[Ffmpeg-devel] [PATCH] BMP encoder

[Michel Bardiaux]

Michael Niedermayer wrote:
> Hi
> 
> On Mon, Oct 30, 2006 at 05:36:24PM +0100, Michel Bardiaux wrote:
>> Michael Niedermayer wrote:
>>> Hi
>>>
>>> On Mon, Oct 30, 2006 at 04:05:39PM +0100, Michel Bardiaux wrote:
>>>> Currently, only straight RGB.
>>> [...]
>>>> +    init_put_bits(&s->pb, buf, buf_size);
>>>> +    // STRUCTURE.field refer to the MSVC documentation
>>>> +    // BITMAPFILEHEADER.bfType
>>>> +    put_bits(&s->pb, 8, 'B');
>>>> +    put_bits(&s->pb, 8, 'M');
>>>> +    // BITMAPFILEHEADER.bfSize
>>>> +    write32(&s->pb, n_bytes);
>>>> +    // BITMAPFILEHEADER.bfReserved1
>>>> +    write16(&s->pb, 0);
>>>> +    // BITMAPFILEHEADER.bfReserved2
>>>> +    write16(&s->pb, 0);
>>>> +    // BITMAPFILEHEADER.bfOffBits
>>>> +    write32(&s->pb, 14 /* BITMAPFILEHEADER */ + 40);
>>>> +    // BITMAPINFOHEADER.biSize
>>>> +    write32(&s->pb, 40);
>>>> +    // BITMAPINFOHEADER.biWidth
>>>> +    write32(&s->pb, avctx->width);
>>>> +    // BITMAPINFOHEADER.biHeight
>>>> +    write32(&s->pb, avctx->height);
>>>> +    // BITMAPINFOHEADER.biPlanes
>>>> +    write16(&s->pb, 1);
>>>> +    // BITMAPINFOHEADER.biBitCount
>>>> +    write16(&s->pb, 24);
>>>> +    // BITMAPINFOHEADER.biCompression
>>>> +    write32(&s->pb, BMP_RGB);
>>>> +    // BITMAPINFOHEADER.biSizeImage
>>>> +    write32(&s->pb, n_bytes_image);
>>>> +    // BITMAPINFOHEADER.biXPelsPerMeter
>>>> +    write32(&s->pb, 0);
>>>> +    // BITMAPINFOHEADER.biYPelsPerMeter
>>>> +    write32(&s->pb, 0);
>>>> +    // BITMAPINFOHEADER.biClrUsed
>>>> +    write32(&s->pb, 0);
>>>> +    // BITMAPINFOHEADER.biClrImportant
>>> see put_bmp_header()
>> Right, I have recoded the wheel. But I cant use it unchanged; first 
>> because it is in libavformat, second because it seems intimately linked 
>> to asf and avi.
>>
>> I see 2 ways: recode it in the bmp encoder; or define our own bitmap 
>> structures to separate setting the values and writing them. For the time 
>> being I will take the 1st way.
> 
> hmm ok but please at least dont interleave the write32() and the comments
> this is confusing, rather put the comments to the right of the write32() and
> nicely aligned ...
> 
> 
>>>
>>>> +    write32(&s->pb, 0);
>>>> +    // BMP files are bottom-to-top
>>>> +    ptr = p->data[0] + (avctx->height - 1) * p->linesize[0];
>>>> +    linesize = -p->linesize[0];
>>>> +    for(i = 0; i < avctx->height; i++){
>>>> +        uint8_t *src = ptr;
>>>> +        int n = 0;
>>>> +        for(j=0;j<avctx->width;j++) {
>>>> +            put_bits(&s->pb, 8, src[0]);
>>>> +            put_bits(&s->pb, 8, src[1]);
>>>> +            put_bits(&s->pb, 8, src[2]);
>>> using the bitstream reader for a purely byte based format is unacceptable
>>> (reason is that its several times slower and more complex)
>> Right. Will change to use put_le16 etc. But see above.
> 
> to put 8bit a simple uint8_t will do ...
> for 16bit a simple uint16_t with le2me_16() will do as long as its
> aligned and i think it is ...
> same for 32bit
> if its not aligned then there are ST16() ST32() in dsputil.h, they could
> be moved into a bytesomething.h (in a seperate patch ...)

Here it is

> 
> 
> [...]


-- 
Michel Bardiaux
R&D Director
T +32 [0] 2 790 29 41
F +32 [0] 2 790 29 02
E mailto:mbardiaux at mediaxim.be

Mediaxim NV/SA
Vorstlaan 191 Boulevard du Souverain
Brussel 1160 Bruxelles
http://www.mediaxim.com/
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: bytestream.pat
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20061030/1f51ff0d/attachment.asc>