[FFmpeg-devel] [PATCH] libopenjpeg wrapper for jpeg2k decoding

Jai Menon jmenon86
Tue Feb 3 07:01:23 CET 2009


Hi,

On Tue, Feb 3, 2009 at 12:58 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
> On Mon, Feb 02, 2009 at 04:48:17PM +0530, Jai Menon wrote:
>> Hi,
>>
>> 2009/1/29 Michael Niedermayer <michaelni at gmx.at>:
>> > On Thu, Jan 29, 2009 at 11:35:31AM +0530, Jai Menon wrote:
>> >> Hi,
>> >>
>> >> On Thu, Jan 29, 2009 at 4:27 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
>> >> > On Wed, Jan 28, 2009 at 02:23:46PM +0530, Jai Menon wrote:
>>
>> [...]
>>
>> > dec is written to twice with no read in between, this applies to more vars
>>
>> fixed.
>>
>> [...]
>>
>> > sec hole when width/height changes
>>
>> fixed (i think).
>>
>> i also reworked most of the code to use get/release_buffer
>> revised patch attached.
> [...]
>> +static int libopenjpeg_decode_frame(AVCodecContext *avctx,
>> +                                    void *data, int *data_size,
>> +                                    const uint8_t *buf, int buf_size)
>> +{
>> +    LibOpenJPEGContext *ctx = avctx->priv_data;
>> +    AVFrame *picture = &ctx->image, *output = data;
>> +    opj_dinfo_t *dec;
>
>> +    opj_cio_t *stream = NULL;
>
> written twice before read

this and another one fixed.

>> +    opj_image_t *image = NULL;
>> +    int width, height, has_alpha = 0, ret = -1;
>> +    int x, y, index;
>> +    uint8_t *img_ptr;
>> +    float scale;
>> +
>> +    *data_size = 0;
>> +
>> +    // Check if input is a raw jpeg2k codestream or in jp2 wrapping
>> +    if((AV_RB32(buf) == 12) &&
>> +       (AV_RB32(buf + 4) == JP2_SIG_TYPE) &&
>> +       (AV_RB32(buf + 8) == JP2_SIG_VALUE)) {
>> +         dec = opj_create_decompress(CODEC_JP2);
>
>> +    }
>> +    else dec = opj_create_decompress(CODEC_J2K);
>
> diego likely wont like this kind of } placement style

changed.

> [...]
>> +    scale = 255.0f / (float)((1 << image->comps[0].prec) - 1);
>> +    for(y = 0; y < height; y++) {
>> +        index = y*width;
>> +        img_ptr = picture->data[0] + y*picture->linesize[0];
>> +        for(x = 0; x < width; x++, index++) {
>> +            *img_ptr++ = image->comps[0].data[index] * scale;
>> +            if(image->numcomps > 2 && check_image_attributes(image)) {
>> +                *img_ptr++ = image->comps[1].data[index] * scale;
>> +                *img_ptr++ = image->comps[2].data[index] * scale;
>> +                if(has_alpha)
>> +                    *img_ptr++ = image->comps[3].data[index] * scale;
>> +            }
>> +        }
>> +    }
>
> lrintf()

hmm...i made this generic and modified it to use integer arithmetic. i
think its better.
revised patch attached.

[...]

-- 
Regards,

Jai
-------------- next part --------------
A non-text attachment was scrubbed...
Name: libopenjpeg_wrapper.patch
Type: text/x-patch
Size: 12143 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090203/b45ef9f2/attachment.bin>



More information about the ffmpeg-devel mailing list