223 lines
5.7 KiB
Python
223 lines
5.7 KiB
Python
import numpy as np
|
||
|
||
# 假设一张 4x4 的灰度图像
|
||
image = np.array([
|
||
[255, 200, 150, 100],
|
||
[180, 140, 110, 80],
|
||
[120, 90, 60, 40],
|
||
[50, 30, 20, 10]
|
||
])
|
||
# 255 = 最亮（白），0 = 最暗（黑）
|
||
# 彩色图像：高度 x 宽度 x 3（红、绿、蓝三通道）
|
||
color_image = np.array([
|
||
[[255, 0, 0], [0, 255, 0], [0, 0, 255]], # 第1行像素：红、绿、蓝
|
||
[[255, 255, 0], [0, 255, 255], [255, 0, 255]], # 第2行像素：黄、青、紫
|
||
[[128, 128, 128],[100, 100, 100],[50, 50, 50]] # 第3行像素：灰度
|
||
])
|
||
# 一维数组（向量）
|
||
vec = np.array([1, 2, 3, 4, 5])
|
||
print(vec) # [1 2 3 4 5]
|
||
|
||
# 二维数组（矩阵）
|
||
mat = np.array([
|
||
[1, 2, 3],
|
||
[4, 5, 6]
|
||
])
|
||
print(mat)
|
||
# [[1 2 3]
|
||
# [4 5 6]]
|
||
# 全0数组 - 常用于初始化
|
||
zeros = np.zeros((3, 4)) # 3行4列的全0矩阵
|
||
print(zeros)
|
||
# [[0. 0. 0. 0.]
|
||
# [0. 0. 0. 0.]
|
||
# [0. 0. 0. 0.]]
|
||
|
||
# 全1数组
|
||
ones = np.ones((2, 3))
|
||
# [[1. 1. 1.]
|
||
# [1. 1. 1.]]
|
||
|
||
# 指定范围内的数组
|
||
arr = np.arange(0, 10, 2) # 0到10，步长2
|
||
print(arr) # [0 2 4 6 8]
|
||
|
||
# 等差数组
|
||
lin = np.linspace(0, 1, 5) # 0到1之间均匀取5个数
|
||
print(lin) # [0. 0.25 0.5 0.75 1. ]
|
||
a = np.array([[1, 2, 3], [4, 5, 6]])
|
||
|
||
print(a.ndim) # 维度数量 = 2（是二维数组/矩阵）
|
||
print(a.shape) # 形状 = (2, 3) 2行3列
|
||
print(a.size) # 总元素数 = 6
|
||
print(a.dtype) # 数据类型 = int64
|
||
|
||
# 对于图像数据，常用 uint8（无符号8位整数，0-255）
|
||
image = np.array([[255, 128], [64, 0]], dtype=np.uint8)
|
||
print(image.dtype) # uint8
|
||
a = np.array([10, 20, 30, 40, 50])
|
||
|
||
print(a[0]) # 第一个元素 = 10
|
||
print(a[-1]) # 最后一个元素 = 50
|
||
print(a[1:3]) # 第2到第3个元素 = [20, 30]
|
||
print(a[:3]) # 前3个元素 = [10, 20, 30]
|
||
print(a[2:]) # 第3个到最后 = [30, 40, 50]
|
||
mat = np.array([
|
||
[1, 2, 3, 4],
|
||
[5, 6, 7, 8],
|
||
[9, 10, 11, 12]
|
||
])
|
||
|
||
# 获取特定元素
|
||
print(mat[0, 0]) # 第1行第1列 = 1
|
||
print(mat[1, 2]) # 第2行第3列 = 7
|
||
print(mat[2, -1]) # 第3行最后一列 = 12
|
||
|
||
# 切片：[行, 列]
|
||
print(mat[0, :]) # 第1行所有列 = [1, 2, 3, 4]
|
||
print(mat[:, 1]) # 所有行第2列 = [2, 6, 10]
|
||
print(mat[0:2, 0:2]) # 取前2行前2列
|
||
# [[1 2]
|
||
# [5 6]]
|
||
|
||
# 负索引
|
||
print(mat[-1, :]) # 最后一行 = [9, 10, 11, 12]
|
||
a = np.array([1, 2, 3])
|
||
b = np.array([4, 5, 6])
|
||
|
||
# 加减乘除（对应元素运算）
|
||
print(a + b) # [5 7 9]
|
||
print(a - b) # [-3 -3 -3]
|
||
print(a * b) # [4 10 18] 注意：这是元素乘法，不是矩阵乘法
|
||
print(a / b) # [0.25 0.4 0.5]
|
||
|
||
# 标量运算（广播）
|
||
print(a * 2) # [2 4 6]
|
||
print(a + 10) # [11 12 13]
|
||
print(a ** 2) # [1 4 9] 平方
|
||
A = np.array([[1, 2], [3, 4]])
|
||
B = np.array([[5, 6], [7, 8]])
|
||
|
||
# 使用 @ 运算符 或 np.dot()
|
||
C = A @ B
|
||
print(C)
|
||
# [[19 22]
|
||
# [41 46]]
|
||
|
||
# 或者
|
||
C = np.dot(A, B)
