We closed this forum 18 June 2010. It has served us well since 2005 as the ALPHA forum did before it from 2002 to 2005. New discussions are ongoing at the new URL http://forum.processing.org. You'll need to sign up and get a new user account. We're sorry about that inconvenience, but we think it's better in the long run. The content on this forum will remain online.
Index › Programming Questions & Help › Programs › RLE compression for TGA output
‹ Previous Topic | Next Topic ›
Page Index Toggle Pages: 1
RLE compression for TGA output (Read 2146 times)
jtarbell
RLE compression for TGA output
Jun 8th, 2005, 7:57am  
Is it possible to enable RLE compression for TGA (TARGA) images saved with the save() or saveFrame() methods?

thanks,
jared
Back to top
 
WWW  
fry
Re: RLE compression for TGA output
Reply #1 - Jun 8th, 2005, 5:09pm  
not currently. the java 1.4 imageio stuff would probably do it (or write to another compressed format), but would require a little hacking around to get the pixels and write to the file. i think elsewhere on the board there was info about this (or maybe on the alpha board).
Back to top
 
WWW  
toxi
Re: RLE compression for TGA output
Reply #2 - Jun 8th, 2005, 5:36pm  
RLE compression, huh! Smiley after a quick consultation of this Targa spec i modified the code to use this simple type of compression. because the maximum chunk length is only 128 pixels, don't expect a huge compression ratio even  if your image uses large regions with flat colours:

Code:
void setup() {
  size(200,200);
  PImage img=loadImage("ndsm.jpg");
  println(saveTGA_RLE(img,"tga0.tga"));
}

boolean saveTGA_RLE(PImage img, String fileName) {
  try {
    int[] pixels=img.pixels;
    int width=img.width;
    int height=img.height;
    OutputStream output = new BufferedOutputStream(
	new FileOutputStream(fileName),
	32768
    );

    byte header[] = new byte[18];

    // set file header info (18 bytes)

    // RGB + RLE Compression
    header[2] = 0x02 + 0x08;
    header[12] = (byte) (width & 0xff);
    header[13] = (byte) (width >> 8);
    header[14] = (byte) (height & 0xff);
    header[15] = (byte) (height >> 8);
    header[16] = 32; // bits per pixel
    header[17] = 0x08 + 0x20; // bits per colour component + origin at top-left

    output.write(header);

    int maxLen = height * width;
    int index = 0;
    int col, prevCol;
    int[] currChunk = new int[128];

    while (index < maxLen) {
	boolean isRLE = false;
	currChunk[0] = col = pixels[index];
	int rle = 1;
	// try to find repeating bytes (min. len = 2 pixels)
	// maximum chunk size is 128 pixels
	while (index + rle < maxLen) {
	  if (col != pixels[index + rle] || rle == 128) {
	    isRLE = (rle > 1); // set flag for RLE chunk
	    break;
	  }
	  rle++;
	}
	// write compressed chunk (RLE+colour)
	if (isRLE) {
	  output.write(128 | (rle - 1));
	  output.write(col & 0xff);
	  output.write(col >> 8 & 0xff);
	  output.write(col >> 16 & 0xff);
	  output.write(col >>> 24 & 0xff);
	} 
	else {
	  // build uncompressed chunk
	  rle = 1; //reset scan offset
	  while (index + rle < maxLen) {
	    if ((col != pixels[index + rle] && rle < 128) || rle < 3) {
		currChunk[rle] = col = pixels[index + rle];
	    } 
	    else {
		// check if the exit condition was the start of a repeating colour 
		if (col == pixels[index + rle]) rle -= 2;
		break;
	    }
	    rle++;
	  }
	  // write uncompressed chunk
	  output.write(rle - 1);
	  for (int i = 0; i < rle; i++) {
	    col = currChunk[i];
	    output.write(col & 0xff);
	    output.write(col >> 8 & 0xff);
	    output.write(col >> 16 & 0xff);
	    output.write(col >>> 24 & 0xff);
	  }
	}
	// update scan offset
	index += rle;
    }
    output.flush();
    return true;
  } 
  catch (IOException e) {
    e.printStackTrace();
    return false;
  }
}


will write to ben about updating this directly in PImage...
Back to top
 
postspectacular.com | toxi.co.uk
WWW  
fry
Re: RLE compression for TGA output
Reply #3 - Jun 8th, 2005, 6:35pm  
bless you toxi. i dug up the tga spec to take a look and found some general code to do it, but that's as far as i got.

