tga 16 bit

[nilsca]

Hi

I'd love to see an option to convert to tga with 16 bit - not seeing that right now.

Thanks
Nils

[fmw42]

What version of IM are you using and what platform? TGA is both read/write capable as far as I know and works under Q16 compile. However, I do not know if 16-bit TGA images can be read or written? Do you have an example of a 16-bit TGA that fails to read? Or some other 16-bit image that fails to write to 16-bit tga? Does TGA support 16-bits? According to http://en.wikipedia.org/wiki/Truevision_TGA, it does not look like TGA supports 16-bits per channel (48-bits rgb). Are you looking for something different? It does seem to support 16-bits total per pixel (16-bit palette). Don't know if IM supports other than 8-bit palettes.

[nilsca]

Hi

ImageMagick 6.9.0-0 on windows reads the 16 bit fine but writes 8, 24 or 32 bit only. The special 16 bit encoding is described here

http://www.fileformat.info/format/tga/egff.htm

When a value of 15 appears in the PixelDepth field of the Image Specification section of the header, there are five bits each of red, green, and blue pixel data and one bit of overlay data in each pixel (see the section below called "Pixel Attribute Bits"). This is the format the Targa 16 uses to store data. Because these 16 bits are stored in two bytes of data, a little shifting and masking is required to read and write these pixel data values.

If I read this correctly ImageMagick writes either 8Bit or 24 bit (plus optionally one byte for alpha)

Code: Select all

static inline void WriteTGAPixel(Image *image,TGAImageType image_type,
  const IndexPacket *indexes,const PixelPacket *p)
{
  if (image_type == TGAColormap || image_type == TGARLEColormap)
    (void) WriteBlobByte(image,(unsigned char) GetPixelIndex(indexes));
  else
    {
      if (image_type == TGAMonochrome || image_type == TGARLEMonochrome)
        (void) WriteBlobByte(image,ScaleQuantumToChar(ClampToQuantum(
          GetPixelLuma(image,p))));
      else
        {
          (void) WriteBlobByte(image,ScaleQuantumToChar(GetPixelBlue(p)));
          (void) WriteBlobByte(image,ScaleQuantumToChar(GetPixelGreen(p)));
          (void) WriteBlobByte(image,ScaleQuantumToChar(GetPixelRed(p)));
          if (image->matte != MagickFalse)
            (void) WriteBlobByte(image,ScaleQuantumToChar(GetPixelAlpha(p)));
        }
    }
}

An option to write that 16bit or to write what was read would complete this. Here's how the read handles this correctly

Code: Select all

...
          case 16:
          {
            QuantumAny
              range;

            /*
              5 bits each of red green and blue.
            */
            j=(unsigned char) ReadBlobByte(image);
            k=(unsigned char) ReadBlobByte(image);
            range=GetQuantumRange(5UL);
            pixel.red=ScaleAnyToQuantum(1UL*(k & 0x7c) >> 2,range);
            pixel.green=ScaleAnyToQuantum((1UL*(k & 0x03) << 3)+
              (1UL*(j & 0xe0) >> 5),range);
            pixel.blue=ScaleAnyToQuantum(1UL*(j & 0x1f),range);
            break;
          }
...

I'll attach an example in a second. There are some consumers that only can read TGA 16 in my case (edge case, for sure).

Thanks for checking
Nils

[nilsca]

The original that I've downloaded (someone made in photoshop) - 16 bit

https://www.dropbox.com/s/9rm25oa6oqt1f ... l.tga?dl=0

the output of convert

https://www.dropbox.com/s/ua3k2tyk8xsfn ... d.tga?dl=0

[dlemstra]

tga16original.tga is a PNG file that has the extension .tga. Did you upload the wrong file?

