Image processing TersorFlow如何期望展平图像？

当使用conv_2D层在TersorFlow中执行2D卷积时，它是否希望像素按如下方式排列

[
  [img[i].red, img[i].green, img[i].blue],
  [img[i+1].red, etc.],
]

或

---

或者其他方式？

2D卷积需要一个4-d张量作为输入，其形状如下：

[批次大小、图像高度、图像宽度、通道大小]

对于rgb图像，通道为三种颜色。因此，像素应排列为：

[
  [
    [img[i,j].red, img[i,j].green, img[i,j].blue], 
    [img[i, j+1].red, img[i, j+1].green, img[i, j+1].blue],
    etc
  ],
  [
    [img[i+1,j].red, img[i+1,j].green, img[i+1,j].blue],
    [img[i+1, j+1].red, img[i+1, j+1].green, img[i+1, j+1].blue],
    etc
  ],
  etc
]

img[y\u坐标，x\u坐标]和img[i，j]=img[i*图像宽度+j]

[
  [
    [img[i,j].red, img[i,j].green, img[i,j].blue], 
    [img[i, j+1].red, img[i, j+1].green, img[i, j+1].blue],
    etc
  ],
  [
    [img[i+1,j].red, img[i+1,j].green, img[i+1,j].blue],
    [img[i+1, j+1].red, img[i+1, j+1].green, img[i+1, j+1].blue],
    etc
  ],
  etc
]