any downside to this being the default when saving .tga?
Back to top
 
WWW  
toxi
Re: RLE compression for TGA output
Reply #4 - Jun 8th, 2005, 7:07pm  
no real downside i know of. it is marginally slower, but the saving in terms of filesize can be quite big (depending on how "busy" the image is). for example the image below gets compressed to 61% (385KB vs. 640K) i have tested reading the exported images with photoshop and QT and they're doing just fine. i'll send you the updated PImage in a minute and you can take it from there! Smiley

...
Back to top
 
postspectacular.com | toxi.co.uk
WWW  
fry
Re: RLE compression for TGA output
Reply #5 - Jun 8th, 2005, 11:13pm  
now integrated for rev 92.. toxi is the hero.
Back to top
 
WWW  
scloopy
Re: RLE compression for TGA output
Reply #6 - Jun 9th, 2005, 1:56am  
Props to toxi! My hard drive is gonna thank you!
Back to top
 
WWW  
jtarbell
Re: RLE compression for TGA output
Reply #7 - Jun 9th, 2005, 11:01pm  
Incredible!  Thank you so much Toxi.

jared
Back to top
 
WWW  
toxi
Re: RLE compression for TGA output
Reply #8 - Jun 10th, 2005, 1:07pm  
pleasure! btw. if you're out to save even more filesize and don't require the alpha channel, apply the patches below to only save 24bit images (and so saving yet another 25%)

1) change the lines:
Code:
header[16] = 32; // bits per pixel
header[17] = 0x08 + 0x20;

into:
Code:
header[16] = 24; // bits per pixel
header[17] = 0x00 + 0x20; // no alpha channel bits

2) remove/comment out both instances of this line:
Code:
output.write(col >>> 24 & 0xff);

good luck!
Back to top
 
postspectacular.com | toxi.co.uk
WWW  
fry
Re: RLE compression for TGA output
Reply #9 - Jun 10th, 2005, 4:30pm  
now integrated into PImage so it'll be all set for 92 (or anyone who's using the cvs version)
Back to top
 
WWW  
fry
Re: RLE compression for TGA output
Reply #10 - Jun 10th, 2005, 4:41pm  
in fact, is there a simple way to do 8 bit so we can save ALPHA format images too (or save things as grayscale?)
Back to top
 
WWW  
toxi
Re: RLE compression for TGA output
Reply #11 - Jun 10th, 2005, 4:44pm  
inspecting now... shouldn't be too hard for grayscale, but will have to pass for proper colour quantization and palette creation!
Back to top
 
postspectacular.com | toxi.co.uk
WWW  
fry
Re: RLE compression for TGA output
Reply #12 - Jun 10th, 2005, 6:21pm  
oh no, nothing that complex.. i just mean when we have a single channel image, is it just a matter of a couple of bits? let's make the assumption that the data is already 8 bit (grayscale or an alpha channel, say) can that be written just as simply?

it was simply that i realized that we have 3 image types, and right now the save() command in PImage is assuming everything is ARGB, so you're gonna get some weird results if you (for some reason) try to save an ALPHA format image, or try to write an RGB image that doesn't have its alpha channel set opaque (often the case). so the 24 bit patch fixes that part, was just wondering if there's a similar 8 bit patch.
Back to top
 
WWW  
toxi
Re: RLE compression for TGA output
Reply #13 - Jun 10th, 2005, 7:54pm  
understood! though i'm not sure if always exporting w/o alpha channel is quite right. the nice thing about Targa is that you did get hold of the alpha... i just found another more detailed format spec and there's another mode mentioned, specifically intended for 8bit grayscale. yeah!

so here's an updated version which takes into account the format of a PImage object and uses the appropriate bit depth to write the TGA file:

Code:
void setup() {
  size(400,400);
  background(255,0,128);
  for(int x=0; x<width; x+=5) {
    stroke(255.0*x/400.0);
    line(x,0,width,height);
    line(0,height,width-x,height-x);
  }
  loadPixels();
  // try one of those...
  g.format=ARGB;
  //g.format=RGB;
  //g.format=ALPHA;
  saveTGA_RLE(g,"targa_test.tga");
}

// new version to takes image format setting into account
// ALPHA images written as 8bit grayscale (uses lowest byte)
// RGB -> 24bits
// RGBA -> 32bits
// all versions are RLE compressed

boolean saveTGA_RLE(PImage img, String fileName) {
  try {
    int[] pixels=img.pixels;
    int width=img.width;
    int height=img.height;
    OutputStream output = new BufferedOutputStream(
    new FileOutputStream(fileName),
    32768
	);

    byte header[] = new byte[18];

