Introduction

library(picohdr)

Read and display a PFM image

PFM (Portable Float Map) is a simple format for storing Gray or RGB colour data. Pixels are stored as 32-bit floating point values in a simple array format.

Load and display a (PFM) image.

Apply tone-mapping to the image and adjust the gamma correction prior to display.

filename <- system.file("image/rstats.pfm.bz2", package = "picohdr") 
im <- read_pfm(filename)
dim(im)
#> [1] 280 420   3

im |> 
  tm_reinhard() |>
  adj_gamma() |> 
  plot()

Read and display an EXR image

OpenEXR images are a more complex HDR image container. Pixel types may be 16-bit floats, 32-bit floats or 32-bit unsigned integers.

Images can contain any number of channels with arbitrary names - this includes data that isn’t related to the colour of the display e.g. image depth, texture coordinates.

The actual image data may be stored as scanlines, tiles or deep images in single- or multi-part formats.

Currently, this package only supports reading single-part scanline images which are using NONE, ZIP or ZIPS compression.

The steps included below:

library(picohdr)

# EXR file of meta-information about the rendered scene
filename <- system.file("image/rstats.exr", package = "picohdr") 

# Load all images
images <- read_exr(filename)
dim(im)
#> [1] 280 420   3

# Channel names. EXR format wants channels arranged alphabetically
dimnames(images)[[3]]
#>  [1] "Albedo.B"           "Albedo.G"           "Albedo.R"          
#>  [4] "B"                  "G"                  "N.X"               
#>  [7] "N.Y"                "N.Z"                "Ns.X"              
#> [10] "Ns.Y"               "Ns.Z"               "P.X"               
#> [13] "P.Y"                "P.Z"                "R"                 
#> [16] "RelativeVariance.B" "RelativeVariance.G" "RelativeVariance.R"
#> [19] "Variance.B"         "Variance.G"         "Variance.R"        
#> [22] "dzdx"               "dzdy"               "u"                 
#> [25] "v"

# Extract RGB channels. Tone-map. Adjust gamma.
images[,,c('R', 'G', 'B')] |>
  tm_reinhard() |>
  adj_gamma() |>
  plot()


# Plot the albedo Green channel
plot(images[, , 'Albedo.G'])

Display a non-colour channel

This EXR file includes information about the surface derivative at each point in the image in the dzdx channel.

This value may be negative or positive, so we will map the values into the standard range [0, 1] so it can be visualised.

# Rescale the derivative channel to the range [0,1] and display
images[,,'dzdx'] |>
  adj_rescale(0, 1) |> 
  plot()

EXR metadata

Meta-information about the contents of an EXR file can be extracted for all EXR image types.

exr_info(filename)
#> $version
#> $version$number
#> [1] 2
#> 
#> $version$single_tile
#> [1] FALSE
#> 
#> $version$long_name
#> [1] FALSE
#> 
#> $version$non_image
#> [1] FALSE
#> 
#> $version$multipart
#> [1] FALSE
#> 
#> $version$desc
#> [1] "single-part scanline"
#> 
#> 
#> $channels
#>                  name type pLinear xSampling ySampling
#> 1            Albedo.B half       0         1         1
#> 2            Albedo.G half       0         1         1
#> 3            Albedo.R half       0         1         1
#> 4                   B half       0         1         1
#> 5                   G half       0         1         1
#> 6                 N.X half       0         1         1
#> 7                 N.Y half       0         1         1
#> 8                 N.Z half       0         1         1
#> 9                Ns.X half       0         1         1
#> 10               Ns.Y half       0         1         1
#> 11               Ns.Z half       0         1         1
#> 12                P.X half       0         1         1
#> 13                P.Y half       0         1         1
#> 14                P.Z half       0         1         1
#> 15                  R half       0         1         1
#> 16 RelativeVariance.B half       0         1         1
#> 17 RelativeVariance.G half       0         1         1
#> 18 RelativeVariance.R half       0         1         1
#> 19         Variance.B half       0         1         1
#> 20         Variance.G half       0         1         1
#> 21         Variance.R half       0         1         1
#> 22               dzdx half       0         1         1
#> 23               dzdy half       0         1         1
#> 24                  u half       0         1         1
#> 25                  v half       0         1         1
#> 
#> $compression
#> [1] "ZIP"
#> 
#> $dataWindow
#> [1]   0   0 359 269
#> 
#> $displayWindow
#> [1]   0   0 359 269
#> 
#> $lineOrder
#> [1] "increasing"
#> 
#> $pixelAspectRatio
#> [1] 1
#> 
#> $renderTimeSeconds
#> [1] 43.84048
#> 
#> $samplesPerPixel
#> [1] 1024
#> 
#> $screenWindowCenter
#> [1] 0 0
#> 
#> $screenWindowWidth
#> [1] 1
#> 
#> $worldToCamera
#>            [,1]       [,2]          [,3]       [,4]
#> [1,]  0.7657048 -0.6431921 -1.233992e-09 -1.2251276
#> [2,] -0.2218857 -0.2641496  9.386115e-01  0.1672952
#> [3,] -0.6037076 -0.7186995 -3.449758e-01 14.9489479
#> [4,]  0.0000000  0.0000000  0.000000e+00  1.0000000
#> 
#> $worldToNDC
#>            [,1]       [,2]       [,3]       [,4]
#> [1,]  4.5627465 -4.4456143 -0.1724879 -0.3088861
#> [2,] -1.7115160 -2.0375190  5.7906065  8.5373173
#> [3,] -0.6037106 -0.7187030 -0.3449775 14.9440222
#> [4,] -0.6037076 -0.7186995 -0.3449758 14.9489479
#> 
#> $chunk_offsets
#>  [1]    1202   79967  158319  238640  338013  456525  582637  720330  863085
#> [10] 1000992 1127091 1242213 1339482 1428107 1514095 1597924 1680690

EXR Implementation Notes

The code for handling EXR images is mostly written in base R with the exception of a core compression/decompression component which is written in C:

The ZIP compression mode applies a predictor and de-interleaves bytes. The byte predictor and interleaving can be in R but is 20x faster in C.

EXR support is for a subset of images which:

This package does not support the following EXR features:

If you would like support for these features please file an issue on GitHub. A link to a free/openly-licensed image containing your requested features would be appreciated.