How to Resize an Image Without Losing Visible Quality

Resize images with fewer visible artifacts by starting from the best source, preserving aspect ratio, avoiding unnecessary upscaling and re-encoding only once.

On this page15 sections
  1. Why resizing can make an image look worse
  2. Downscaling is different from quality loss caused by compression
  3. Start from the highest-quality source you have
  4. Resize from the destination backward
  5. Preserve the aspect ratio
  6. Avoid unnecessary upscaling
  7. Use one controlled resize-and-encode pass
  8. How to judge visible quality correctly
  9. Photographs, screenshots, and graphics need different care
  10. Do higher-density displays require larger images?
  11. Worked example
  12. A quality-preserving resize checklist
  13. Resize an image in the browser
  14. Frequently asked questions
  15. Sources and further reading

“Resize an image without losing quality” is a useful goal, but it needs one correction: when you reduce the pixel dimensions of a raster image, information is discarded. The real objective is to resize without creating unnecessary visible degradation at the size where the image will actually be used.

A 6000×4000 photograph reduced to 1200×800 no longer contains the same number of pixels. That is unavoidable. What you can control is whether the downscaled result still looks clean, whether fine edges survive well, whether the aspect ratio remains correct, and whether the new file is encoded without avoidable artifacts.

Short version: start from the best original, resize to a dimension tied to the real display slot, preserve aspect ratio, avoid enlarging small sources, and encode the final delivery image once. Judge the result at its intended display size—not at 400% zoom.

Why resizing can make an image look worse

Raster images are grids of pixels. Changing that grid requires resampling: software estimates what the new pixel values should be. Downscaling combines information from many source pixels into fewer output pixels. Upscaling has the opposite problem: it must create additional pixel values that were not captured by the original camera or artwork.

The result can look worse for several different reasons:

  • the output is enlarged beyond the useful detail in the source;
  • the aspect ratio is changed accidentally, stretching the image;
  • the file is repeatedly resized and lossy-compressed across several generations;
  • the output quality setting is too aggressive;
  • the image is displayed larger than its new intrinsic dimensions;
  • fine textures, text, thin lines, or high-contrast edges reveal resampling and compression artifacts more readily than broad photographic areas.

Downscaling is different from quality loss caused by compression

Downscaling removes pixels. Lossy compression approximates image information to encode it more efficiently. Those processes can happen together, but they are not the same.

If you take a 4000×3000 source, resize it to 1200×900, and save it at a reasonable quality setting, most viewers may perceive the delivery version as excellent at 600 or 900 CSS pixels wide. If you keep the original dimensions but push compression too far, the image may show blocking, ringing, smearing, or banding even though the pixel count did not change.

That is why the correct workflow combines this guide with image compression quality settings rather than treating the quality slider as a resizing control.

Start from the highest-quality source you have

Every derivative should ideally be generated from a master or the best available original. Avoid taking a small web JPEG, enlarging it, compressing it, then later using that derivative as the source for another export.

Lossy generations compound. Fine texture and edge detail that disappeared in the first export cannot be restored by the next resize. Keeping a master also gives you flexibility when a future layout needs a different crop or larger source width.

Resize from the destination backward

Do not ask “what is the best image width?” before knowing where the image will appear. Measure the destination.

If the largest article column is 900 CSS pixels wide, you might create a source around 900 pixels for standard-density use and a larger candidate around 1800 pixels for higher-density displays. If the same image appears as a 300-pixel card elsewhere, responsive markup can let the browser choose a smaller candidate.

The key is that the image dimensions should be connected to the rendered slot. Our complete image resizing guide explains intrinsic pixels, CSS pixels and device pixel ratio in detail.

Preserve the aspect ratio

If an image is 3:2, proportional resizing should keep it 3:2. A 3000×2000 source becomes 1500×1000, 1200×800, or 900×600. It should not become 1200×700 unless you intentionally crop or distort it.

On the web, CSS can preserve an image’s intrinsic ratio when one dimension is automatic. MDN also documents object-fit: cover and contain for fitting images into boxes without stretching them. If your intended layout uses a different ratio, cropping is usually the cleaner solution; see Crop vs Resize.

Avoid unnecessary upscaling

Upscaling a 600-pixel source to 1800 pixels does not reveal detail that was never present. Traditional interpolation can make the file larger and the edges smoother, but it cannot recreate original optical information.

AI upscaling can synthesize plausible texture and edge detail, which can be useful in restoration workflows, but it is not equivalent to recovering the original capture. For normal web publishing, the safer rule is simple: choose a source that already contains enough pixels for the intended display size.

