theora - Nov 2005 - theoraToRGB, and a problem with it

hi folks!

maybe you can remember, months ago i wanted to know about YUV to RGB
conversion in the theora case.

finally i managed to go on with this issue, and here is a function i
made which needs a
theora_info and a yuv_buffer passed and will create an interleaved 24
bit RGB buffer.

just one big problem: i am coding since days, and the only remaining
problem which i cannot seem to solute is that in the last part, the
conversion from interleaved YUV to interleaved RGB seems to have
problems. theoretically (and when tried external with input and output
files it works perfectly) it should work.
but when dumping rgbbuf with fwrite(), the output is completely empty :(

maybe someone of you can help with this :)

greets,

---david

code:

#include <stdio.h>
#include "theora/theora.h"

/*
* theora2rgb 0.1
*
* Converts a yuv_buffer filled by theora_decode_YUVout to
* a 24bit RGB buffer, ordered R-G-B, interleaved.
* Currently, the color conversion is not very efficient.
* In addition, encoder specified color is ignored, resulting colors
* are a little red-weighted, and no resolution scaling is done.
*/

#define clip(a) (((a)&256) ? ((a)>>31)^(-1) : (a))

void theoraToRGB(theora_info ti, yuv_buffer yuv) {

  // create yuvbuf in interleaved YUV format.

  int i, j, crop_offset = 0;

  char* yuvbuf = (char *)malloc(ti.width * ti.height * 3);
  char* rgbbuf = (char *)malloc(ti.width * ti.height * 3);

  crop_offset=ti.offset_x+yuv.y_stride*ti.offset_y;

  int k=0;
  i = 0;
  while(i<ti.height) {
    for(j=0;j<ti.width;j++) {
      yuvbuf[k]=yuv.y[crop_offset+yuv.y_stride*i+j];
      k += 3;
    }
    i++;
  }

  crop_offset=(ti.offset_x/2)+(yuv.uv_stride)*(ti.offset_y/2);

  k=1;
  i=0;
  while(i<ti.height/2) {
    for(j=0;j<ti.width/2;j++) {
      yuvbuf[k++]=yuv.u[crop_offset+yuv.uv_stride*i+j];
      yuvbuf[k++]=yuv.v[crop_offset+yuv.uv_stride*i+j];
      k++;
      yuvbuf[k++]=yuv.u[crop_offset+yuv.uv_stride*i+j];
      yuvbuf[k++]=yuv.v[crop_offset+yuv.uv_stride*i+j];
      k++;
    }
    for(j=0;j<ti.width/2;j++) {
      yuvbuf[k++]=yuv.u[crop_offset+yuv.uv_stride*i+j];
      yuvbuf[k++]=yuv.v[crop_offset+yuv.uv_stride*i+j];
      k++;
      yuvbuf[k++]=yuv.u[crop_offset+yuv.uv_stride*i+j];
      yuvbuf[k++]=yuv.v[crop_offset+yuv.uv_stride*i+j];
      k++;
    }
    i++;
  }

  // convert yuvbuf to rgbbuf, interleaved RGB.

  int offset, y, u, v = 0;

  while(offset < ti.width * ti.height * 3) {

    y = yuvbuf[offset++] - 16;
    u = yuvbuf[offset++] - 128;
    v = yuvbuf[offset++] - 128;

    offset -= 3;

    rgbbuf[offset++] = clip((298 * y + 409 * v + 128) >> 8);
    rgbbuf[offset++] = clip((298 * y + 100 * u - 208 * v + 128) >> 8);
    rgbbuf[offset++] = clip((298 * y + 516 * u + 128) >> 8);
  }

  // just for testing: dump RGB data to RAW file

  FILE * ofp = fopen("rgb.raw", "wb");
  fwrite(rgbbuf, sizeof(char), ti.width * ti.height * 3, ofp);
/*
when using
  fwrite(yuvbuf, sizeof(char), ti.width * ti.height * 3, ofp);
instead, you can see that the YUV buffer is perfectly correct.
*/

  fclose(ofp);

  free(yuvbuf);
  free(rgbbuf);
}

>   int offset, y, u, v = 0;
int offset = 0, y, u, v = 0;

--
Makoto SETOH

Makoto SETOH schrieb:
>>  int offset, y, u, v = 0;
>>    
>>
>
>int offset = 0, y, u, v = 0;
>
>--
>Makoto SETOH
>
>
>  
>
hmm, thanks a lot. i always thought my way was correct, but now it seems 
definitely wrong.

ok, now that is fixed. but the problem which i am fighting with since 
weeks is still present.
the RGB output looks weird, like if a color channel is missing.

but when using the conversion in some external program, with infile and 
outfile, it works perfectly (see code on bottom).

this problem just occurs when working with 2 buffers instead of 2 
files.... :(

demo code:

#include <stdio.h>

#define clip(a) (((a)&256) ? ((a)>>31)^(-1) : (a))

int main(void) {

  FILE* ifp = fopen("yuv.raw", "rb");
  FILE* ofp = fopen("rgb.raw", "ab");

  int offset = 0, y, u, v;

  while(offset < 512 * 384 * 3) {

    y = fgetc(ifp) - 16;
    u = fgetc(ifp) - 128;
    v = fgetc(ifp) - 128;

    fputc(clip((298 * y + 409 * v + 128) >> 8), ofp);
    fputc(clip((298 * y + 100 * u - 208 * v + 128) >> 8), ofp);
    fputc(clip((298 * y + 516 * u + 128) >> 8), ofp);

    offset += 3;
  }

  fclose(ifp);
  fclose(ofp);

  return 0;
}