当ffplay指定启动参数-showmode [0/1/2]时,值为0,正常播放(默认);值为1,显示波形;值为2,显示频谱图。
在ffplay中由show_mode变量控制显示方式取值分别为:SHOW_MODE_VIDEO/SHOW_MODE_WAVES/SHOW_MODE_RDFT
在函数video_display(…)中判断是否有音频流并且show_mode不等于SHOW_MODE_VIDEO,如果成立调用函数video_audio_display(…)
进行波谱或频谱显示
下面详细分析video_audio_display(…)
static void video_audio_display(VideoState *s)
{
int i, i_start, x, y1, y, ys, delay, n, nb_display_channels;
int ch, channels, h, h2;
int64_t time_diff;
int rdft_bits, nb_freq;
//计算离散傅里叶变换输入数组的长度log2的值,绘制频谱时使用双通道,并且填充显示区域高度
for (rdft_bits = 1; (1 << rdft_bits) < 2 * s->height; rdft_bits++)
;
nb_freq = 1 << (rdft_bits - 1);
/* compute display index : center on currently output samples */
channels = s->audio_tgt.channels;
nb_display_channels = channels;
if (!s->paused) {
//该语句块计算当前波/频谱图显示音频数据的开始索引
int data_used= s->show_mode == SHOW_MODE_WAVES ? s->width : (2*nb_freq);
n = 2 * channels;
delay = s->audio_write_buf_size;
delay /= n;
/* to be more precise, we take into account the time spent since
the last buffer computation */
if (audio_callback_time) {
time_diff = av_gettime_relative() - audio_callback_time;
delay -= (time_diff * s->audio_tgt.freq) / 1000000;
}
delay += 2 * data_used;
if (delay < data_used)
delay = data_used;
i_start= x = compute_mod(s->sample_array_index - delay * channels, SAMPLE_ARRAY_SIZE);
if (s->show_mode == SHOW_MODE_WAVES) {
//如果为波谱图查找振幅变化最大的音频部分数据的索引
h = INT_MIN;
for (i = 0; i < 1000; i += channels) {
int idx = (SAMPLE_ARRAY_SIZE + x - i) % SAMPLE_ARRAY_SIZE;
int a = s->sample_array[idx];
int b = s->sample_array[(idx + 4 * channels) % SAMPLE_ARRAY_SIZE];
int c = s->sample_array[(idx + 5 * channels) % SAMPLE_ARRAY_SIZE];
int d = s->sample_array[(idx + 9 * channels) % SAMPLE_ARRAY_SIZE];
int score = a - d;
if (h < score && (b ^ c) < 0) {
h = score;
i_start = idx;
}
}
}
s->last_i_start = i_start;
} else {
i_start = s->last_i_start;
}
if (s->show_mode == SHOW_MODE_WAVES) {
//设置颜色为白色
SDL_SetRenderDrawColor(renderer, 255, 255, 255, 255);
/* total height for one channel */
h = s->height / nb_display_channels;//每一个通道波谱显示区域占有的高度
/* graph height / 2 */
h2 = (h * 9) / 20;
for (ch = 0; ch < nb_display_channels; ch++) {
i = i_start + ch;
y1 = s->ytop + ch * h + (h / 2); /* position of center line */
for (x = 0; x < s->width; x++) {
y = (s->sample_array[i] * h2) >> 15;//y的绝对值一定小于h
if (y < 0) {
y = -y;
ys = y1 - y;
} else {
ys = y1;
}
fill_rectangle(s->xleft + x, ys, 1, y);//绘制高度为1的矩形
i += channels;
if (i >= SAMPLE_ARRAY_SIZE)
i -= SAMPLE_ARRAY_SIZE;
}
}
SDL_SetRenderDrawColor(renderer, 0, 0, 255, 255);
//绘制不同声道波谱图的分割线
for (ch = 1; ch < nb_display_channels; ch++) {
y = s->ytop + ch * h;
fill_rectangle(s->xleft, y, s->width, 1);
}
} else {
if (realloc_texture(&s->vis_texture, SDL_PIXELFORMAT_ARGB8888, s->width, s->height, SDL_BLENDMODE_NONE, 1) < 0)
return;
nb_display_channels= FFMIN(nb_display_channels, 2);//显示最多2声道的频谱图
if (rdft_bits != s->rdft_bits) {
av_rdft_end(s->rdft);
av_free(s->rdft_data);
s->rdft = av_rdft_init(rdft_bits, DFT_R2C);//初始化离散傅里叶变换的Context
s->rdft_bits = rdft_bits;
s->rdft_data = av_malloc_array(nb_freq, 4 *sizeof(*s->rdft_data));//双声道,每一通道的输入数组大小为2*nb_freq
}
if (!s->rdft || !s->rdft_data){
av_log(NULL, AV_LOG_ERROR, "Failed to allocate buffers for RDFT, switching to waves display\n");
s->show_mode = SHOW_MODE_WAVES;
} else {
FFTSample *data[2];
SDL_Rect rect = {.x = s->xpos, .y = 0, .w = 1, .h = s->height};
uint32_t *pixels;
int pitch;
for (ch = 0; ch < nb_display_channels; ch++) {
data[ch] = s->rdft_data + 2 * nb_freq * ch;
i = i_start + ch;
for (x = 0; x < 2 * nb_freq; x++) {
double w = (x-nb_freq) * (1.0 / nb_freq);
data[ch][x] = s->sample_array[i] * (1.0 - w * w);//减小了音频值,为何这样处理??
i += channels;
if (i >= SAMPLE_ARRAY_SIZE)
i -= SAMPLE_ARRAY_SIZE;
}
av_rdft_calc(s->rdft, data[ch]);//进行离散傅里叶变换
}
/* Least efficient way to do this, we should of course
* directly access it but it is more than fast enough. */
if (!SDL_LockTexture(s->vis_texture, &rect, (void **)&pixels, &pitch)) {
pitch >>= 2;//每一个像素4字节,pixels为uint32_t指针
pixels += pitch * s->height;
//for循环填充每一个像素
for (y = 0; y < s->height; y++) {
double w = 1 / sqrt(nb_freq);
//声道1频率值转换为颜色值,显示为红色
int a = sqrt(w * sqrt(data[0][2 * y + 0] * data[0][2 * y + 0] + data[0][2 * y + 1] * data[0][2 * y + 1]));
//声道2频率值转换为颜色值,显示为绿色
int b = (nb_display_channels == 2 ) ? sqrt(w * hypot(data[1][2 * y + 0], data[1][2 * y + 1]))
: a;
a = FFMIN(a, 255);
b = FFMIN(b, 255);
pixels -= pitch;//上一行
*pixels = (a << 16) + (b << 8) + ((a+b) >> 1);
}
SDL_UnlockTexture(s->vis_texture);
}
SDL_RenderCopy(renderer, s->vis_texture, NULL, NULL);
}
if (!s->paused)
s->xpos++;
if (s->xpos >= s->width)
s->xpos= s->xleft;
}
}