Use one controlled resize-and-encode pass

When possible, resize and encode the final delivery file from the original in one operation. Repeatedly opening a lossy JPEG, resizing it a little, exporting again, then repeating that process can compound artifacts.

This is one reason our browser tool combines dimensions, format, and quality in the same workflow. You can choose the target size and output format before creating the new file rather than building a chain of derivatives.

How to judge visible quality correctly

Inspect at two scales:

  1. Intended display size. This is the primary judgment. Does the image look crisp enough in the real layout?
  2. A modest detail crop. Check high-risk areas such as hair, foliage, product texture, small lettering, diagonal edges, and smooth gradients.

Do not make the decision only at extreme zoom. At 400% every resampled raster will reveal its pixel structure. Users experience the image at the rendered size.

Photographs, screenshots, and graphics need different care

Photographs usually tolerate downscaling well because natural detail blends across pixels. Screenshots containing text and UI elements can expose softness immediately. Diagrams, logos, and illustrations with hard edges may be better delivered as SVG when the source is vector, or at carefully chosen raster dimensions when SVG is not appropriate.

If the image contains text that must remain readable, check it at the actual CSS size. A large screenshot that looks perfect in the source can become illegible after aggressive downscaling even when the resize itself is technically clean.

Do higher-density displays require larger images?

Often, yes—but that does not mean every image should be exported at twice the layout width and served to everyone. MDN explains that browsers can combine srcset, expected slot size, and device pixel ratio when selecting an image candidate.

For a 600 CSS-pixel slot, a higher-density screen may prefer a source closer to 1200 pixels. A standard-density screen may not need it. Responsive images let the browser make that choice.

Worked example

Imagine a 5000×3333 product photograph used in a content area that reaches 800 CSS pixels on desktop and 360 CSS pixels on mobile.

A sensible workflow is to keep the 5000-pixel master, create perhaps 480w, 960w, and 1600w delivery candidates from that master, then encode them in an appropriate web format. The 1600-pixel version provides headroom for a high-density 800 CSS-pixel display without forcing a phone to download that candidate.

Compare the 960- and 1600-pixel outputs at the real rendered sizes. If fine product texture is important, favor the candidate and quality setting that preserves it. If the image is a decorative background, the quality budget may be lower.

A quality-preserving resize checklist

  1. Keep the original master.
  2. Measure the real maximum rendered size.
  3. Preserve aspect ratio unless you intentionally crop.
  4. Avoid enlarging a source that is already too small.
  5. Create only the responsive widths the layout can use.
  6. Resize and encode from the master rather than from a previous derivative.
  7. Use an output format appropriate to the content.
  8. Inspect the result at intended display size and in a difficult detail area.
  9. Verify which source the browser actually downloads.

Resize an image in the browser

Open the image tool, choose a source, expand Advanced options, and set the target width or height. Keep the ratio proportional, choose the output format and quality, then inspect the result before downloading.

Frequently asked questions

Does reducing image dimensions always reduce quality?

It removes pixel information, but the visible result can remain excellent when the output still contains enough detail for its display size. “Quality” should be judged against the intended use, not only against the source at full zoom.

Why does my resized image look blurry?

Common causes are upscaling a small source, displaying the output larger than its intrinsic dimensions, using an overly aggressive compression setting, or resizing a text-heavy screenshot too far.

Should I resize before compressing?

Think of them as one delivery workflow. Decide the needed dimensions first, then encode the resized output at an appropriate quality. Starting with dimensions prevents you from spending bytes on pixels the layout cannot use.

Is a 2× image always necessary for Retina displays?

No. It can be useful, but responsive source selection is better than a universal 2× rule. Provide sensible candidates and let the browser choose based on slot size and display density.

Sources and further reading

Continue the series8 parts

Image Resizing & Dimensions

A practical series on choosing, resizing, cropping, and delivering the right image dimensions for the web.

  1. Image Resizing: The Complete Guide to Dimensions, Resolution and Aspect Ratio
  2. How to Resize an Image Without Losing Visible Quality
  3. Crop vs Resize: What’s the Difference?
  4. Aspect Ratio Explained: How to Resize Images Without Stretching Them
  5. Best Image Size for a Website: Hero, Banner, Blog and Content Images
  6. Best Hero Image Size for Desktop and Mobile
  7. Pixels vs Resolution vs File Size: What Actually Matters on the Web?
  8. Does DPI Matter for Web Images? The 72 DPI Myth Explained