Re: Rendering grayscale on opengl using nsbitmaprep

[Fred Kiefer]

Sorry, I still don't understand your problem. Why are you converting the
image? The standard code should work fine in displaying it. Actually the
code I pointed you to is used for that.
>From your screen shot I would expect that your drawing code uses the
wrong sizes. To verify this it would help if you sent a bit more of your
code.

Fred

PS: For drawing pre-multiplied images get used. Better convert in to
that format. But this is completely unrelated to your problem.


Am 15.02.2011 20:34, schrieb indieAN:
> Hi,
>   I copied the source  of convertToFormatBitsPerSample and implemented
> in a category to NSBitmapImageRep. Now I am converting 
> all image format to a standard format using 
> 
>    bitmap = [bitmap convertToFormatBitsPerSample: 8
>                                      samplesPerPixel: 4
>                                             hasAlpha: YES
>                                             isPlanar: NO
>                                       colorSpaceName:
> NSDeviceRGBColorSpace
>                                         bitmapFormat:
> NSAlphaNonpremultipliedBitmapFormat 
>                                          bytesPerRow: 0
>                                         bitsPerPixel: 0];
> 
> 
> All other images (RBGA) renders great. But still no luck on this gray
> scale image. I am assuming that above code converts gray scale image to
> std RBGA format. So same rendering code should works (It is not included
> here ). I am attaching output that got from rendering.
> 
> Thanks to all,
> -ANIndie
>  
> 
> 
> On Mon, 2011-02-14 at 17:31 +0100, Fred Kiefer wrote:
>> I tried to render your image with a normal GNUstep test application and
>> it seems to display correctly. From that I would say that it gets loaded
>> by the PNG code correctly. This makes me wonder why you think that there
>> might be a bug in the grayscale rendering? As far as I see rendering
>> isn't involved in your case. More likely the bug is in your own
>> conversion code. I tried to have a quick look at your code, but could
>> not make much sense out of it. Why would you want to put in the width
>> and the hight separately from the image? As far as I know these values
>> might be completely off.
>> If you just want to learn how to convert between different image format,
>> have a look that the method
>>
>> - (NSBitmapImageRep *) _convertToFormatBitsPerSample: (int)bps
>>                                      samplesPerPixel: (int)spp
>>                                             hasAlpha: (BOOL)alpha
>>                                             isPlanar: (BOOL)isPlanar
>>                                       colorSpaceName:
>> (NSString*)colorSpaceName
>>                                         bitmapFormat:
>> (NSBitmapFormat)bitmapFormat
>>                                          bytesPerRow: (int)rowBytes
>>                                         bitsPerPixel: (int)pixelBits
>>
>> I added that to NSBitmapImage a few years ago just to document the
>> proper way of conversion. Even in GNUstep itself we had a lot of
>> incomplete image conversions.
>>
>> Cheers
>> Fred
>>
>>
>> Am 13.02.2011 18:40, schrieb Indie AN:
>>> Hi,
>>>  I had attached the image along this  mail.  Its  part of
>>> cocos2d-iphone project
>>> (https://github.com/ANindie/cocos2d-GNUstep/raw/1219df69bd64feb6aa2ae83073e1f9e380e6f9ad/src/Resources/Images/stars2-grayscale.png).
>>>   As per 'Gimp 'it contains 2 channels, alpha and luminosity. 64*64 pixels.
>>> GNUstep NSBitmapimagerep reports that samplesPerPixel=2,
>>> BytesPerRow=128. So I tried to use the data provided by
>>> NSBitmapimagerep  to generate RBGA pixel data. Either my approach
>>> going wrong or NSBitmapimagerep doesn't  give proper data.
>>>  Also I tried rendering image assuming BytesPerRow = 64*4 half of the
>>> image renders properly rest is black/garbage.
>>>
>>>
>>> On 2/12/11, Gregory Casamento <greg.casamento@gmail.com> wrote:
>>>> Which image?  I don't see an attachment or a link.
>>>>
>>>> On Thursday, February 10, 2011, indieAN <an.indian.indie@gmail.com> wrote:
>>>>> Hi,
>>>>>   I trying to render a gray scale image (contains 2 channels luminosity
>>>>> and alpha) using opengl. Actually this image is a png image. I am
>>>>> decoding it using nsbitmaprep. But only garbage  gets rendered. Can any
>>>>> body tell whats wrong with following code. Or their is bug in
>>>>> nsbitmapImagerep for rendering grayscale?
>>>>>
>>>>>
>>>>> -(id) initPremultipliedATextureWithBitmap:(NSBitmapImageRep*)bitmap
>>>>> pixelsWide:(NSUInteger)POTWide pixelsHigh:(NSUInteger)POTHigh
>>>>> {
>>>>>
>>>>>     short bytesPerPixel = [bitmap bitsPerPixel]>>3;//2
>>>>>         unsigned char * data = calloc(1, POTHigh  * POTWide * 4);
>>>>>         const unsigned char * originaldata = [bitmap bitmapData];
>>>>>
>>>>>
>>>>>
>>>>>     NSSize size=NSMakeSize([bitmap pixelsWide],[bitmap
>>>>> pixelsHigh]); //64*64
>>>>>     NSUInteger bytesPerRow = [bitmap bytesPerRow];//128
>>>>>     NSUInteger pixelPerRow =  bytesPerRow / bytesPerPixel;//64
>>>>>     CCTexture2DPixelFormat      pixelFormat;
>>>>>
>>>>>     NSUInteger inputPixelPerRow = POTWide * 4;
>>>>>
>>>>>
>>>>>
>>>>>      int spp = [bitmap samplesPerPixel];  //2
>>>>>
>>>>>
>>>>> #define FLIPPED_BITMAP 0
>>>>>      int i;
>>>>>
>>>>>      for (i = 0; i < size.height ; i++)
>>>>>     {
>>>>>
>>>>>
>>>>> #if FLIPPED_BITMAP
>>>>>      int vrow=(int)size.height -i-1;
>>>>> #else
>>>>>      int vrow=i;
>>>>> #endif
>>>>>
>>>>>
>>>>>       if(spp == 2)
>>>>>       {
>>>>>
>>>>>
>>>>>         unsigned char * dest =  data + (vrow * POTWide * 4);
>>>>>         const unsigned char * src =  originaldata + (i * bytesPerRow);
>>>>>
>>>>>         int k=0;
>>>>>         for(k=0; k<size.width ; k++)
>>>>>         {
>>>>>
>>>>>                 *(dest +k*4+3)=*(src+k*2+1); //alpha
>>>>>
>>>>>                  *(dest +k*4) = *(src+k*2+2);//r
>>>>>                  *(dest +k*4 + 1) = *(src+k*2);//g
>>>>>                  *(dest +k*4 + 2) = *(src+k*2);//b
>>>>>
>>>>>         }
>>>>>       }
>>>>>        else
>>>>>        memcpy(data + (vrow * POTWide * 4), originaldata + (i *
>>>>> bytesPerRow), bytesPerRow);
>>>>>    }
>>>>>
>>>>>
>>>>>
>>>>>         self = [self initWithData:data pixelFormat:pixelFormat
>>>>> pixelsWide:POTWide pixelsHigh:POTHigh contentSize:size
>>>>>                         unpackRowLength:POTWide];
>>>>>
>>>>>
>>>>>         return self;
>>>>> };
>