Resize Best Practices: Lanczos3 vs Bilinear
Understand the math and visual outcomes of popular image resampling algorithms.
MV
Marcus Vance
Contributing Author · Squoosh Next BlogWhen downscaling images, choosing the right interpolation algorithm determines whether output appears crisp or blurry. Bilinear interpolation computes each output pixel as a weighted average of the four nearest source pixels, producing smooth but sometimes soft results. Bicubic interpolation uses a 4×4 neighborhood with cubic weighting, sharpening edges at the cost of minor ringing artifacts.
Lanczos3 applies a windowed sinc filter over a 6×6 pixel neighborhood, preserving high-frequency detail most accurately of the three but requiring more computation. For thumbnail generation where speed matters, bilinear is acceptable. For hero image downscaling where quality is paramount, Lanczos3 is the professional standard.
Never upscale with nearest-neighbor for photographic content as it produces visible pixel blocks.
Key Takeaways
When downscaling images, choosing the right interpolation algorithm determines whether output appears crisp or blurry.
Bilinear interpolation computes each output pixel as a weighted average of the four nearest source pixels, producing smooth but sometimes soft results.
Bicubic interpolation uses a 4×4 neighborhood with cubic weighting, sharpening edges at the cost of minor ringing artifacts.
Lanczos3 applies a windowed sinc filter over a 6×6 pixel neighborhood, preserving high-frequency detail most accurately of the three but requiring more computation.
Try It in the Workspace
Everything discussed in this article can be tested directly in Squoosh Next — no sign-up, no upload, 100% client-side.