Every time you resize an image in Photoshop, the software has to decide what to do with the pixels. Make a photo bigger and it must invent pixels that were never captured. Make it smaller and it must decide which pixels to throw away. The rule Photoshop uses to make those decisions is called image interpolation (also known as resampling), and the method you pick has a visible effect on sharpness, noise, halos and detail.
The options have changed a lot in recent years. Preserve Details 2.0, Camera Raw’s Super Resolution and Generative Upscale now sit alongside the classic Bicubic, Bilinear and Nearest Neighbor methods. This guide explains how each one works, when to use it, and where to find the settings.
What is Image Interpolation in Photoshop?
In simpler terms, image interpolation in Photoshop is the process used by Photoshop to determine how new pixels are created or existing ones are adjusted when you resize an image.
Imagine a beautiful mosaic – resizing it without interpolation would simply mean adding or removing tiles, resulting in a distorted and blurry image. Interpolation, on the other hand, acts like a skilled craftsperson, meticulously analysing the surrounding tiles and creating new ones that seamlessly blend in, preserving the overall integrity of the mosaic (your image).
Quick Answer: Which Resample Method Should You Use?
| What you’re doing | Best choice | Why |
| Enlarging a photo | Preserve Details 2.0 | Keeps detail while controlling noise and artifacts |
| Shrinking a photo | Bicubic Sharper (or Automatic) | Keeps edges crisp after pixels are removed |
| Enlarging pixel art, icons or hard-edged graphics | Nearest Neighbor | Keeps edges hard, with no blur |
| Everyday resizing when you’re unsure | Automatic | Photoshop picks a method based on the job |
| Enlarging a portrait or landscape and you want AI to rebuild detail | Generative Upscale | Reconstructs fine detail, but can change it |
| Enlarging a graphic with text or line art | Preserve Details 2.0 | Scales existing pixels instead of generating new ones |

What Is Image Interpolation in Photoshop?
Image interpolation is the calculation Photoshop uses to work out the colour of a pixel that doesn’t exist in the original image. It looks at the pixels around the new position and estimates a value from them.
Think of it as filling in the gaps in a connect-the-dots drawing. If you stretch the drawing to twice the size, there are suddenly gaps between every dot. Different interpolation methods fill those gaps in different ways: some copy the nearest dot, some blend between neighbouring dots, and some use trained AI models to guess what the missing detail should look like.
One important point: interpolation can only estimate, never recover. No method can restore detail that wasn’t captured by the camera. It can only produce a smoother, sharper or more convincing enlargement. That’s why a clean, high-resolution original will always beat an upscaled small file.
When Does Interpolation Actually Happen?
Interpolation only happens when Photoshop changes the number of pixels. This is the most common source of confusion, so it’s worth a moment.
In Image > Image Size, the Resample checkbox controls everything:
- Resample ON: you can change the pixel dimensions. Photoshop adds or removes pixels and uses the interpolation method you choose.
- Resample OFF: the pixel dimensions are locked. Changing the width, height or resolution only changes how large the image will print, not how many pixels it contains. No interpolation takes place and image quality is unaffected.
For example, a 6000 × 4000 px photo prints at 20 × 13.3 inches at 300 ppi. Turn Resample off, change the resolution to 150 ppi, and the print size doubles to 40 × 26.7 inches with exactly the same pixels. Turn Resample on and ask for 12000 × 8000 px, and Photoshop has to invent 72 million new pixels. That’s when your interpolation choice matters.

