virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-15 14:13:16 +02:00
Code Issues Releases Activity

avfilter/avf_showcwt: fix truncation of low Hz kernels

Browse Source
This commit is contained in:
Paul B Mahol
2023-07-29 17:11:28 +02:00
parent c0f867bf50
commit 37d3000e3e
1 changed files with 32 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
libavfilter/avf_showcwt.c
View File

@@ -583,7 +583,8 @@ static int run_channel_cwt(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo
AVComplexFloat *isrc = (AVComplexFloat *)s->ifft_in->extended_data[jobnr]; AVComplexFloat *isrc = (AVComplexFloat *)s->ifft_in->extended_data[jobnr];
AVComplexFloat *idst = (AVComplexFloat *)s->ifft_out->extended_data[jobnr]; AVComplexFloat *idst = (AVComplexFloat *)s->ifft_out->extended_data[jobnr];
const int output_padding_size = s->output_padding_size; const int output_padding_size = s->output_padding_size;
const float scale = 1.f / s->input_padding_size; const int input_padding_size = s->input_padding_size;
const float scale = 1.f / input_padding_size;
const int ihop_size = s->ihop_size; const int ihop_size = s->ihop_size;
const int count = s->frequency_band_count; const int count = s->frequency_band_count;
const int start = (count * jobnr) / nb_jobs; const int start = (count * jobnr) / nb_jobs;
@@ -599,20 +600,37 @@ static int run_channel_cwt(AVFilterContext *ctx, void *arg, int jobnr, int nb_jo
const int kernel_start = s->kernel_start[y]; const int kernel_start = s->kernel_start[y];
const int kernel_stop = s->kernel_stop[y]; const int kernel_stop = s->kernel_stop[y];
const int kernel_range = kernel_stop - kernel_start + 1; const int kernel_range = kernel_stop - kernel_start + 1;
int offset;
if (kernel_start >= 0) {
offset = 0;
memcpy(srcx, fft_out + kernel_start, sizeof(*fft_out) * kernel_range); memcpy(srcx, fft_out + kernel_start, sizeof(*fft_out) * kernel_range);
} else {
offset = -kernel_start;
memcpy(srcx+offset, fft_out, sizeof(*fft_out) * (kernel_range-offset));
memcpy(srcx, fft_out+input_padding_size-offset, sizeof(*fft_out)*offset);
}
s->fdsp->vector_fmul_scalar((float *)srcx, (const float *)srcx, scale, FFALIGN(kernel_range * 2, 4)); s->fdsp->vector_fmul_scalar((float *)srcx, (const float *)srcx, scale, FFALIGN(kernel_range * 2, 4));
s->fdsp->vector_fmul((float *)dstx, (const float *)srcx, s->fdsp->vector_fmul((float *)dstx, (const float *)srcx,
(const float *)kernel, FFALIGN(kernel_range * 2, 16)); (const float *)kernel, FFALIGN(kernel_range * 2, 16));
memset(isrc, 0, sizeof(*isrc) * output_padding_size); memset(isrc, 0, sizeof(*isrc) * output_padding_size);
if (offset == 0) {
for (int i = 0; i < kernel_range; i++) { for (int i = 0; i < kernel_range; i++) {
const unsigned n = index[i + kernel_start]; const unsigned n = index[i + kernel_start];
isrc[n].re += dstx[i].re; isrc[n].re += dstx[i].re;
isrc[n].im += dstx[i].im; isrc[n].im += dstx[i].im;
} }
} else {
for (int i = 0; i < kernel_range; i++) {
const unsigned n = (i-kernel_start) & (output_padding_size-1);
isrc[n].re += dstx[i].re;
isrc[n].im += dstx[i].im;
}
}
s->itx_fn(s->ifft[jobnr], idst, isrc, sizeof(*isrc)); s->itx_fn(s->ifft[jobnr], idst, isrc, sizeof(*isrc));
@@ -636,7 +654,6 @@ static int compute_kernel(AVFilterContext *ctx)
const int fsize = s->frequency_band_count; const int fsize = s->frequency_band_count;
int *kernel_start = s->kernel_start; int *kernel_start = s->kernel_start;
int *kernel_stop = s->kernel_stop; int *kernel_stop = s->kernel_stop;
const int hsize = size >> 1;
unsigned *index = s->index; unsigned *index = s->index;
int kernel_min = INT_MAX; int kernel_min = INT_MAX;
int kernel_max = 0, ret = 0; int kernel_max = 0, ret = 0;
@@ -649,17 +666,17 @@ static int compute_kernel(AVFilterContext *ctx)
for (int y = 0; y < fsize; y++) { for (int y = 0; y < fsize; y++) {
AVComplexFloat *kernel = s->kernel[y]; AVComplexFloat *kernel = s->kernel[y];
int start = 0, stop = 0; int start = 0, stop = 0;
float frequency = s->frequency_band[y*2] * correction; const float frequency = s->frequency_band[y*2] * correction;
float deviation = 1.f / (s->frequency_band[y*2+1] * const float deviation = 1.f / (s->frequency_band[y*2+1] *
output_sample_count * correction); output_sample_count * correction);
const int range = 8.f*M_PI*sqrtf(1.f/deviation);
memset(tkernel, 0, size * sizeof(*tkernel)); memset(tkernel, 0, size * sizeof(*tkernel));
for (int n = 0; n < size; n++) { for (int n = -range; n < size-range; n++) {
float ff, f = n+0.5f-frequency; float ff, f = n+0.5f-frequency;
f = hsize - fabsf(f - hsize);
ff = expf(-f*f*deviation); ff = expf(-f*f*deviation);
tkernel[n] = ff; tkernel[n+range] = ff;
} }
for (int n = 0; n < size; n++) { for (int n = 0; n < size; n++) {
@@ -682,8 +699,8 @@ static int compute_kernel(AVFilterContext *ctx)
} }
} }
kernel_start[y] = start; kernel_start[y] = start - range;
kernel_stop[y] = stop; kernel_stop[y] = stop - range;
kernel = av_calloc(FFALIGN(stop - start + 1, 16), sizeof(*kernel)); kernel = av_calloc(FFALIGN(stop - start + 1, 16), sizeof(*kernel));
if (!kernel) { if (!kernel) {