|
||
print(C)
|
||
a = np.array([1, 2, 3, 4, 5])
|
||
|
||
print(np.sum(a)) # 求和 = 15
|
||
print(np.mean(a)) # 平均值 = 3.0
|
||
print(np.max(a)) # 最大值 = 5
|
||
print(np.min(a)) # 最小值 = 1
|
||
print(np.std(a)) # 标准差
|
||
print(np.argmax(a)) # 最大值的索引 = 4
|
||
print(np.argmin(a)) # 最小值的索引 = 0
|
||
|
||
# 对于二维数组，可以指定轴
|
||
mat = np.array([[1, 2, 3], [4, 5, 6]])
|
||
print(np.sum(mat, axis=0)) # 按列求和 = [5 7 9]
|
||
pra = np.array([1, 2, 3, 4, 5, 6])
|
||
|
||
# 展平（flatten）: 多维变一维
|
||
print(a.flatten()) # [1 2 3 4 5 6]
|
||
|
||
# 改变形状（reshape）
|
||
b = a.reshape(2, 3) # 变成2行3列
|
||
print(b)
|
||
# [[1 2 3]
|
||
# [4 5 6]]
|
||
|
||
# 转置（transpose）: 行变列
|
||
print(b.T)
|
||
# [[1 4]
|
||
# [2 5]
|
||
# [3 6]]
|
||
|
||
# 翻转
|
||
c = np.array([[1, 2, 3], [4, 5, 6]])
|
||
print(np.fliplr(c)) # 左右翻转
|
||
# [[3 2 1]
|
||
# [6 5 4]]
|
||
print(np.flipud(c)) # 上下翻转
|
||
# [[4 5 6]
|
||
# [1 2 3]]int(np.sum(mat, axis=1)) # 按行求和 = [6 15]
|
||
a = np.array([1, 2, 3, 4, 5, 6])
|
||
|
||
# 展平（flatten）: 多维变一维
|
||
print(a.flatten()) # [1 2 3 4 5 6]
|
||
|
||
# 改变形状（reshape）
|
||
b = a.reshape(2, 3) # 变成2行3列
|
||
print(b)
|
||
# [[1 2 3]
|
||
# [4 5 6]]
|
||
|
||
# 转置（transpose）: 行变列
|
||
print(b.T)
|
||
# [[1 4]
|
||
# [2 5]
|
||
# [3 6]]
|
||
|
||
# 翻转
|
||
c = np.array([[1, 2, 3], [4, 5, 6]])
|
||
print(np.fliplr(c)) # 左右翻转
|
||
# [[3 2 1]
|
||
# [6 5 4]]
|
||
print(np.flipud(c)) # 上下翻转
|
||
# [[4 5 6]
|
||
# [1 2 3]]
|
||
a = np.array([10, 20, 30, 40, 50])
|
||
|
||
# 找出满足条件的元素
|
||
print(a > 25) # [False False False True True]
|
||
print(a[a > 25]) # [30 40 50] 只保留大于25的值
|
||
|
||
# 统计满足条件的数量
|
||
print(np.sum(a > 25)) # 3
|
||
|
||
# 替换值
|
||
b = np.where(a > 25, a, 0) # 大于25保留，否则改成0
|
||
print(b) # [0 0 30 40 50]
|
||
a = np.array([3, 4])
|
||
|
||
# L2范数（欧几里得距离）
|
||
norm = np.linalg.norm(a)
|
||
print(norm) # 5.0 (因为 3² + 4² = 25, √25 = 5)
|
||
|
||
# 点积（dot product）
|
||
b = np.array([1, 2])
|
||
dot = np.dot(a, b)
|
||
print(dot) # 3*1 + 4*2 = 11
|
||
|
||
# 余弦相似度
|
||
def cosine_similarity(a, b):
|
||
return np.dot(a, b) / (np.linalg.norm(a) * np.linalg.norm(b))
|
||
|
||
v1 = np.array([1, 0, 1])
|
||
v2 = np.array([1, 1, 0])
|
||
print(cosine_similarity(v1, v2)) # 约 0.707
|
||
a = np.array([1, 2, 3])
|
||
b = a # 这是引用，不是复制！
|
||
b[0] = 100
|
||
print(a[0]) # 100，a也跟着变了
|
||
|
||
# 如果想复制，应该用
|
||
c = a.copy()
|
||
c[0] = 999
|
||
print(a[0]) # 100，a不变
|
||
# 亮度调整：所有像素值加一个常数
|
||
brighter = image + 30 # 变亮
|
||
darker = image - 30 # 变暗
|
||
|
||
# 对比度调整：像素值乘以一个系数
|
||
contrasted = image * 1.5 # 对比度增强
|
||
# 裁剪：取矩阵的一部分
|
||
top_left = image[0:2, 0:2] # 取左上角2x2区域
|
||
|
||
# 旋转：矩阵转置和翻转
|
||
rotated = np.transpose(image) # 转置
|
||
flipped = np.fliplr(image) # 左右翻转 |