Resize Pictures

Resize Picture at its best!

Scaling Methods

Nearest Neighbor
The new pixel is assigned the color value of the closest neighbor. This means that no new color values are generated. The method works well for discrete data such as line drawings, but not very well with 'normal' photos. It is very fast. Inclined edges easily take on a 'staircase' look.

Linear
The color value(s) of the inserted pixel(s) is/are calculated according to a linear function between the two original pixels which have been separated. Very fast, likelihood of aliasing and can lead to prominent color values. The image appears smoother than with the Nearest Neighbour method.

Bilinear
Also called quadratic interpolation. With this method, four neighbouring pixels are used to determined the color values of the new pixel(s). Equivalent to Linear in both X and Y directions, i.e. mean value of four neighboring pixels. Effect: softening; i.e. less sharp and lessens aliasing. Good for pixel magnification.

Box
= Nearest Neighbor (!) - generates mean color value from pixels within a frame and assigns this to new pixel. Effect: strong softening, almost no aliasing, fast. Good for magnification but bad for image reduction.

Triangle
Medium quality and fairly fast. Usually leaves rough edges. Bad for image reduction.

Bell
Weights color values of nearer pixels more strongly (bell curve) when determining new pixel. Quite fast, typical softening filter, not so suitable for resizing.

B-Spline (Bicubic)
PhotoPerfect uses bicubic interpolation, not 'cubic spline interpolation'. It uses 16 neighboring pixels to determine the color value(s) of the new pixel(s). Effect: softer transitions and much less loss of sharpness than with Bilinear, practically no aliasing. Even fine structures are preserved as it even has a slightly sharpening effect. Slow method making high demands on resources. Suitable for image reduction as well. (The classical B-Spline is a powerful 'softener' and causes a high degree of fuzziness.)

Lanczos3
One of the conjugated gradient methods, a bi-orthogonalizing algorithm. Produces good sharpness, minimal softening and slightly rough edges. Similarly good results as with fractal methods, very slow. Good for magnification. If the image bokeh displays doughnuts effects, additional sharpening may help.

Mitchell
Produces good sharpness and fine details. The softening is similar to that of the Triangle method. Not very good in respect of rough edges, slow, reduces image size, good for magnification. Represents a compromise between Lanczos3, Box, Triangle and Bell. Here too, subsequent sharpening can help.

PhotoPerfect ClearType
Optimizes the display of downscaled images on LCD monitors. Has no effect on the display of the image on cathode ray monitors or printed images.


 Learn more...