The Reshade Algorithm

The main purpose of the algorithm is to upsize (increase the resolution of) images without deteriorating image details. Most of the existing upsizing algorithms either give a very blurry result or an artificial look, which is not desirable generally. Any zoom (upscaling size) is possible theoretically. The limitations for the online resizer are due to limited available processing time on the server. The following are a few of the factors considered when creating the reshade algorithm:

Maintaining edge clarity

The classic trade-offs when enlarging an image were either jagged, blurred or artificially smooth edges. Reshading succeeds in keeping sharp and naturally smooth edges, breaking the "tradition".

Creating texture

When enlarging an image, finer texture almost entirely disappears. Studies on perception show that the lack of lower order texturing has a degrading effect on the perceived image quality. Some solutions to this go as far as adding noise (film grain) intentionally to compensate for this reduction in detail. Of course noise doesn't really do much good, but it's helpful if no other alternative exists. Reshading creates small texture by finely distorting existing details to match the original image. This gives a better feel to the resulting textures in the enlarged image. They seem more natural at that scale and more pleasing to the eye.

Reducing noise

For most other algorithms a small dot in the original image may become a large speckle in the enlarged version. Not so with reshading. By selectively manipulating the details, noise is usually reduced to manageable amounts. There is also a denoise setting available so that you can remove more noise from the image.

Removing sharpening-halo effects

When sharpening with unsharp-mask like filters you create greater contrast near edges. This looks good in the original image: the eye perceives only accentuated edges and not small halos around them. When zooming in, this changes: now you can easily see the halos and this is not that pleasant. With reshading the impact of sharpening-halos on the enlarged image is reduced.

Reducing jpeg artifacts

The same thing happens with jpeg artifacs. When viewing jpegs at low compression these artifacts may not be noticeable. But when zooming in they become annoying. Reshade fixes this completely sometimes and in other cases reduces dramatically the degrading effect on image quality automatically. You can explicitly tell reshade to remove artifacts by using the smooth feature.

Enhancing blurry images

A very important aspect is keeping the focus of the image while resizing. Normal resizing methods like bicubic (from Photoshop) introduce a noticeable blur starting from very low zoom factors just about 120%. Reshade is able to maintain focus well over 400%. If the original image doesn't contain high detail textures, enlargements can reach even 2000% without a great deal of blur.