[平均RGB图像超出4D np.array

Question

我有一个4D数据np.array，其中包含150个分解为3个通道的图像50X70。数据的形状为（150、50、70、3）。我需要按通道（形状为（50，70，3））的那150张图像的平均值（其中平均R通道将是150个R通道的平均值，依此类推）我已经尝试过：

average = data.mean(axis=0)
averageimage = Image.fromarray(average, 'RGB')
averageimage.show()

但是即使它给出正确的形状，图像看起来也像随机的颜色噪声。

编辑：我已经尝试过

def average_image(a_lot_of_images):
    avg = np.zeros((a_lot_of_images.shape[1], a_lot_of_images.shape[2], a_lot_of_images.shape[3]))
    for i in range(a_lot_of_images.shape[0]):
        avg[:,:,0] += a_lot_of_images[i,:,:,0]
        avg[:,:,1] += a_lot_of_images[i,:,:,1]
        avg[:,:,2] += a_lot_of_images[i,:,:,2]
    for i in [0,1,2]:
        avg[:,:,i] = avg[:,:,i]/a_lot_of_images.shape[0]
    return avg

并且输出仍然看起来像彩色噪点。

Answer 1

def average_image(a_lot_of_images):
    avg = np.zeros((a_lot_of_images.shape[1], a_lot_of_images.shape[2], a_lot_of_images.shape[3]), dtype=float)
    for i in range(a_lot_of_images.shape[0]):
        avg[:, :, 0] += a_lot_of_images[i, :, :, 0]
        avg[:, :, 1] += a_lot_of_images[i, :, :, 1]
        avg[:, :, 2] += a_lot_of_images[i, :, :, 2]
    for i in [0,1,2]:
        avg[:, :, i] = avg[:, :, i]/a_lot_of_images.shape[0]
    avg = np.array(np.round(avg), dtype=np.uint8)
    return avg

有效！有没有办法以更优雅的方式做到这一点？