# NaN-aware \`imresize\`
**URL:** https://discourse.julialang.org/t/nan-aware-imresize/117526
**Category:** General Usage
**Tags:** question
**Created:** [July 26, 2024, 5:02pm UTC](https://discourse.julialang.org/t/nan-aware-imresize/117526 "2024-07-26T17:02:48Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@gdkrmr](https://discourse.julialang.org/u/gdkrmr)
#### Post date: [July 26, 2024, 5:02pm UTC](https://discourse.julialang.org/t/nan-aware-imresize/117526/1 "2024-07-26T17:02:48Z")
Is there an easy way hack `ImageTransforms.imresize` to not propagate `NaN` values?
```julia
julia> using ImageTransformations
julia> im = Float32[
1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 NaN
]
4×4 Matrix{Float32}:
1.0 2.0 3.0 4.0
5.0 6.0 7.0 8.0
9.0 10.0 11.0 12.0
13.0 14.0 15.0 NaN
julia> imresize(im, (2, 2))
2×2 Matrix{Float32}:
3.5 5.5
11.5 NaN
```
In result the lower right corner should be `mean(11, 12, 15)`.
---
### Author:  [@juliohm](https://discourse.julialang.org/u/juliohm)
#### Post date: [July 26, 2024, 6:50pm UTC](https://discourse.julialang.org/t/nan-aware-imresize/117526/2 "2024-07-26T18:50:30Z")
Perhaps you can clarify if you need a solution with ImageTransforms.jl specifically. There are various other packages that can achieve this.
---
### Author:  [@gdkrmr](https://discourse.julialang.org/u/gdkrmr)
#### Post date: [July 26, 2024, 7:09pm UTC](https://discourse.julialang.org/t/nan-aware-imresize/117526/3 "2024-07-26T19:09:54Z")
No need to use ImageTransforms.jl.
---
### Author:  [@juliohm](https://discourse.julialang.org/u/juliohm)
#### Post date: [July 26, 2024, 7:27pm UTC](https://discourse.julialang.org/t/nan-aware-imresize/117526/4 "2024-07-26T19:27:51Z")
One possible solution is to use the `Aggregate` transform from GeoStats.jl:
```julia
Aggregate(domain, var₁ => agg₁, var₂ => agg₂, ..., varₙ => aggₙ)
Aggregate([g₁, g₂, ..., gₙ], var₁ => agg₁, var₂ => agg₂, ..., varₙ => aggₙ)
Aggregate variables var₁, var₂, ..., varₙ over geospatial domain using aggregation functions agg₁, agg₂, ..., aggₙ. Alternatively, aggregate variables over geometries g₁,
g₂, ..., gₙ. Default aggregation function is mean for continuous variables and first otherwise.
Examples
≡≡≡≡≡≡≡≡
Aggregate(domain, 1 => last, 2 => maximum)
Aggregate(domain, :a => first, :b => minimum)
Aggregate(domain, "a" => last, "b" => maximum)
Aggregate(geoms, "a" => last, "b" => maximum)
```
Pass your own aggregation function `aggfun` to ignore `NaN` values:
```julia
img = GeoIO.load("foo.jpg")
grid = CartesianGrid(M, N)
img |> Aggregate(grid, "color" => aggfun)
```
Of course this solution is mostly useful if you are already taking advantage of the full stack we are building for geospatial data science. Others can suggest simpler solutions with smaller packages.
In case you want to learn more: [Geospatial Data Science with Julia](https://juliaearth.github.io/geospatial-data-science-with-julia/)