How to control the image after transform to show all original content before transform?

[nissansz]

Sample code below may cause image after transform to show beyong original size.

from PIL import Image
import numpy as np
import cv2

# 读取图像
image = Image.open(r"F:\23.png")

# 定义透视变换矩阵
matrix = np.array([
    [1, 0.2, 0],   # x 轴拉伸因子
    # [0.5, 1, 0],   # y 轴拉伸因子
    [0, 1, 0],   # y 轴拉伸因子
    [0, 0, 1]      # 平移因子
])


# # matrix = np.array([
# #     # [1, 0.5, 0],   # x 轴拉伸因子
# #     [0.5, 0, 0.5],   # x 轴拉伸因子
# #     # [0.5, 1, 0],   # y 轴拉伸因子
# #     [0, 1, 0],   # y 轴拉伸因子
# #     [0, 0, 1]      # 平移因子
# # ])
# 
# width, height = image.size
# # 计算变换后的图像尺寸
# new_width = int(width * 1.5)
# new_height = int(height * 1.5)
# # 进行透视变换
# # trans_image = image.transform(image.size, Image.AFFINE, matrix.flatten())
# trans_image = image.transform((new_width, new_height), Image.AFFINE, matrix.flatten())

# # 保存变换后的图像
# trans_image.save("output_image.jpg")
img = cv2.cvtColor(np.asarray(trans_image),cv2.COLOR_RGB2BGR)
cv2.imshow("OpenCV",img)
cv2.waitKey()

[radarhere]

from PIL import Image
import numpy as np
image = Image.open("F23.png")
matrix = np.array([
  [1, 0.2, 0],
  [0, 1, 0],
  [0, 0, 1]
])
new_width = int(image.width * 1.5)
new_height = int(image.height * 1.5)
transformed_image = image.transform((new_width, new_height), Image.AFFINE, matrix.flatten())
transformed_image.save("output_image.jpg")

gives

You can use crop() to manually remove the black box.

But I think you're asking for a way to reverse the matrix operation? You can simply flip() the matrix.

from PIL import Image
import numpy as np
image = Image.open("F23.png")
matrix = np.array([
  [1, 0.2, 0],
  [0, 1, 0],
  [0, 0, 1]
])
new_width = int(image.width * 1.5)
new_height = int(image.height * 1.5)
transformed_image = image.transform((new_width, new_height), Image.AFFINE, matrix.flatten())

matrix = np.flip(matrix)
reversed_image = image.transform(image.size, Image.AFFINE, matrix.flatten())
reversed_image.save("reversed_image.png")

[radarhere]

The numbers are x, y pairs of co-ordinates of the shape that is being transformed into the final image.

The original image is 128px by 128px.

The first two numbers are 0, 0, because the top left corner should stay the same.
The second two numbers are 0, 128, because the bottom left corner should stay the same.
The third two numbers are 128, 158, because I'm pulling in 30 pixels from below the image.
The fourth two numbers are 128, -30 because I'm pulling in 30 pixels from above the image.

[nissansz]

transform = ImageTransform.QuadTransform((0, 0, 0, height, width, height+30, width, -30))
It works now. Thank you.

[nissansz]

from PIL import Image
import numpy as np
image = Image.open("F23.png")
matrix = np.array([
  [1, 0.2, 0],
  [0, 1, 0],
  [0, 0, 1]
])
new_width = int(image.width * 1.5)
new_height = int(image.height * 1.5)
transformed_image = image.transform((new_width, new_height), Image.AFFINE, matrix.flatten())
transformed_image.save("output_image.jpg")

gives

You can use crop() to manually remove the black box.

But I think you're asking for a way to reverse the matrix operation? You can simply flip() the matrix.

from PIL import Image
import numpy as np
image = Image.open("F23.png")
matrix = np.array([
  [1, 0.2, 0],
  [0, 1, 0],
  [0, 0, 1]
])
new_width = int(image.width * 1.5)
new_height = int(image.height * 1.5)
transformed_image = image.transform((new_width, new_height), Image.AFFINE, matrix.flatten())

matrix = np.flip(matrix)
reversed_image = image.transform(image.size, Image.AFFINE, matrix.flatten())
reversed_image.save("reversed_image.png")

matrix = np.array([
[1, 0.2, 0],
[0, 1, 0],
[0, 0, 1]
])

If I change 0.2 to 0.8 etc., the image will be like below, some characters in the beiginning disappeared, how to solve such issue?

[nissansz]

I copied original image to a white blank image width*1.2, then transform, can show whole image now

[radarhere]

Thank you for finding the solution.