53                                    int order, 
float cutoff_ratio,
 
   62                "low-pass filter mode\n");
 
   67                "even filter orders\n");
 
   71     wa = 2 * tan(
M_PI * 0.5 * cutoff_ratio);
 
   74     for(i = 1; i < (order >> 1) + 1; i++)
 
   75         c->
cx[i] = c->
cx[i - 1] * (order - i + 1LL) / i;
 
   79     for(i = 1; i <= order; i++)
 
   80         p[i][0] = p[i][1] = 0.0;
 
   81     for(i = 0; i < order; i++){
 
   83         double th = (i + (order >> 1) + 0.5) * 
M_PI / order;
 
   84         double a_re, a_im, c_re, c_im;
 
   91         zp[0] = (a_re * c_re + a_im * c_im) / (c_re * c_re + c_im * c_im);
 
   92         zp[1] = (a_im * c_re - a_re * c_im) / (c_re * c_re + c_im * c_im);
 
   94         for(j = order; j >= 1; j--)
 
   98             p[j][0] = a_re*zp[0] - a_im*zp[1] + p[j-1][0];
 
   99             p[j][1] = a_re*zp[1] + a_im*zp[0] + p[j-1][1];
 
  101         a_re    = p[0][0]*zp[0] - p[0][1]*zp[1];
 
  102         p[0][1] = p[0][0]*zp[1] + p[0][1]*zp[0];
 
  105     c->
gain = p[order][0];
 
  106     for(i = 0; i < order; i++){
 
  108         c->
cy[i] = (-p[i][0] * p[order][0] + -p[i][1] * p[order][1]) /
 
  109                    (p[order][0] * p[order][0] + p[order][1] * p[order][1]);
 
  111     c->
gain /= 1 << order;
 
  118                               float cutoff_ratio, 
float stopband)
 
  120     double cos_w0, sin_w0;
 
  126                "high-pass and low-pass filter modes\n");
 
  134     cos_w0 = cos(
M_PI * cutoff_ratio);
 
  135     sin_w0 = sin(
M_PI * cutoff_ratio);
 
  137     a0 = 1.0 + (sin_w0 / 2.0);
 
  140         c->
gain  =  ((1.0 + cos_w0) / 2.0)  / 
a0;
 
  141         x0       =  ((1.0 + cos_w0) / 2.0)  / 
a0;
 
  142         x1       = (-(1.0 + cos_w0))        / 
a0;
 
  144         c->
gain  =  ((1.0 - cos_w0) / 2.0)  / 
a0;
 
  145         x0       =  ((1.0 - cos_w0) / 2.0)  / 
a0;
 
  146         x1       =   (1.0 - cos_w0)         / a0;
 
  148     c->
cy[0] = (-1.0 + (sin_w0 / 2.0)) / a0;
 
  149     c->
cy[1] =  (2.0 *  cos_w0)        / a0;
 
  162                                                 int order, 
float cutoff_ratio,
 
  163                                                 float stopband, 
float ripple)
 
  168     if (order <= 0 || order > 
MAXORDER || cutoff_ratio >= 1.0)
 
  207 #define CONV_S16(dest, source) dest = av_clip_int16(lrintf(source)); 
  209 #define CONV_FLT(dest, source) dest = source; 
  211 #define FILTER_BW_O4_1(i0, i1, i2, i3, fmt)         \ 
  212     in = *src0 * c->gain                            \ 
  213          + c->cy[0]*s->x[i0] + c->cy[1]*s->x[i1]    \ 
  214          + c->cy[2]*s->x[i2] + c->cy[3]*s->x[i3];   \ 
  215     res =  (s->x[i0] + in      )*1                  \ 
  216          + (s->x[i1] + s->x[i3])*4                  \ 
  218     CONV_##fmt(*dst0, res)                          \ 
  223 #define FILTER_BW_O4(type, fmt) {           \ 
  225     const type *src0 = src;                 \ 
  227     for (i = 0; i < size; i += 4) {         \ 
  229         FILTER_BW_O4_1(0, 1, 2, 3, fmt);    \ 
  230         FILTER_BW_O4_1(1, 2, 3, 0, fmt);    \ 
  231         FILTER_BW_O4_1(2, 3, 0, 1, fmt);    \ 
  232         FILTER_BW_O4_1(3, 0, 1, 2, fmt);    \ 
  236 #define FILTER_DIRECT_FORM_II(type, fmt) {                                  \ 
  238     const type *src0 = src;                                                 \ 
  240     for (i = 0; i < size; i++) {                                            \ 
  243         in = *src0 * c->gain;                                               \ 
  244         for(j = 0; j < c->order; j++)                                       \ 
  245             in += c->cy[j] * s->x[j];                                       \ 
  246         res = s->x[0] + in + s->x[c->order >> 1] * c->cx[c->order >> 1];    \ 
  247         for(j = 1; j < c->order >> 1; j++)                                  \ 
  248             res += (s->x[j] + s->x[c->order - j]) * c->cx[j];               \ 
  249         for(j = 0; j < c->order - 1; j++)                                   \ 
  250             s->x[j] = s->x[j + 1];                                          \ 
  251         CONV_##fmt(*dst0, res)                                              \ 
  252         s->x[c->order - 1] = in;                                            \ 
  258 #define FILTER_O2(type, fmt) {                                              \ 
  260     const type *src0 = src;                                                 \ 
  262     for (i = 0; i < size; i++) {                                            \ 
  263         float in = *src0   * c->gain  +                                     \ 
  264                    s->x[0] * c->cy[0] +                                     \ 
  265                    s->x[1] * c->cy[1];                                      \ 
  266         CONV_##fmt(*dst0, s->x[0] + in + s->x[1] * c->cx[1])                \ 
  276                    const int16_t *src, 
int sstep, int16_t *
dst, 
int dstep)
 
  280     } 
else if (c->
order == 4) {
 
  289                        const float *src, 
int sstep, 
float *
dst, 
int dstep)
 
  293     } 
else if (c->
order == 4) {
 
  323     float cutoff_coeff = 0.4;
 
  329                                         cutoff_coeff, 0.0, 0.0);
 
  332     for (i = 0; i < 
SIZE; i++) {
 
  333         x[i] = 
lrint(0.75 * INT16_MAX * sin(0.5*
M_PI*i*i/SIZE));
 
  338     for (i = 0; i < 
SIZE; i++)
 
  339         printf(
"%6d %6d\n", x[i], y[i]);