Numpy 图像数组到 pyglet 图像

Question

我正在尝试使用 imageio 将图像加载到 numpy 数组，并使用 pyglet 显示它。尽管我可以看到一些结构，但最终结果是乱码。代码：

import pyglet as pg
import imageio.v3 as im
import numpy as np



window = pg.window.Window()

#Load image, and get_shape
np_image = im.imread("test.png")[100:400, 100:400] #Get smaller section from much larger image (~3Kx3K)
height = np_image.shape[0]
width = np_image.shape[1]
depth = np_image.shape[2]

#Create pyglet image and load image data to it (+ set anchor for displaying)
pg_image = pg.image.create(width, height)
pg_image.set_data("RGB", width*3, np_image)
pg_image.anchor_x = width//2
pg_image.anchor_y = height//2

#Print shapes and dtype, all should be correct
print(np_image.shape)
print(width, height, depth)
print(np_image.dtype)

#Put into sprite
gp_sprite = pg.sprite.Sprite(pg_image, x = window.width//2, y=window.height//2)

@window.event
def on_draw():
    window.clear()
    gp_sprite.draw()s


pg.app.run()

最终结果是：

我在这里做错了什么？

编辑：

调试打印为：

(300, 300, 3)
300 300 3
uint8

Answer 1

Numpy 数组按行主序存储数据。然而，对于图像，数据必须按列主顺序存储。所以你需要

transpose

numpy 数组。此外，您还必须将颜色通道从 BGR 转换为 RGB：

np_image = numpy.transpose(np_image, (1, 0, 2))
np_image[:, :, [0, 2]] = np_image[:, :, [2, 0]]