very much the same way that adding dithering to digital audio allows us to hear further down into the sound.

Our brains are very good at filtering out high frequency broadband noise to get to the detail in the noise, as long as that noise isn’t overwhelming to the point of being distracting. The trick is riding that balance between helping the perceived image quality and hurting it.

Till next time.

Author: Adrian Bacon

Photographer. Videographer. Coder.