Combine command in EBImage package in R

Question

I am using a combine command in EBImage in R. My images files are upto 100 images I have also set expression with following command

options(expressions=500000)

While I use combine function from EBImage package with following code

combine(Images_0)

I get a error command as Error: C stack usage 15925536 is too close to the limit

What am I doing wrong?

Answer 1

This assumes the Bioconductor package EBImage is available.

This many not exactly answer your question but I hope it is helpful. I am not certain of where the code is having problems but I find that combine() can be very inefficient. There may be a do.call() call buried in the call to combine, which can consume significant memory.

I find that abind() of the abind package is often a very effective solution. Without knowing more about the images you are trying to combine (it does help to provide minimal reproducible examples), let me suggest the following possibility.

# load library and read example color image
  library(EBImage)
  img <- readImage(system.file("images", "sample-color.png", package="EBImage"))
  dim(img)
> [1] 768 512   3

# create a list of 48 images differing in lightness
  images <- lapply(runif(48, 0.4, 1/0.4), function(x) img/x)

This is a reasonably sized list of color images. You can view the first image with print(img) . The last command created 48 copies of the image that vary in lightness.

Now compare the time required to combine the images in that list with combine() versus abind() . Incidentally, abind() has been slightly modified by EBImage to handle dimnames differently.

# First with combine()
  system.time(x <- combine(images)) # rather fast AMD Ryzen 12-Core CPU 
>  user  system elapsed 
> 12.21    3.68   15.90

# Now with abind() where along = 4 adds a fourth dimension needed for the combination 
  system.time(y <- abind(images, along = 4)) # but much faster here
>  user  system elapsed 
>  0.53    0.02    0.55 

# The combined images are the same
    identical(x, y)
> [1] TRUE

The timing would likely vary with machines and the need for disk access but it seems like a pretty big difference to me. With many images, abind is really helpful. You can examine the combination with print(y, nx = 8, all = TRUE) .

By the way, I would be curious to know if this fixed the "stack usage error."