PROSAGA码农传奇-ffmpeg-当使用奇异的分辨率时，FFmpeg正在创建无效的输出

<div class =“post-text”itemprop =“text”>
  <P>
    发送到编码器的帧的所有平面都与线条大小对齐，线条大小是32（或更大）的倍数。对于格式YUV420P，色度平面的宽度是亮度分量的一半。因此，对于1440的帧宽，色度宽度为720，并且720％32！= 0.但是U和V缓冲区已被指定为宽度x高度的朴素大小。所以当
     <code>
      memcpy
    </code>
     是第一个进行的
     <code>
      (width/2 - (width/2) % 32)
    </code>
     将下一行的元素复制到发送到编码器的帧的步幅填充元素中。这将产生如Q图像中所见的视觉失真。
  </p>
  <P>
    校正是将原始色度平面缓冲区填充到跨步对齐的大小。 OP的修改与评论相关联
  </p>
   <pre>
    <code>
      void fill_yuv_imageY(unsigned char **pict, int frame_index, int width, int height) {
    int x, y, i;
    i = frame_index;

for (y = 0; y < height; y++)
        for (x = 0; x < width; x++)
            pict[0][y * width + x] = x + y + i * 3;
}

void fill_yuv_imageUV(unsigned char **pict, int frame_index, int halfWidth, int height) {
    int x, y, i;
    for (y = 0; y < height / 2; y++) {
        for (x = 0; x < halfWidth; x++) {
            pict[1][y * halfWidth + x] = 128 + y + i * 2;
            pict[2][y * halfWidth + x] = 64 + x + i * 5;
        }
    }
}

int roundUp(int numToRound, int multiple){
    if (multiple == 0)
        return numToRound;

int remainder = numToRound % multiple;
    if (remainder == 0)
        return numToRound;

return numToRound + multiple - remainder;
}

//Allocating test frames
unsigned char*** frames = new unsigned char**[frameCount];
for (int i = 0; i < frameCount; i++) {
    frames[i] = new unsigned char*[3];
    frames[i][0] = new unsigned char[width * height];
    fill_yuv_imageY(frames[i], i, width, height);
    frames[i][1] = new unsigned char[roundUp(width / 2, 32) * (height / 2)];
    frames[i][2] = new unsigned char[roundUp(width / 2, 32) * (height / 2)];
    fill_yuv_imageUV(frames[i], i, roundUp(width / 2, 32), height);
}

</code>
  </pre>
</DIV>