Inovix Image SDK High Precision Re-size - Report

A user may both want to inspect an image at coarse scale or to study some detail at fine scale. To this end, interpolation operations like zooming-in and -out are useful. The range of applications goes from fingerprint, medical imaging, satellite imaging and actually never stops... Interpolation is almost never a goal in itself, yet it affects both the desired results and the ways to obtain them.

Existing software and programming language modules include such image re-sizing functions, unfortunately the end result after re-sizing is either distorted or most of the details get blurred out in the process.

To illustrate the performance of the image re-sizing routines inside the Inovix Image SDK we took this standard image

and increased it's size by 25% to 640x640 pixels then decreased it size back to the original 512x512 pixels in order to overlap with the original image. The absolute difference in pixel intensities is computed between the original and restored image and displayed as a difference image. The darker difference image indicates the best re-sizing algorithm. We display the image after rotation to the left and difference image to the right

The results for Inovix Image SDK resizing algorithm :

The results with C-Sharp:

The results with IrfanView:

The results with Java:

The results with Perl:

The results with Python:

Inovix Image SDK offers the best performance with sub-pixel precision as can be seen by examining the resized and difference images.