Is it possible to save a 32bit BMP image (with alpha)

Is it possible to save a 32bit BMP image. i.e. image size (imheight x imwidth x 4) where the 4.th channel consist of alpha values?

Answer 1

[Walter Roberson]

imwrite does not support alpha in BMP. See http://www.mathworks.com/matlabcentral/answers/6335-writing-32-bit-images for information on what can be used.

[Slimane]

thank you Walter for your relpy..

I tried also using TIFF and RAS, where i add alpha layer but the size of the new image remains 24BPP as input image.. here is a code sample:

%%%%%%%%%%%%%%%%%%%%%

%Input image (376x240x3)
InpIm = imread('inputImage.bmp'); % InpIm is of uint8 type
[rows,cols,chan] = size(InpIm);

%Build alpha layer
InpIm(:,:,4) = 0.5*ones(rows,cols,'uint8');

%Save as Tiff image
imwrite(InpIm,'New.tiff','TIFF'); %New.tiff is also 24BPP !!

%%%%%%%%%%%%%%%%%%%%%

Does any one have an example where one can save an image as 32BPP (RGBA)??

[Walter Roberson]

uint8(0.5*uint8(1)) is uint8(0.5) which is 1, so the alpha value you would be writing would be 1 out of 256. You should be writing 0.5 * intmax('uint8') if you want alpha of 1/2

Anyhow, if you have a new enough MATLAB, use the Tiff class, http://www.mathworks.com/help/techdoc/ref/tiffclass.html

[Slimane]

I am actually more concerned by the final size of the image. the resulted image is always 24BPP and not as expected 32BPP, (since i added an extra layer i.e. InpIm(:,:,4))

[Walter Roberson]

I just tried,

foo = uint8(rand(64,64,4) * 256);

imwrite(foo,'New.tiff','TIFF');

imfinfo('New.tiff')

ans =

Filename: 'New.tiff'

FileModDate: '11-May-2011 13:50:06'

FileSize: 16734

Format: 'tif'

FormatVersion: []

Width: 64

Height: 64

BitDepth: 32

ColorType: 'CMYK'

FormatSignature: [73 73 42 0]

ByteOrder: 'little-endian'

NewSubFileType: 0

BitsPerSample: [8 8 8 8]

Compression: 'PackBits'

PhotometricInterpretation: 'CMYK'

StripOffsets: [2x1 double]

SamplesPerPixel: 4

RowsPerStrip: 32

StripByteCounts: [2x1 double]

XResolution: 72

YResolution: 72

ResolutionUnit: 'Inch'

Colormap: []

PlanarConfiguration: 'Chunky'

TileWidth: []

TileLength: []

TileOffsets: []

TileByteCounts: []

Orientation: 1

FillOrder: 1

GrayResponseUnit: 0.01

MaxSampleValue: [255 255 255 255]

MinSampleValue: 0

Thresholding: 1

What does imfinfo() tell you about your file?

[Slimane]

Yes i can see that bitdepth is 32, but when i open the saved image in e.g. 'irfanView' it says the image size is 240x376x24BPP. Here is the imfinfo i receive:

ans =

                     Filename: 'New.tiff'
                  FileModDate: '11-May-2011 21:52:55'
                     FileSize: 364374
                       Format: 'tif'
                FormatVersion: []
                        Width: 376
                       Height: 240
                     BitDepth: 32
                    ColorType: 'unknown'
              FormatSignature: [73 73 42 0]
                    ByteOrder: 'little-endian'
               NewSubFileType: 0
                BitsPerSample: [8 8 8 8]
                  Compression: 'PackBits'
    PhotometricInterpretation: 'CMYK'
                 StripOffsets: [48x1 double]
              SamplesPerPixel: 4
                 RowsPerStrip: 5
              StripByteCounts: [48x1 double]
                  XResolution: 72
                  YResolution: 72
               ResolutionUnit: 'Inch'
                     Colormap: []
          PlanarConfiguration: 'Chunky'
                    TileWidth: []
                   TileLength: []
                  TileOffsets: []
               TileByteCounts: []
                  Orientation: 1
                    FillOrder: 1
             GrayResponseUnit: 0.0100
               MaxSampleValue: [255 255 255 255]
               MinSampleValue: 0
                 Thresholding: 1

[Walter Roberson]

It might be an irfanView issue. I don't have an image processing program around that I can cross-check with.

[Slimane]

thanks a lot Walter, i guess i will just use c++ it is much easier :)

[DGM]

This problem is occurring because imwrite() interprets a 4-channel RGBA image as CMYK (regardless of any asserted colorspace parameter). So anything that decodes the file will likely wind up treating it as CMYK and converting it back into an opaque 24b RGB equivalent.

As far as I know, writing an RGBA TIFF this way isn't an option with imwrite(). It should be possible with the tiff class.

Answer 2

[John]

Here's a demo I did a while back on how to create a TIFF with an alpha layer.

% This image is included with MATLAB
D = imread('street1.jpg');
image(D);  title('RGB image without alpha layer.');

h = size(D,1);
w = size(D,2);

% Create a gradually increasing transparency mask 
% from left-to-right across the image.  
T = zeros(h,w,'uint8');
for j=1:w
  T(:,j) = min(255,j) * ones(h,1); 
end

% The dimensions of our image are h x w x 3 (remember, it's RGB).  If 
% we wish to add an alpha component, we need to give the image data a 4th
% component.  
rgbaData = cat(3,D,T);

% Let's create our new image.
t = Tiff('alpha.tif','w');

% The photometric interpretation remains RGB, even though we will be
% adding an extra sample.
t.setTag('Photometric',Tiff.Photometric.RGB);

% These three tags define the dimensions of the image in the file. 
t.setTag('ImageWidth',w);
t.setTag('ImageLength',h);
t.setTag('BitsPerSample',8);
t.setTag('SamplesPerPixel',4);

% We will make the image consist of 64x64 tiles.
t.setTag('TileWidth',64);
t.setTag('TileLength',64);

t.setTag('PlanarConfiguration',Tiff.PlanarConfiguration.Chunky);

% A photometric interpretation of RGB means that the TIFF file is expecting
% three components.  We have to tell it that there will really be four
% components (the alpha layer is the 4th).
t.setTag('ExtraSamples',Tiff.ExtraSamples.UnassociatedAlpha);

t.write(rgbaData);
t.close();