    // save ALPHA images as 8bit grayscale
    if (ALPHA==img.format) {
	header[2] = 0x0B;
	header[16] = 0x08;
	header[17] = 0x28;
    } 
    else {
	header[2] = 0x0A;
	// use 32bit colour for ARGB
	// else 24bit for RGB formatted images
	if (ARGB==img.format) {
	  header[16] = 32;
	  header[17] = 0x28;
	} 
	else {
	  header[16] = 24;
	  header[17] = 0x20;
	}
    }
    // set image dimensions lo-hi byte order
    header[12] = (byte) (width & 0xff);
    header[13] = (byte) (width >> 8);
    header[14] = (byte) (height & 0xff);
    header[15] = (byte) (height >> 8);

    output.write(header);

    int maxLen = height * width;
    int index = 0;
    int col, prevCol;
    int[] currChunk = new int[128];

    // 8bit image exporter is in separate loop
    // to avoid excessive conditionals...
    if (ALPHA==img.format) {
	while (index < maxLen) {
	  boolean isRLE = false;
	  int rle = 1;
	  currChunk[0] = col = pixels[index] & 0xff;
	  while (index + rle < maxLen) {
	    if (col != (pixels[index + rle]&0xff) || rle == 128) {
		isRLE = (rle > 1);
		break;
	    }
	    rle++;
	  }
	  if (isRLE) {
	    output.write(0x80 | (rle - 1));
	    output.write(col);
	  } 
	  else {
	    rle = 1;
	    while (index + rle < maxLen) {
		int cscan=pixels[index + rle]&0xff;
		if ((col != cscan && rle < 128) || rle < 3) {
		  currChunk[rle] = col = cscan;
		} 
		else {
		  if (col == cscan) rle -= 2;
		  break;
		}
		rle++;
	    }
	    output.write(rle - 1);
	    for (int i = 0; i < rle; i++) output.write(currChunk[i]);
	  }
	  index += rle;
	}
    } 
    else {
	// export 24/32 bit TARGA
	while (index < maxLen) {
	  boolean isRLE = false;
	  currChunk[0] = col = pixels[index];
	  int rle = 1;
	  // try to find repeating bytes (min. len = 2 pixels)
	  // maximum chunk size is 128 pixels
	  while (index + rle < maxLen) {
	    if (col != pixels[index + rle] || rle == 128) {
		isRLE = (rle > 1); // set flag for RLE chunk
		break;
	    }
	    rle++;
	  }
	  if (isRLE) {
	    output.write(128 | (rle - 1));
	    output.write(col & 0xff);
	    output.write(col >> 8 & 0xff);
	    output.write(col >> 16 & 0xff);
	    if (ARGB==img.format) output.write(col >>> 24 & 0xff);
	  }  
	  else {
	    rle = 1;
	    while (index + rle < maxLen) {
		if ((col != pixels[index + rle] && rle < 128) || rle < 3) {
		  currChunk[rle] = col = pixels[index + rle];
		}  
		else {
		  // check if the exit condition was the start of a repeating colour  
		  if (col == pixels[index + rle]) rle -= 2;
		  break;
		}
		rle++;
	    }
	    // write uncompressed chunk
	    output.write(rle - 1);
	    if (ARGB==img.format) {
		for (int i = 0; i < rle; i++) {
		  col = currChunk[i];
		  output.write(col & 0xff);
		  output.write(col >> 8 & 0xff);
		  output.write(col >> 16 & 0xff);
		  output.write(col >>> 24 & 0xff);
		}
	    } 
	    else {
		for (int i = 0; i < rle; i++) {
		  col = currChunk[i];
		  output.write(col & 0xff);
		  output.write(col >> 8 & 0xff);
		  output.write(col >> 16 & 0xff);
		}
	    }
	  }
	  index += rle;
	} 
    }

    output.flush();
    return true;
  } 
  catch (IOException e) {
    e.printStackTrace();
    return false;
  }
}


good weekend all! :)
Back to top
 
postspectacular.com | toxi.co.uk
WWW  
fry
Re: RLE compression for TGA output
Reply #14 - Jun 10th, 2005, 9:45pm  
right, in cvs i had it flipping between RGB and ARGB as necessary, but in doing so had realized that we don't have a good option for single channel (format == ALPHA, or maybe grayscale if we ever support that).

but hey, this is even easier and i'll just apply it. Wink
Back to top
 
WWW  
Page Index Toggle Pages: 1
‹ Previous Topic | Next Topic ›