Python 将图像中矩形四角二维坐标转化为归一化一维坐标

计算原理：根据矩形四角的二维坐标得出矩形中心的坐标，再计算出矩形的宽和高，由此可将四个二维坐标转化为四个一维坐标 。归一化即x轴的值除以图像宽度，y轴的值除以图像高度。

[中心坐标x, 中心坐标y, 矩形宽, 矩形高]

def normalize_bbox_coordinates(bbox_coords, image_width=500, image_height=500):
    """
    将矩形框坐标归一化到[0,1]范围

    :param bbox_coords: 矩形框四个坐标点 [(x1,y1), (x2,y2), (x3,y3), (x4,y4)]
    :param image_width: 图片宽度，默认500
    :param image_height: 图片高度，默认500
    :return list: 归一化后的坐标 [x_center, y_center, width, height]
    """
    # 提取所有x坐标和y坐标
    x_coords = [point[0] for point in bbox_coords]
    y_coords = [point[1] for point in bbox_coords]

    # 计算边界框的最小和最大坐标
    x_min = min(x_coords)
    x_max = max(x_coords)
    y_min = min(y_coords)
    y_max = max(y_coords)

    # 计算中心点坐标和宽高
    width = x_max - x_min
    height = y_max - y_min
    x_center = x_min + width / 2
    y_center = y_min + height / 2

    # 归一化到[0,1]范围
    x_center_norm = x_center / image_width
    y_center_norm = y_center / image_height
    width_norm = width / image_width
    height_norm = height / image_height

    return [x_center_norm, y_center_norm, width_norm, height_norm]


bbox_coords = [(125, 125), (125, 375), (375, 375), (375, 125)]
normalized_coords = normalize_bbox_coordinates(bbox_coords)
print(f"原始坐标: {bbox_coords}")
print(f"归一化坐标: {normalized_coords}")
print(f"验证结果: {[round(coord, 1) for coord in normalized_coords]}")

原始坐标: [(125, 125), (125, 375), (375, 375), (375, 125)]
归一化坐标: [0.5, 0.5, 0.5, 0.5]
验证结果: [0.5, 0.5, 0.5, 0.5]

Python 将图像中归一化一维坐标转化为矩形四角二维坐标

def denormalize_coordinates(normalized_coords, image_width=500, image_height=500):
    """
    将归一化坐标复原到原始图像尺寸

    :param normalized_coords: 归一化坐标 [x_center_norm, y_center_norm, width_norm, height_norm]
    :param image_width: 图片宽度，默认500
    :param image_height: 图片高度，默认500
    :return tuple: (center_x, center_y, width, height) 原始坐标和尺寸
    """
    if len(normalized_coords) != 4:
        raise ValueError("归一化坐标必须是4个值: [x_center, y_center, width, height]")

    x_center_norm, y_center_norm, width_norm, height_norm = normalized_coords

    # 复原中心坐标
    center_x = x_center_norm * image_width
    center_y = y_center_norm * image_height

    # 复原宽度和高度
    width = width_norm * image_width
    height = height_norm * image_height

    return center_x, center_y, width, height

def denormalize_to_corners(normalized_coords, image_width=500, image_height=500):
    """
    将归一化坐标复原为四个角点坐标

    :param normalized_coords: 归一化坐标 [x_center_norm, y_center_norm, width_norm, height_norm]
    :param image_width: 图片宽度，默认500
    :param image_height: 图片高度，默认500
    :return list: 四个角点坐标 [(x1,y1), (x2,y2), (x3,y3), (x4,y4)]
    """
    center_x, center_y, width, height = denormalize_coordinates(normalized_coords, image_width, image_height)

    # 计算角点坐标
    x_min = center_x - width / 2
    x_max = center_x + width / 2
    y_min = center_y - height / 2
    y_max = center_y + height / 2

    # 返回四个角点（顺时针顺序）
    corners = [
        (x_min, y_min),  # 左上
        (x_min, y_max),  # 左下
        (x_max, y_max),  # 右下
        (x_max, y_min)   # 右上
    ]

    return corners

# 归一化坐标
normalized_coords = [0.5, 0.5, 0.5, 0.5]

# 复原中心坐标和尺寸
center_x, center_y, width, height = denormalize_coordinates(normalized_coords)
print(f"归一化坐标: {normalized_coords}")
print(f"复原中心坐标: ({center_x}, {center_y}), 尺寸: {width}x{height}")

# 复原为四个角点
corners = denormalize_to_corners(normalized_coords)
print(f"复原角点坐标: {corners}")

# 验证复原结果是否与原始坐标一致
original_corners = [(125, 125), (125, 375), (375, 375), (375, 125)]
print(f"原始角点坐标: {original_corners}")
print(f"复原是否正确: {corners == original_corners}")

归一化坐标: [0.5, 0.5, 0.5, 0.5]
复原中心坐标: (250.0, 250.0), 尺寸: 250.0x250.0
复原角点坐标: [(125.0, 125.0), (125.0, 375.0), (375.0, 375.0), (375.0, 125.0)]
原始角点坐标: [(125, 125), (125, 375), (375, 375), (375, 125)]
复原是否正确: True