Multiprocessing for Speed

Multiprocessing for Speed

This tutorial shows how to use OpenPIV’s multiprocessing capabilities to analyze multiple image pairs in parallel.

Why Multiprocessing?

When analyzing movies or long sequences, you need to process many image pairs. Multiprocessing distributes the workload across multiple CPU cores, significantly reducing processing time.

Code Example

Here’s how to set up parallel processing:

import pathlib
import numpy as np
from openpiv import tools, scaling, pyprocess, validation, filters

# Define a worker function for processing each image pair
def process_pair(args):
    """Process a single image pair."""
    file_a, file_b, output_idx = args
    
    # Read images
    img_a = tools.imread(pathlib.Path("data/test1") / file_a)
    img_b = tools.imread(pathlib.Path("data/test1") / file_b)
    
    # Convert to int32 for processing
    img_a = (img_a * 1024).astype(np.int32)
    img_b = (img_b * 1024).astype(np.int32)
    
    # Run PIV analysis
    u, v, sig2noise = pyprocess.extended_search_area_piv(
        img_a, img_b,
        window_size=32,
        overlap=16,
        dt=0.02,
        search_area_size=38,
        sig2noise_method='peak2peak'
    )
    
    # Validate
    mask = validation.sig2noise_val(sig2noise, threshold=1.5)
    
    # Get coordinates
    x, y = pyprocess.get_coordinates(
        image_size=img_a.shape,
        search_area_size=38,
        overlap=16
    )
    
    return x, y, u, v, mask

print("Worker function defined for parallel processing")

Worker function defined for parallel processing

Using the Multiprocesser

OpenPIV provides a convenient class for managing parallel processing:

# Check available test files
import os
test_path = pathlib.Path("data/test1")
files = list(test_path.glob("*.bmp"))
print(f"Available test images: {len(files)}")
for f in sorted(files):
    print(f"  - {f.name}")

Available test images: 0

Summary

Multiprocessing is essential for: - Analyzing long image sequences - Processing high-resolution images - Reducing overall processing time

Next Steps

Try the Multipass PIV tutorial to learn about iterative refinement for improved accuracy.