Skip to main content

export - High memory use while exporting a BMP


Something mysterious is going on when exporting images. I'm using Mathematica 9.0.1 on Windows 7 (64 bit), and find that the memory use is excessive when exporting to a bmp. But it doesn't happen with other image formats.


For example, running the following (with different values for k):


k = 1;  extensions = {"png", "tif", "jpg", "bmp"};  SeedRandom[42];
image = Image@RandomReal[{0, 1}, {1500, 1500, 3}];
base = MaxMemoryUsed[];

Export["C:\\mem." <> extensions[[k]], image];
{extensions[[k]], (MaxMemoryUsed[] - base)/1024.^2}
Quit[]

gives


{"png",  79.987}
{"tif", 28.617}
{"jpg", 60.453}
{"bmp", 927.602}


file sizes (added by s0rce):


png: 12.8 MB
tif: 12.8 MB
jpg: 1.29 MB
bmp: 6.43 MB

Why does exporting a bmp require over 10 times the memory of other image formats? And what can I do about it?


EDIT:


Here's the results of some more testing (code at the end).


format  texture  type    memory(MB)  time(s)  size(KB)


png random real 89 3.4 15 028
bit16 0 2.6 15 028
byte 0 1.5 7 515
smooth real 75 1.3 726
bit16 0 1.2 726
byte 0 0.5 85

tif random real 17 0.2 15 000
bit16 0 0.1 15 000

byte 0 0.1 7 500
smooth real 17 0.3 15 000
bit16 0 0.1 15 000
byte 0 0.2 7 500

jpg random real 68 0.4 1 500
bit16 0 0.8 1 503
byte 0 0.7 1 500
smooth real 68 0.3 50
bit16 0 0.3 50

byte 0 0.3 50

bmp random real 1069 25.3 7 500
bit16 967 22.8 7 500
byte 69 $Aborted 0
smooth real 1055 100.8 7 500
bit16 953 32.4 7 500
byte 71 1.4 7 500

Notes:




  • By default Mathematica will treat the image type as Real.

  • When exporting, a Real type image will need to be converted to an integer type. This requires extra memory and time.

    • tif and png support 16 bit pixels, so this will be used by default for Real images.

    • jpg and bmp only support up to 8 bits per pixel.



  • My computer was thrashing the memory disk cache for the bmp smooth real, so it took extra time.

  • The bmp random byte was using 100% for 5 minutes before I gave up.



Conclusions:



  • Use tif for speed.

  • Use png for saving disk space.

  • Never use jpg because it is lossy.

  • Never use bmp because it is broken.


Code:


k = 1; (* 1 .. 24 *)

{e, t, c} = Tuples[{Range[4], Range[3], Range[2]}][[k]];
extensions = {"png", "tif", "jpg", "bmp"};
types = {"real", "byte", "bit16"};
compress = {"random", "smooth"};
res = 1600;
SeedRandom[42];
data = Switch[compress[[c]],
"random", RandomReal[{0, 1}, {res, res, 3}],
"smooth", Table[{i, j, 1 - i j}, {i, 0., 1, 1/(res - 1)}, {j, 0., 1, 1/(res - 1)}]
];

image = Switch[types[[t]],
"real", Image@data,
"byte", Image[Round[(2^8 - 1) data], "Byte"],
"bit16", Image[Round[(2^16 - 1) data], "Bit16"]
];
filename = "C:\\_" <> types[[t]] <> "_" <> compress[[c]] <> "." <> extensions[[e]];
base = MaxMemoryUsed[];
time = AbsoluteTiming[Export[filename, image]][[1]];
CellPrint@TextCell[ToString@{
filename,(MaxMemoryUsed[] - base)/1024.^2, time, FileByteCount[filename]/1024.

}, "Text", CellAutoOverwrite -> False]
Quit[]

Answer



I believe this is a bug in Mathematica. It can be verified that the unreasonable memory usage is due to array unpacking by


On["Packing"]
ExportString[Image@RandomReal[1, {30, 30}], "BMP"];

So please report it to support@wolfram.com so this can be fixed for the benefit of all of us, and point them to this discussion.


Some further playing with Simon Woods's excellent spelunker and the (much underrated) built-in debugger revealed the culprit. It is an innocent looking line inside System`Convert`BitmapDump`ExportBMP24:


Module[{bitmap, ...},

...
bitmap = Floor[(255 bitmap)/max]; (* <-- this unpacks *)
...
]

Now bitmap appears to be a packed array of integer values and max is an integer as well. Did you already guess why this unpacks? Hover below for the spoiler.



