### 使用Python和OpenCV测试CUDA的示例代码 为了验证OpenCV是否能够成功利用CUDA进行加速，可以通过编写简单的图像处理程序来检测性能差异。下面是一个具体的例子，展示了如何加载图片并应用高斯模糊滤镜，分别在CPU和GPU上执行。 #### CPU版本实现 ```python import cv2 import time def cpu_blur(image_path): img = cv2.imread(image_path) start_time = time.time() blurred_img = cv2.GaussianBlur(img, (15, 15), 0) end_time = time.time() print(f'CPU processing took {end_time - start_time:.4f} seconds') return blurred_img ``` #### GPU版本实现 对于GPU上的操作，则需引入`cv2.cuda`模块中的相应类与方法： ```python import numpy as np import cv2 from matplotlib import pyplot as plt import time def gpu_blur(image_path): # 初始化CUDA流 stream = cv2.cuda.Stream() # 加载图像到GPU内存中 img_host = cv2.imread(image_path) # 将主机端数据上传至设备端（即显卡） upload_start_time = time.time() img_device = cv2.cuda_GpuMat() img_device.upload(img_host, stream=stream) # 创建Gaussian Filter对象 gaussian_filter = cv2.cuda.createGaussianFilter( srcType=cv2.CV_8UC3, dstType=cv2.CV_8UC3, ksize=(15, 15), sigma1=0 ) # 应用过滤器 filter_start_time = time.time() result_device = cv2.cuda_GpuMat() gaussian_filter.apply(img_device, result_device, stream=stream) # 下载结果回到主机侧 download_start_time = time.time() result_host = result_device.download(stream=stream).get() total_end_time = time.time() print(f'CUDA Upload Time: {(filter_start_time-upload_start_time):.4f}') print(f'CUDA Processing Time: {(download_start_time-filter_start_time):.4f}') print(f'CUDA Download Time: {(total_end_time-download_start_time):.4f}') return result_host ``` 这段代码不仅实现了基本的功能，还提供了详细的计时信息以便比较不同平台之间的效率差距[^1]。