What is the default resize filter?

[cj-holmes]

Hi there, I have a question/help wanted issue.

I'm trying to work out what the default filter used with image_resize() is.
I see from the R package documentation that it is set to NULL image_resize(image, geometry = NULL, filter = NULL)

So I looked in the Magick++ documentation (https://www.imagemagick.org/Magick++/STL.html) and see that it might be Lanczos

Filter to use when resizing image. The reduction filter employed has a sigificant effect on the time required to resize an image and the resulting quality. The default filter is Lanczos which has been shown to produce good results when reducing images.

But when I compare an image resized with the default filter = NULL argument, and one resized by specifying filter = "Lanczos" - they are not the same (see example below). Is anyone able to help?

Thanks!

waldo::compare(
  image_read('https://www.r-project.org/logo/Rlogo.png') |> 
  image_resize("50x", filter = NULL) |>
  image_raster(),
  
  image_read('https://www.r-project.org/logo/Rlogo.png') |> 
  image_resize("50x", filter = "Lanczos") |>
  image_raster()
)

Output

old vs new
                      col
  old[9, ]    transparent
  old[10, ]   transparent
  old[11, ]   transparent
- old[12, ]   transparent
+ new[12, ]   #c0c2c502  
- old[13, ]   transparent
+ new[13, ]   #bfc2c504  
- old[14, ]   transparent
+ new[14, ]   #bebfc201  
  old[15, ]   transparent
  old[16, ]   transparent
- old[17, ]   #bec0c40b  
+ new[17, ]   transparent
- old[18, ]   #bec0c327  
+ new[18, ]   #bec1c416  
and 24 more ...

[cj-holmes]

I have just discovered that using filter Undefined makes the outputs the same.
So what resize method is being used when filter = 'Undefined'?

waldo::compare(
  image_read('https://www.r-project.org/logo/Rlogo.png') |> 
    image_resize("50x", filter = NULL) |>
    image_raster(),
  
  image_read('https://www.r-project.org/logo/Rlogo.png') |> 
    image_resize("50x", filter = "Undefined") |>
    image_raster()
)

✔ No differences

[cj-holmes]

OK, so I seem to be answering my own question here! But Ive just tested and it looks like the Undefined filter and the Mitchell filter return the same for me.

So is it the Mitchell filter that is used by default? Would love some help with this is possible!

o <-
  lapply(
  filter_types(),
  function(x){
    waldo::compare(
      magick::image_read('https://www.r-project.org/logo/Rlogo.png') |> 
        magick::image_resize("10x", filter = NULL) |>
        magick::image_raster(),
      
      magick::image_read('https://www.r-project.org/logo/Rlogo.png') |> 
        magick::image_resize("10x", filter = x) |>
        magick::image_raster()
    )
  })

filter_types()[which(lengths(o) == 0)]
[1] "Undefined" "Mitchell"