Image Recognition and replacement tool

Hi guys, i'm new around here...

I need an information... with KNIME Image Processing  can i accomplish this:

i have a folder "A" with 3000 images, all are 400px x 400px and named from 1 to 3000.

Also I have an other folder "B" with the original images, the same of folder "A"  but in hi-resolution.. All are 2000px x 2000px named with the original names...

I want to find a tool that helps me to recognize images from the folder "A" in low resolution and replace them with the original ones and keeping the name from the folder "A", so from 1 to 3000.

Hi appraily,

giving a precise answer to your question is a bit hard, without having seen the images, as the success of your planned analysis heavily depends on several factors:

* is there always a high-res image for a low res image?

* are the low res images aquired the same way (angle, position etc) as the high-res images?

* are the low res images just down-scaled versions of the high-res images?

* what kind of images are you analyzing (microscope data, MRT/CT, holiday? :-)).

if you can provide e.g. 10 examples for low-res and high-res images, then I can try to setup a small workflow.



Hi Christian, thanks for the quick answer.

I'm restyling a Music Shop, Italian Folklore Music. These images are the covers of sold albums. The previous website system was downgrading the original images to 400px x 400px or 600px x600px and rename them with a sequential number.

Here I've created a folder of examples: ((( Don't lol at the high level of trash in the pictures, this Italian top-class trash music )))

There is always a hi-res image for the low res image, but the folder who contain Hi-res images has like 7k images, so only 3k images have been downgraded.

Not all images have 1:1 ratio (Cd covers) , some have 61:91 ratio (DVD covers)

Not all original images have the same dimensions: some one could be 2000px x 2000px others 1400px x 1400px, plus some one could be 300dpi other 72dpi. Low res ones are always 72dpi.

I hope this information are good.

Kind regards,



Hi Michele,

Interesting! We will definitevly try to set something up (will be a combination of KNIME Image Processing and the KNIME Pyton Integration). However, I'm afraid we can't make it before new year.

I hope that is fine?



i don't know how to thank you, in the mean time  i'll be on understanding how Knime Tool are working and reading some RES.

Have a great holidays!



Hi Michele,

I create a KNIME Workflow which should solve your problem.

This workflow matches the images by subtracting the downscaled highres images from the lowres images.

I get a perfect match for all of your example images :)

Happy New Year!