[nilsca]

Apologies, I've messed up the files - here's an actual TGA input (says 5 bit which I find weird) and the 8 bit output that ImageMagick (built from source) converts to

convert ~/Dropbox/Forum/original5bit.tga ~/Dropbox/Forum/converted8bit.tga

converts to

https://www.dropbox.com/s/4sfetbnyivj3k ... t.tga?dl=0

nils@nmeier ~/workspace/ImageMagick-6.9.0-0 $ utilities/identify ~/Dropbox/Forum/original5bit.tga
/home/nils/Dropbox/Forum/original5bit.tga TGA 128x128 128x128+0+0 5-bit sRGB 32.8KB 0.000u 0:00.000

https://www.dropbox.com/s/q5csfacvsxfnq ... t.tga?dl=0

nils@nmeier ~/workspace/ImageMagick-6.9.0-0 $ utilities/identify ~/Dropbox/Forum/converted8bit.tga
/home/nils/Dropbox/Forum/converted8bit.tga TGA 128x128 128x128+0+0 8-bit sRGB 65.6KB 0.000u 0:00.000

Files available here - I'm on the wrong computer right now to provide the 16 bit file (iirc) I was playing with yesterday. I'm reading a bit into the code to see what options for 5?/15/16 added to 8/24/32 would require. It's a hurdle for me though.

Thanks
Nils

[dlemstra]

It is reporting 5 bit because the depth is per channel. 5*3 + 1 = 16

[nilsca]

Ah, yes, I'm mixing up bits_per_pixel and depth, thanks

[dlemstra]

I just added support for writing 16-bit TGA files. This will be available in the next release of ImageMagick (6.9.0-1). You can write it by specifying a channel depth of 5:

Code: Select all

convert logo: -depth 5 logo-16bit.tga

[nilsca]

Excellent, thank you for the quick turn-around. If you'd like I could test out the tga 5 bit depth support before the release - it's probably just a diff on tga.c required?

In any case, thanks again

[dlemstra]

You are correct, you only need to change tga.c. You can find my changes in revision 17275 of our SVN repository.

[nilsca]

Hi, built and tested and it works with one caveat and I'm not sure that's an issue or not.

If I start with an indexed png and convert it to 5 bit tga, all fine

https://www.dropbox.com/s/4p093jw4i0cd6 ... d.png?dl=0
https://www.dropbox.com/s/rm5i77zcdmz5k ... t.tga?dl=0

~/workspace/ImageMagick $ utilities/identify ~/Dropbox/test-indexed.png
~/Dropbox/test-indexed.png PNG 96x96 96x96+0+0 8-bit sRGB 149c 1.12KB 0.000u 0:00.000
~/workspace/ImageMagick $ utilities/identify ~/Dropbox/test-indexed-5bit.tga
~/Dropbox/test-indexed-5bit.tga TGA 96x96 96x96+0+0 5-bit sRGB 149c 9.53KB 0.000u 0:00.000

I can load this in Gimp (not quite the standard but hey).

If I start with an RGB png and convert it to 5 bit tga, Gimp won't show the picture anymore (again, might be gimp)

https://www.dropbox.com/s/ztprx92u2uls612/test.png?dl=0
https://www.dropbox.com/s/g26oaqdzsqu4c ... t.tga?dl=0


~/workspace/ImageMagick $ utilities/identify ~/Dropbox/test.png
~/Dropbox/test.png PNG 96x96 96x96+0+0 8-bit sRGB 1.58KB 0.000u 0:00.000
~/workspace/ImageMagick $ utilities/identify ~/Dropbox/test-5bit.tga
~/Dropbox/test-5bit.tga TGA 96x96 96x96+0+0 5-bit sRGB 36.9KB 0.000u 0:00.000

Converting both tgas back to png is fine for the indexed 5 bit, the other non-index(?) tga converts into a garbled picture.


Thanks
Nils

[nilsca]

.. one more thing while you're looking if I may - an option for enabling RLE compression explicitly would be completing things

Nighty
Nils

[dlemstra]

Are you sure you applied my patch (http://trac.imagemagick.org/changeset/17275) properly? Your file "test-5bit.tga" is incorrect but when I create it with my local build I get a correct tga file.

And you can already force RLE compression:

Code: Select all

convert test.png -depth 5 -compress rle test-5bit.tga

[nilsca]

on compress - ah, yes, I missed "convert -list compress", sorry