深度图像边缘提取及转储

import cv2

def extract_frames(video_path, interval):
    # 打开视频文件
    cap = cv2.VideoCapture(video_path)
    # 计算视频总帧数和帧率
    frame_count = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = cap.get(cv2.CAP_PROP_FPS)
    # 计算抽帧间隔
    interval_frames = int(interval * fps)
    # 初始化帧计数器和关键帧列表
    count = 0
    frames = []
    # 逐帧遍历视频
    while True:
        ret, frame = cap.read()
        if not ret:
            break
        count += 1
        # 如果是关键帧，将其添加到关键帧列表中
        if count % interval_frames == 0:
            frames.append(frame)
    # 关闭视频文件
    cap.release()
    return frames

frames = extract_frames('video.mp4', 1)  # 抽取间隔为1秒的关键帧
for frame in frames:
    cv2.imshow('frame', frame)
    cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np

def extract_depth_edges(depth_img):
    # 计算Sobel算子的卷积核
    sobelx = cv2.Sobel(depth_img, cv2.CV_64F, 1, 0, ksize=3)
    sobely = cv2.Sobel(depth_img, cv2.CV_64F, 0, 1, ksize=3)
    # 计算梯度幅值和方向
    grad_mag = np.sqrt(sobelx ** 2 + sobely ** 2)
    grad_dir = np.arctan2(sobely, sobelx)
    # 将梯度幅值归一化到0-255之间
    grad_mag_norm = cv2.normalize(grad_mag, None, 0, 255, cv2.NORM_MINMAX, cv2.CV_8U)
    # 将梯度方向转换为角度
    grad_dir_deg = (grad_dir * 180 / np.pi) % 180
    # 应用非极大值抑制
    grad_mag_nms = cv2.Canny(grad_mag_norm, 100, 200)
    return grad_mag_nms

depth_img = cv2.imread('depth_image.png', cv2.IMREAD_GRAYSCALE)
edges = extract_depth_edges(depth_img)
cv2.imshow('depth_edges', edges)
cv2.waitKey(0)
cv2.destroyAllWindows()

import numpy as np

# 假设我们已经提取了深度图像的边缘信息，存储在名为edge_img的NumPy数组中

# 将边缘值缩放到0到1之间
edge_img = edge_img / 255.0

# 将边缘信息转换为字符串格式
edge_str = np.array2string(edge_img, separator=',', formatter={'float_kind':lambda x: "%.5f" % x})

# 将字符串写入txt文件
with open('edge_info.txt', 'w') as f:
    f.write(edge_str)

import cv2
import numpy as np

def extract_edge(frame, threshold):
    # 使用高斯模糊平滑图像
    blurred = cv2.GaussianBlur(frame, (3, 3), 0)
    # 转换为灰度图像
    gray = cv2.cvtColor(blurred, cv2.COLOR_BGR2GRAY)
    # 使用Canny算法提取边缘
    edges = cv2.Canny(gray, threshold, threshold * 2)
    return edges

def save_edges_to_txt(edges, filename):
    # 将边缘值缩放到0到1之间
    edges = edges / 255.0
    # 将边缘信息转换为字符串格式
    edge_str = np.array2string(edges, separator=',', formatter={'float_kind':lambda x: "%.5f" % x})
    # 将字符串写入txt文件
    with open(filename, 'w') as f:
        f.write(edge_str)

# 读取深度图像
depth_img = cv2.imread('depth_img.png')
# 指定抽帧间隔
interval = 10
# 提取深度图像边缘
edges = extract_edge(depth_img, 50)
# 抽帧，保留每隔interval个像素
sampled_edges = edges[::interval, ::interval]
# 将边缘信息保存到txt文件中
save_edges_to_txt(sampled_edges, 'edge_info.txt')

import cv2
import numpy as np

def load_edges_from_txt(filename, shape):
    # 从txt文件中读取边缘信息
    edge_str = np.loadtxt(filename, delimiter=',')
    # 创建全零数组
    edges = np.zeros(shape)
    # 将边缘信息复制到全零数组的对应位置上
    np.put(edges, np.arange(shape[0]*shape[1]), edge_str)
    # 对全零数组进行插值操作
    edges = cv2.resize(edges, (shape[1], shape[0]))
    # 对插值后的边缘图像进行二值化处理
    ret, edges = cv2.threshold(edges, 0, 255, cv2.THRESH_BINARY)
    return edges

# 读取深度图像
depth_img = cv2.imread('depth_img.png')
# 获取深度图像大小
height, width = depth_img.shape[:2]
# 从txt文件中加载边缘信息，并恢复成图像
edges = load_edges_from_txt('edge_info.txt', (height//10, width//10))
# 显示原始深度图像和恢复的边缘图像
cv2.imshow('depth_img', depth_img)
cv2.imshow('edges', edges)
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2

# 读取原始图像
img = cv2.imread('original_image.png')
# 缩放图像
new_img = cv2.resize(img, (10, 10), interpolation=cv2.INTER_AREA)
# 显示原始图像和缩放后的图像
cv2.imshow('original', img)
cv2.imshow('new', new_img)
cv2.waitKey(0)
cv2.destroyAllWindows()