import numpy as np
import pandas as pd
from scipy.interpolate import griddata
from PIL import Image


# 读取CSV文件并填充矩阵A
data = pd.read_csv('PHA SID230828-2003.csv', header=None)
matrix_A = data.values
# matrix_A = np.flipud(matrix_A)


# 获取已知点和对应的值
known_points = []
known_values = []
for u in range(120):
    for v in range(120):
        if matrix_A[u, v] != 0:
            x = int(0.211 * u - 4.538 * v + 894.754)
            y = int(-4.594 * u - 0.1044 * v + 800.7384)
            if u == 27 and v == 15:
                print(x, y)
            known_points.append([x, y])
            known_values.append(matrix_A[u, v])

known_points = np.array(known_points)
known_values = np.array(known_values)

# 创建一个1280x1024的矩阵
matrix_size = (1280, 1024)
matrix = np.zeros(matrix_size)

# 根据已知点和对应的值填充矩阵
for i in range(len(known_points)):
    x, y = known_points[i]
    value = known_values[i]
    matrix[x, y] = value
print(matrix.shape)
# 生成网格点坐标
grid_x, grid_y = np.mgrid[0:matrix_size[0], 0:matrix_size[1]]

# 使用双线性插值填充未知点
points = np.transpose(np.nonzero(matrix == 0))
filled_values = griddata(known_points, known_values, points, method='linear')

# 将填充后的值放入矩阵
for i in range(len(points)):
    x, y = points[i]
    value = filled_values[i]
    matrix[x, y] = value

# 将矩阵中的未知位置填充为0
matrix = np.nan_to_num(matrix, nan=0)

# 将矩阵保存为CSV文件
df = pd.DataFrame(matrix)
df.to_csv('matrix_filled.csv', index=False, header=False)

# # 将矩阵保存为CSV文件
# df = pd.DataFrame(matrix)
# df.to_csv('matrix.csv', index=False, header=False)

matrix_B = matrix.copy()
print(matrix_B.shape)
matrix_c = np.where(matrix_B != 0, matrix_B * 217, matrix_B)

image_filled = Image.fromarray(matrix_c.astype(np.uint8))
# 保存为BMP图像
image_filled.save('filled.bmp')

# 将矩阵B中的非零值取反
matrix_B = np.where(matrix_B != 0, -matrix_B, matrix_B)

# 对负数进行+1操作
matrix_B = np.where(matrix_B < 0, matrix_B + 1, matrix_B)

# 再次对负数进行+1操作
matrix_B = np.where(matrix_B < 0, matrix_B + 1, matrix_B)

# 将矩阵B保存为CSV文件
df = pd.DataFrame(matrix_B)
df.to_csv('PHA_bilinear_reversal.csv', header=False, index=False)

# 将矩阵B中的非零值映射到灰度值范围
matrix_B = np.where(matrix_B != 0, matrix_B * 217, matrix_B)

# 将矩阵B转换为8位灰度图像
image = Image.fromarray(matrix_B.astype(np.uint8))

# 保存为BMP图像
image.save('output.bmp')