255/max evaluates to a Rational and of course multiplying integers by Rationals will typically give a Rational and unpack the integer array. Using Floor[(255. bitmap)/max] would fix the problem.



This shows how counterintuitive Mathematica can be at times. I would have fallen into the same trap.





To verify that what I'm saying here is correct, one can do the following experiment:


MaxMemoryUsed[ExportString[Image@RandomReal[1, {1500, 1500}], "BMP"];]

(* ==> 1007548312 , i.e. 1 GB *)

Now evaluate DownValues[System`Convert`BitmapDump`ExportBMP24], search for Floor, change 255 to 255. and set the resulting expression back to DownValues[System`Convert`BitmapDump`ExportBMP24] = .... Let's measure the memory usage again:


MaxMemoryUsed[ExportString[Image@RandomReal[1, {1500, 1500}], "BMP"];]

(* ==> 289673368 i.e. 290 MB *)


Now it uses much less memory and ExportString finishes much faster.


The reason why it still uses quite a bit of memory is other flaws in the code which cause partial unpacking further down the pipeline.


Comments

Popular posts from this blog

plotting - Plot 4D data with color as 4th dimension

I have a list of 4D data (x position, y position, amplitude, wavelength). I want to plot x, y, and amplitude on a 3D plot and have the color of the points correspond to the wavelength. I have seen many examples using functions to define color but my wavelength cannot be expressed by an analytic function. Is there a simple way to do this? Answer Here a another possible way to visualize 4D data: data = Flatten[Table[{x, y, x^2 + y^2, Sin[x - y]}, {x, -Pi, Pi,Pi/10}, {y,-Pi,Pi, Pi/10}], 1]; You can use the function Point along with VertexColors . Now the points are places using the first three elements and the color is determined by the fourth. In this case I used Hue, but you can use whatever you prefer. Graphics3D[ Point[data[[All, 1 ;; 3]], VertexColors -> Hue /@ data[[All, 4]]], Axes -> True, BoxRatios -> {1, 1, 1/GoldenRatio}]

plotting - Filling between two spheres in SphericalPlot3D

Manipulate[ SphericalPlot3D[{1, 2 - n}, {θ, 0, Pi}, {ϕ, 0, 1.5 Pi}, Mesh -> None, PlotPoints -> 15, PlotRange -> {-2.2, 2.2}], {n, 0, 1}] I cant' seem to be able to make a filling between two spheres. I've already tried the obvious Filling -> {1 -> {2}} but Mathematica doesn't seem to like that option. Is there any easy way around this or ... Answer There is no built-in filling in SphericalPlot3D . One option is to use ParametricPlot3D to draw the surfaces between the two shells: Manipulate[ Show[SphericalPlot3D[{1, 2 - n}, {θ, 0, Pi}, {ϕ, 0, 1.5 Pi}, PlotPoints -> 15, PlotRange -> {-2.2, 2.2}], ParametricPlot3D[{ r {Sin[t] Cos[1.5 Pi], Sin[t] Sin[1.5 Pi], Cos[t]}, r {Sin[t] Cos[0 Pi], Sin[t] Sin[0 Pi], Cos[t]}}, {r, 1, 2 - n}, {t, 0, Pi}, PlotStyle -> Yellow, Mesh -> {2, 15}]], {n, 0, 1}]

plotting - Mathematica: 3D plot based on combined 2D graphs

I have several sigmoidal fits to 3 different datasets, with mean fit predictions plus the 95% confidence limits (not symmetrical around the mean) and the actual data. I would now like to show these different 2D plots projected in 3D as in but then using proper perspective. In the link here they give some solutions to combine the plots using isometric perspective, but I would like to use proper 3 point perspective. Any thoughts? Also any way to show the mean points per time point for each series plus or minus the standard error on the mean would be cool too, either using points+vertical bars, or using spheres plus tubes. Below are some test data and the fit function I am using. Note that I am working on a logit(proportion) scale and that the final vertical scale is Log10(percentage). (* some test data *) data = Table[Null, {i, 4}]; data[[1]] = {{1, -5.8}, {2, -5.4}, {3, -0.8}, {4, -0.2}, {5, 4.6}, {1, -6.4}, {2, -5.6}, {3, -0.7}, {4, 0.04}, {5, 1.0}, {1, -6.8}, {2, -4.7}, {3, -1....