Cropping BMP and filesizes

[chuko]

Hi,

I am using the CLI imagemagick convert command to crop a batch of BMP files at different geometry offsets.
The original images are BMP and have the same geometry and filesize.

When converting a pair using the following commands:

convert -crop 4008x1916+0+285 +repage -extent 4008x1916 -type palette -strip set01_0001A.bmp set01crop_0001A.bmp
convert -crop 4008x1916+0+285 +repage -extent 4008x1916 -type palette -strip set01_0001B.bmp set01crop_0001B.bmp

it creates two cropped images of different sizes which are LARGER than the original files (10710454).
set01crop_0001A.bmp - filesize: 12129946
set01crop_0001B.bmp - filesize: 11944798

Is there anyway to force them to the same filesize as each other? And make them smaller than the original?

[fmw42]

You are not using the correct syntax for IM 6 which has the input image right after the convert, e.g.

convert input options output


That may not make a difference but try (rearranging and adding +dither -colors 256 since you are outputting an 8-bit palette format)

convert input -strip -crop 4008x1916+0+285 +repage -extent 4008x1916 +dither -colors 256 -type palette output

However consider that IM produces BMP4 by default and your input images may be BMP2 or BMP3.

For more details and limitations, see http://www.imagemagick.org/Usage/formats/#bmp

also see http://www.imagemagick.org/Usage/basics/#why

[chuko]

Thanks for the quick reply.

Your advice about IM default output of BMP4 led me to test BMP2 and BMP3, the former providing a consistent filesize for a fixed geometry. However, the program I use to process this data incorrectly reads the dimensions [125570984 x 524289] when a BMP2 format is input. So, I guess I need to use BMP3 or BMP4 - however a requirement is that the filesize is consistent.

I don't understand why two BMP3 and BMP4 files of the same dimensions would have different file sizes - is it the header content?

You are not using the correct syntax for IM 6 which has the input image right after the convert, e.g.

convert input options output


That may not make a difference but try (rearranging and adding +dither -colors 256 since you are outputting an 8-bit palette format)

convert input -strip -crop 4008x1916+0+285 +repage -extent 4008x1916 +dither -colors 256 -type palette output

However consider that IM produces BMP4 by default and your input images may be BMP2 or BMP3.

For more details and limitations, see http://www.imagemagick.org/Usage/formats/#bmp

also see http://www.imagemagick.org/Usage/basics/#why

[fmw42]

without -colors 256 it is possible that BMP4 is not 8-bit but 24-bit and perhaps your input image was truly 8-bit. compare the verbose info for each image to see its bit-depth

identify -verbose image

also as you suggest, it could be the headers.

[chuko]

Hi,

I've been using identify to try and find out more. The original image is:

Code: Select all

set01_0001A.bmp BMP 4008x2672 PseudoClass 256c 10.2mb 
Image: set01_0001A.bmp
  Format: BMP (Microsoft Windows bitmap image)
  Geometry: 4008x2672
  Class: PseudoClass
  Type: Grayscale
  Endianess: Undefined
  Colorspace: Gray
  Channel depth:
    Gray: 8-bits
  Channel statistics:
    Gray:
      Min: 0 (0)
      Max: 255 (1)
      Mean: 58.6435 (0.229975)
      Standard deviation: 37.7538 (0.148054)
  Colors: 256
  Histogram:
<LOTS OF HISTOGRAM INFO>
  Rendering-intent: Undefined
  Units: Undefined
  Filesize: 10.2mb
  Interlace: None
  Background Color: white
  Border Color: #DFDFDF
  Matte Color: grey74
  Dispose: Undefined
  Iterations: 0
  Compression: Undefined
  Orientation: Undefined
  Signature: 0c239517c97f7d2eaa5e9816f9b140a53e2e708edc17aac19d100cbfe6224785
  Tainted: False
  Version: ImageMagick 6.2.4 02/10/07 Q16 http://www.imagemagick.org

and the cropped one is

Code: Select all

set01crop_0001A.bmp BMP 4008x1916 PseudoClass 256c 13.6mb 
Image: set01crop_0001A.bmp
  Format: BMP (Microsoft Windows bitmap image)
  Geometry: 4008x1916
  Class: PseudoClass
  Type: Grayscale
  Endianess: Undefined
  Colorspace: Gray
  Channel depth:
    Gray: 8-bits
  Channel statistics:
    Gray:
      Min: 0 (0)
      Max: 255 (1)
      Mean: 58.9603 (0.231217)
      Standard deviation: 37.7652 (0.148099)
  Colors: 256
  Histogram:
<LOTS OF HISTOGRAM INFO>
  Rendering-intent: Undefined
  Units: Undefined
  Filesize: 13.6mb
  Interlace: None
  Background Color: white
  Border Color: #DFDFDF
  Matte Color: grey74
  Dispose: Undefined
  Iterations: 0
  Compression: Undefined
  Orientation: Undefined
  Signature: 8c757a13cf7979d12d1275faa6f3b423c9cf19b3d0e01130769dd33591d5e1e9
  Tainted: False
  Version: ImageMagick 6.2.4 02/10/07 Q16 http://www.imagemagick.org

They contain the same info, apart from filesize and geometry using

Code: Select all

convert set01_0001A.bmp -strip -crop 4008x1916+0+285 +repage -extent 4008x1916 +dither -colors 256 -type palette set01crop_0001A.bmp

and the same result if I specify BMP2, except with different filesize.

So, I suspect at the moment that I need to investigate the header and info header, and a fixed palette (if it is added to BMP).

[fmw42]

look at the length (size) of the color palette in each image.

IM 6.2.4 is rather old. You may find that upgrading fixes lots of problems and perhaps this. You are about 380 versions old.

[fmw42]

Code: Select all

convert set01_0001A.bmp -strip -crop 4008x1916+0+285 +repage -extent 4008x1916 +dither -colors 256 -type palette set01crop_0001A.bmp

-extent 4008x1916 is superfluous since your have already cropped your image to 4008x1916.

Try removing it and see what happens to the file size.

[fmw42]

see http://en.wikipedia.org/wiki/BMP_file_format (regarding headers and compression)

perhaps your input image was compressed (even though IM verbose info did not recognize it) and your output is not compressed.

[Drarakel]

Probably it's the other way round - you have uncompressed input files, but the output files are compressed. This doesn't get displayed in the verbose output (I'm using current IM versions though). I think, the default for ImageMagick when it has to write 8bit BMPs is to use a BMP V3 header and RLE compression (which can be inefficient).
You either have to write as 'BMP2:' (which creates a uncompressed file with BMP V1 header), or specify '-compress none'.
So, use something like this:

Code: Select all

convert set01_0001A.bmp -crop 4008x1916+0+285 +repage -compress none set01crop_0001A.bmp

[chuko]

Hi all,

thank you for all your advice - this worked. However, what is interesting is that the file size was smaller when -compress none was included. In fact it was the size of the 8bit x geometry + a little bit extra. Without it, it was double the size. The BMP2 worked to keep the size the same, but excluded some header info that the program I fed it to must have required.

Probably it's the other way round - you have uncompressed input files, but the output files are compressed. This doesn't get displayed in the verbose output (I'm using current IM versions though). I think, the default for ImageMagick when it has to write 8bit BMPs is to use a BMP V3 header and RLE compression (which can be inefficient).
You either have to write as 'BMP2:' (which creates a uncompressed file with BMP V1 header), or specify '-compress none'.
So, use something like this:

Code: Select all

convert set01_0001A.bmp -crop 4008x1916+0+285 +repage -compress none set01crop_0001A.bmp