Jump to content

  • Log In with Google      Sign In   
  • Create Account

unsigned char <-> float


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
8 replies to this topic

#1 gnomgrol   Members   -  Reputation: 622

Like
0Likes
Like

Posted 11 July 2012 - 05:28 AM

Hy,

Im loading a bitmap from a file as unsigned char array. Then I convert it to floats, do some math with it, and then put it back into another bitmap.
Im using Visual C++ 2010 express, and when Im running the programm in Debugmode it works fine. But as soon as I put it into Releasemode, the output is only a completly black bitmap.
Does anyone have a clue what could cause this? I dont get any warnings at all (with high warninglevel).


That should be everything that is important:
struct f3{
   float r,g,b;
};

unsigned char * shadowMapImage;
f3 *shadowMap;

shadowMapImage = new unsigned char[mapWidth*mapWidth*3];
shadowMap = new f3[mapWidth*mapWidth*3];

LoadFromFile();   // Fills shadowMapImage with   fread(shadowMapImage, 1, imageSize, filePtr);


k=0;
for(int z = 0; z<mapWidth; z++){
		  for(int x = 0; x<mapWidth; x++){
				shadowMap[z*mapWidth+x].r = shadowMapImage[k];
				shadowMap[z*mapWidth+x].g = shadowMapImage[k+1];
				shadowMap[z*mapWidth+x].b = shadowMapImage[k+2];
		  k+=3;
	  }
	}

SomeMathFunctionThatNeedsFloats();


k=0;
for(int z = 0; z<mapWidth; z++){
		  for(int x = 0; x<mapWidth; x++){
				 shadowMapImage[(z*mapWidth+x)*3] = (unsigned char)shadowMap[(z*mapWidth+x)].r;
				 shadowMapImage[(z*mapWidth+x)*3+1] = (unsigned char)shadowMap[(z*mapWidth+x)].g;
				 shadowMapImage[(z*mapWidth+x)*3+2] = (unsigned char)shadowMap[(z*mapWidth+x)].b;
			  k+=3;
	  }
	}

SaveToFile();


Thanks

Edited by gnomgrol, 11 July 2012 - 06:38 AM.


Sponsor:

#2 doeme   Members   -  Reputation: 718

Like
1Likes
Like

Posted 11 July 2012 - 05:38 AM

WIthout any code to see I'm guessing that data structure or variable might not be initialized the right way or might not be initialized at all or that your conversion between unsigend char and floats is not ok. Could you post the code-snippet?

#3 wack   Members   -  Reputation: 1345

Like
0Likes
Like

Posted 11 July 2012 - 07:25 AM

Is the code that works on these floats a third party library?

I belive it's common for image libraries that work on floats to expect the value to be in the range 0..1

If this is the case, you would need to convert them so that 255 becomes 1.0: float f = (float)byteval / 255.0f

#4 gnomgrol   Members   -  Reputation: 622

Like
0Likes
Like

Posted 11 July 2012 - 07:30 AM

No, its only normal c++ code.

#5 Ryan_001   Prime Members   -  Reputation: 1456

Like
0Likes
Like

Posted 11 July 2012 - 07:42 AM

shadowMap = new f3[mapWidth*mapWidth*3];
shouldn't that be:
shadowMap = new f3[mapWidth*mapWidth];
since f3 has the rgb already in it?

Also your second loop should probably read:
shadowMapImage[k] = (unsigned char)shadowMap[(z*mapWidth+x)].r;
shadowMapImage[k+1] = (unsigned char)shadowMap[(z*mapWidth+x)].g;
shadowMapImage[k+2] = (unsigned char)shadowMap[(z*mapWidth+x)].b;

None-the-less I still don't see that accounting for the errors. As was pointed out, usually converting from a float to an unsigned char you need to multiply by 255, and divide by 255 when going from unsigned char to float, as usually the float's are treated as normalized. You did neither of these but it still seemed to work in debug mode so perhaps this wasn't necessary in your case. Its hard to say really given what little code is there.

Edited by Ryan_001, 11 July 2012 - 07:49 AM.


#6 gnomgrol   Members   -  Reputation: 622

Like
0Likes
Like

Posted 11 July 2012 - 08:31 AM

The only thing the Mathfunction really does is change the values in shadowMapImage with some +-. There really is no more code to be concerned about.
I dont normalize the floats because I need them as they are, from 0 to 255.

Edited by gnomgrol, 11 July 2012 - 08:33 AM.


#7 doeme   Members   -  Reputation: 718

Like
0Likes
Like

Posted 11 July 2012 - 08:51 AM

My first guess would also have been a normalization-issue, but if that is not the cause I guess the best course of action is to narrow down the point of failure. Is the image loaded correctly? Does it look ok, if you save it without any modification?

#8 Álvaro   Crossbones+   -  Reputation: 13897

Like
0Likes
Like

Posted 11 July 2012 - 09:03 AM

In situations like this you should try to come up with a minimal program that still shows the problem. In the process of simplifying your program, you are likely to figure out what the problem was. Even if you don't, you'll have something to post here so we can test it ourselves.

#9 gnomgrol   Members   -  Reputation: 622

Like
0Likes
Like

Posted 11 July 2012 - 09:54 AM

I think I got the problem. Its kinda wierd.
It HAD to do with the Mathfunction, but why it worked in debugging mode is a miracle.
Actually, the numbers were calculated invert (means 255 was 0, 0 was 255). But where should have been 0, there was 211. I cant tell why.
Breaking it into smaller pieces really helped, thanks!




Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS