ResizeX

ResizeX is a Fast Image Resizer which can resize images with free size values or with fixed static values whereas the algorithms with static resizing values are faster than the general resizing algorithms. Some of the available algorithms, as for the ScaleXx, are adding also additional pixel information into the image.

Scale2x/Scale3x Examples

Original size
Double size (Scale2x)
Triple size (Scale3x)

     

Available algorithm groups

Algorithm	Description
ResizeX	Very fast picture resize with given factor by just cloning the pixels
ScaleX	Fast quality picture resize with given factor and a ScaleX algorithm adding details
FastResize	Resizes picture faster but probably with less details as the default .net Bitmap class scaler
ResizeNET	Resizes picture with the default .net Bitmap class scaler (Reference)
SharpResize	Resizes picture with more sharpness but slower as the default .net Bitmap class scaler

Information about the Scale2X/Scale3X algorithm can be found at the Scale2x homepage.

Available algorithms

For every algorithm group are 8 algorithms available which scale to 200%, 300%, 400%, 600%, 800%, 900%, 1200% or 1600%.

Remark: some image/scaler combinations may produce errors when images get to big in size/memory consumption!

Library

All the image resizing functions are bundled in the Feuster.Imaging.Resizing.Lib

The library and a nuget package are available in the library release bundle.

Performance Test

The following tables show an example performance comparision of all scaler algorithms on a Intel(R) Core(TM) i5-6500 CPU @ 3.20GHz whereas ResizeXXX is the default .net image scaler. The performance test measures the pure conversion duration within a algorithm group without the time for file handling as loading or saving the image, so the results are independent of the type of storage where the image I/O is done. The performance test is integrated in ResizeX, see commandline reference for details.

Reference Image

The following image is used as reference source for the performance test and can be found in the Examples folder of the project.

Performance Results

Resizing the reference picture to 200% size:

Resize2x	Scale2x	FastResize200	Resize200	SharpResize200
60ms	73ms	140ms	143ms	241ms

Resizing the reference picture to 300% size:

Resize3x	Scale3x	FastResize300	Resize300	SharpResize300
52ms	255ms	290ms	360ms	918ms

Resizing the reference picture to 400% size:

Resize4x	Scale4x	FastResize400	Resize400	SharpResize400
149ms	325ms	584ms	600ms	1172ms

Resizing the reference picture to 600% size:

Resize6x	Scale6x	FastResize600	Resize600	SharpResize600
260ms	852ms	1265ms	1357ms	3806ms

Resizing the reference picture to 800% size:

Resize8x	Scale8x	FastResize800	Resize800	SharpResize800
587ms	1272ms	2230ms	2477ms	4813ms

Resizing the reference picture to 900% size:

Resize9x	Scale9x	FastResize900	Resize900	SharpResize900
548ms	1748ms	2778ms	3061ms	10115ms

Resizing the reference picture to 1200% size:

Resize12x	Scale12x	FastResize1200	Resize1200	SharpResize1200
999ms	2961ms	4900ms	5393ms	16283ms

Resizing the reference picture to 1600% size:

Resize16x	Scale16x	FastResize1600	Resize1600	SharpResize1600
2483ms	4670ms	8586ms	9725ms	19596ms

Conclusion:

Overall the ResizeXx, ScaleXx and FastResizeXXX algorithms are faster than the reference ResizeXXX algorithm. With the ResizeXx as fastest algorithm this can be easily up to 2 to 4 times faster than the reference ResizeXXX algorithm!

Usage

ResizeX is a commandline tool which means it is controlled via commandline arguments within a shell or script.

For a full help on the available commandline arguments type:

Resize.exe -h  or  Resize.exe --help

For help on the available algorithms type:

Resize.exe -l

A standard example for resizing a picture with a fixed scaler algorithm is:

Resize.exe -i [path_to_input_image] -o [path_to_output_image] -s [scalertype]

A standard example for resizing a picture with the FastResize scaler algorithm is:

Resize.exe -i [path_to_input_image] -o [path_to_output_image] -f -x [new_width] -y [new_height]

A standard example for a performance test without saving the result images:

Resize.exe -i [path_to_input_image] -n

A standard example for a performance test with saving the result images:

Resize.exe -i [path_to_input_image] -o [path_to_output_image] -p

     

Copyright

© Alexander Feuster 2023

ResizeX and Feuster.Imaging.Resizing.Lib are licensed unter the GNU GENERAL PUBLIC LICENSE Version 2