Interpolation also happens whenever you scale a layer with Free Transform or the Transform tools, and whenever you rotate, skew or warp pixels. That’s why scaling a pixel layer up and down repeatedly softens it. (There’s a way around that with Smart Objects, covered below.)
Photoshop’s Resample Methods Explained
Here are the methods you’ll find in the Image Size dialog, from simplest to most advanced. [Confirm the exact list and wording in your current Photoshop version.]
Automatic
Photoshop chooses a method for you based on the document type and whether you’re scaling up or down. It’s a sensible default and a good choice if you’re unsure. If you care about the result, such as a big print or a client delivery, choose the method yourself.
Nearest Neighbor (Hard Edges)
The simplest and fastest method. Each new pixel simply copies the closest original pixel. There’s no blending, so edges stay perfectly hard.
- Best for: pixel art, icons, screenshots, QR codes and graphics with hard edges where you want to avoid smoothing.
- Avoid for: photos and anything with curves or diagonal lines, which will look jagged or blocky.
Bilinear
Bilinear works out each new pixel by averaging the four nearest original pixels (a 2 × 2 block). The result is smoother than Nearest Neighbor, but softer and less accurate than the bicubic methods.
- Best for: quick resizing where speed matters more than quality.
- Avoid for: final-quality photo enlargements.
Bicubic (Smooth Gradients)
Bicubic samples a larger neighbourhood (16 pixels, a 4 × 4 block) and weights them to calculate each new pixel. It gives smoother gradients and more accurate tones than Bilinear. Standard Bicubic is a good all-rounder, particularly for images with smooth tonal transitions.
Bicubic Smoother (Enlargement)
A bicubic variant tuned for enlarging. It smooths the result to hide the harshness that enlargement can create. The trade-off is that it can soften fine detail. It was long the go-to choice for photo enlargement, but Preserve Details 2.0 is now generally the better option for photos.
Bicubic Sharper (Reduction)
A bicubic variant designed for reducing image size. It adds sharpening during the resize so downsized images don’t look soft. If you shrink photos for the web, email or social media, this is usually the one to choose. Watch for over-sharpened edges or halos on high-contrast areas, and switch to standard Bicubic if you see them.
Preserve Details (Enlargement) – Legacy
The original Preserve Details algorithm introduced with Photoshop CC. It includes a Reduce Noise slider and was a big step up from Bicubic Smoother for enlargement. It has been largely replaced by Preserve Details 2.0.
Preserve Details 2.0
The most advanced of the traditional resample methods, and the one to reach for when enlarging photos. It uses a trained AI model to identify important detail and texture, and keeps them sharp without exaggerating noise or over-sharpening everything else. It includes a Noise Reduction slider.
Crucially, Preserve Details 2.0 scales your existing pixels rather than generating new content, so it won’t invent features or rewrite text.

How to Use Preserve Details 2.0 (Step by Step)
- Open your image and go to Image > Image Size.
- Make sure the chain link between Width and Height is on so the proportions stay locked.
- Tick Resample.
- From the Resample dropdown, choose Preserve Details 2.0.
- Enter the new width or height. [Check the dialog in your version and note whether you also need to switch units to Pixels.]
- Adjust the Noise Reduction slider. A low-to-moderate setting (around 10–20%) is a sensible starting point for clean files; increase it for grainy or heavily compressed images. Watch out for too much, as it smooths away fine texture.
- Zoom to 100% in the preview and check detail areas such as eyes, hair, foliage and text.
- Click OK.
If you don’t see Preserve Details 2.0 in the dropdown, open Edit > Preferences > Technology Previews (on Mac: Photoshop > Settings > Technology Previews) and check that Enable Preserve Details 2.0 Upscale is ticked, then restart Photoshop. [Verify the menu path and default state in your version; on recent builds it is on by default.]

Beyond Interpolation: AI Upscaling Options
Traditional interpolation methods work by estimating pixels from their neighbours. Newer tools go further and generate new detail from a trained AI model. Photoshop now gives you three AI-based routes, and they behave differently.
| Method | Where to find it | Scale control | How it works | Best for |
| Preserve Details 2.0 | Image > Image Size | Any custom size | AI-guided resampling of your existing pixels | Photos, and anything with text or line art |
| Super Resolution | Camera Raw > Enhance | 2× per side (4× the pixels) | Machine-learning enhancement of raw and standard files | Clean photos that need a quick, modest boost |
| Generative Upscale | Image > Generative Upscale | 2×, 3× or 4× | Generative AI predicts and reconstructs fine detail | Low-resolution photos where you want maximum perceived detail |
Generative Upscale: Impressive, But Check the Result
Generative Upscale can produce striking results on faces and photographic textures, because it’s predicting what detail should be there. That’s also its risk. Because it reconstructs detail rather than just scaling it, it can alter fine features. Adobe’s own guidance on a reported issue with embedded text and technical drawings is to use Preserve Details 2.0 for those images, since it scales existing pixels without generative reconstruction.
A practical rule: use Preserve Details 2.0 when accuracy matters (product photos, logos, documents, screenshots, anything with text or line art) and try Generative Upscale when you simply need a low-resolution photo to look as good as possible, then check the result at 100% for artifacts or changed details.

