[FFmpeg-devel] [PATCH 1/2] Add biquad high-pass and low-pass IIR filters.

Ronald S. Bultje rsbultje
Fri Jan 21 00:44:06 CET 2011


Hi,

On Thu, Jan 20, 2011 at 4:26 PM, Justin Ruggles
<justin.ruggles at gmail.com> wrote:
> ---
> ?libavcodec/iirfilter.c | ? 50 ++++++++++++++++++++++++++++++++++++++++++++++++
> ?libavcodec/iirfilter.h | ? ?1 +
> ?2 files changed, 51 insertions(+), 0 deletions(-)
[..]
> +    a0 = 1.0 + (sin_w0 / 2.0);
> +
> +    if (filt_mode == FF_FILTER_MODE_HIGHPASS) {
> +        c->gain  =  ((1.0 + cos_w0) / 2.0)  / a0;
> +        x0       = (-(1.0 + cos_w0))        / a0;
> +        x1       =  ((1.0 + cos_w0) / 2.0)  / a0;
> +    } else { // FF_FILTER_MODE_LOWPASS
> +        c->gain  =  ((1.0 - cos_w0) / 2.0)  / a0;
> +        x0       =   (1.0 - cos_w0)         / a0;
> +        x1       =  ((1.0 - cos_w0) / 2.0)  / a0;
> +    }
> +    c->cy[0] =  (2.0 *  cos_w0)        / a0;
> +    c->cy[1] = (-1.0 + (sin_w0 / 2.0)) / a0;
> +
> +    // divide by gain to make the x coeffs integers.
> +    // during filtering, the delay state will include the gain multiplication
> +    c->cx[0] = lrintf(x0 / c->gain);
> +    c->cx[1] = lrintf(x1 / c->gain);
> +    c->cy[0] /= c->gain;
> +    c->cy[1] /= c->gain;

Patch is fine. At some point you should do igain=1/c->gain and then
change div->mul, same for a0, make a0=1/a0; and then div->mul. But
let's not halt patches for micro-optimizations in init funcs... So OK.

Ronald



More information about the ffmpeg-devel mailing list