| Introduction
Image resizing (magnification or reduction) is a standard operation in image processing.
The main purpose of this demo is to show that we can improve the results obtained by the standard resizing method (interpolation and resampling) in terms of visual quality; our method reduces the artifacts as it approximates the continuous model by its projection onto a given space prior to resampling (which is a generalized version of anti-aliasing prefiltering). We have implemented an algorithm that allows to compute both oblique and orthogonal projections (least-squares approximations) for splines of any degree n.
The demonstration will allow you to test the improvements by yourself. To demonstrate that the loss of information is less important for our method than in the standard case, we propose to perform image reduction, which may create artifacts due to aliasing. Note that you have the possibility to use either the standard method (as many programs of image processing do), or to obtain better results applying the projection method.
- Original image.
- Resized image (by a factor inferior to 100%).
- The previous image is enlarged to obtain an image of the same size as the original one.
- Difference between the original and the enlarged images.
|