Super Resolution in Camera Raw
Super Resolution doubles the width and height of an image (four times the pixel count) using machine learning. In Camera Raw, open the image, right-click and choose Enhance (or press Ctrl/Cmd + Shift + D), tick Super Resolution and click Enhance. It saves the result as a new DNG file, which is handy when you want a clean 2× boost with minimal fuss. [Confirm the steps in your Camera Raw version.]
Shrinking Images: Which Method Is Best?
Downsizing throws pixels away, so the risks are different: lost detail, jagged edges and moiré patterns. For most photos, Bicubic Sharper (Reduction) or Automatic gives a crisp, clean result. If edges look over-sharpened, try standard Bicubic. For hard-edged graphics such as screenshots or pixel art, Nearest Neighbor keeps lines sharp but can look uneven at non-integer scales.
For best results when downsizing for the web, resize once from the original rather than repeatedly shrinking an already-resized copy, and apply any final output sharpening after resizing.
How to Change the Default Interpolation Method
The method you choose in the Image Size dialog applies to that resize only. Separately, Photoshop has a default interpolation setting that controls what happens when you scale pixel layers with Free Transform and other transform tools.
- Go to Edit > Preferences > General (Windows) or Photoshop > Settings > General (Mac).
- Find the Image Interpolation dropdown.
- Choose your preferred method (for example Bicubic Smoother for mostly enlarging, or Bicubic Sharper for mostly reducing) and click OK. [The Preferences list is shorter than the Image Size list; note which methods your version offers.]

You can also override it on the fly: when you use Free Transform (Ctrl/Cmd + T), the options bar at the top includes an interpolation dropdown, so you can choose a method for that transform alone.

Pro Tip: Use Smart Objects to Avoid Repeated Quality Loss
Every time you scale a regular pixel layer, Photoshop resamples it, and the quality can degrade a little more each time. Convert the layer to a Smart Object first (right-click the layer > Convert to Smart Object). A Smart Object keeps the original pixel data intact, so you can scale it up and down as often as you like and Photoshop always re-renders from the original instead of from the previous result. Note that scaling a Smart Object above its original size still means interpolation is happening, so the same limits apply.
Common Mistakes to Avoid
- Enlarging more than you need. Every extra percent asks Photoshop to invent more detail. [Add your own tested guidance, for example how far Preserve Details 2.0 holds up on a typical photo before artifacts appear.]
- Resizing the same file repeatedly. Always resize from the original, or use a Smart Object.
- Leaving Resample ticked when you only want to change print size. If you only need a different print size or ppi value, untick Resample so no pixels are added or removed.
- Using AI upscaling on text, logos and technical drawings without checking the result. Generative methods can change fine detail.
- Over-sharpening a downsized image. Bicubic Sharper already sharpens, so extra sharpening can create halos.
- Judging at fit-on-screen zoom. Differences between methods only show at 100% or higher.
Frequently Asked Questions
What is the best interpolation method in Photoshop?
There isn’t one best method. For enlarging photos, Preserve Details 2.0 is generally the strongest built-in choice. For reducing photos, Bicubic Sharper is usually best. For pixel art and hard-edged graphics, use Nearest Neighbor. If you’re unsure, Automatic is a sensible default.
What is the difference between bilinear and bicubic interpolation?
Bilinear averages the four nearest pixels, while bicubic weights a larger group of sixteen. Bicubic produces smoother gradients and more accurate detail, and it takes slightly longer.
Is Preserve Details 2.0 AI?
It uses a trained machine-learning model to guide how pixels are enlarged, but it scales your existing pixels rather than generating new content the way Generative Upscale does.
Does interpolation reduce image quality?
Resizing always involves a trade-off. Enlarging can’t add real detail, and shrinking discards it. A good method minimises visible quality loss, but it can’t restore detail that was never captured.
Should I use Generative Upscale or Preserve Details 2.0?
Use Preserve Details 2.0 when you need accuracy, such as with text, logos or product shots. Try Generative Upscale for low-resolution photos when you want the best-looking detail and are happy to check the result for changes.
Why does my image look blurry after resizing?
The usual causes are enlarging too far, using a smoothing method like Bicubic Smoother on an image that needs sharper edges, or resizing an already-resized file. Try Preserve Details 2.0 with a small amount of noise reduction, or resize again from the original.
Further Reading
- Resize Images Without Losing Quality in Photoshop
- AI Upscaling vs. Bicubic Interpolation: Is Resizing Dead in the Water?
- Transforming Objects in Photoshop (Adobe)
- Resampling options in Photoshop (Adobe)
Final Thoughts
Image interpolation is one of those behind-the-scenes settings that quietly shapes every resize you do. The short version: use Preserve Details 2.0 for enlarging photos, Bicubic Sharper for shrinking them, Nearest Neighbor for hard-edged graphics, and treat Generative Upscale as a powerful tool that deserves a close look at 100% before you trust it. Test on your own images, resize from the original, and you’ll get clean results every time.
