basic-usage.md

layout	default
title	Basic Usage Example
description	Basic Mini-OpenCV usage example - loading, processing, and saving images.

Basic Usage Example

Complete walkthrough of loading an image, applying operations, and saving the result.

Full Code

#include "gpu_image/gpu_image_processing.hpp"
#include <iostream>

using namespace gpu_image;

int main() {
    // Step 1: Check CUDA availability
    if (!isCudaAvailable()) {
        std::cerr << "CUDA not available!" << std::endl;
        return 1;
    }
    std::cout << "Using: " << getDeviceInfo() << std::endl;

    // Step 2: Create processor
    ImageProcessor processor;

    // Step 3: Create test image
    HostImage input = ImageUtils::createHostImage(512, 512, 3);
    for (int y = 0; y < 512; ++y) {
        for (int x = 0; x < 512; ++x) {
            input.at(x, y, 0) = (x * 255) / 512;  // R gradient
            input.at(x, y, 1) = (y * 255) / 512;  // G gradient
            input.at(x, y, 2) = 128;               // B constant
        }
    }

    // Step 4: Upload to GPU
    GpuImage gpu = processor.loadFromHost(input);

    // Step 5: Apply operations
    GpuImage blurred = processor.gaussianBlur(gpu, 5, 1.5f);
    GpuImage edges = processor.sobelEdgeDetection(blurred);

    // Step 6: Download result
    HostImage output = processor.downloadImage(edges);

    // Step 7: Verify output
    std::cout << "Output: " << output.width << "x" << output.height 
              << "x" << output.channels << std::endl;

    return 0;
}

Step-by-Step Explanation

1. CUDA Check

Always verify CUDA is available before processing.

2. Processor Creation

ImageProcessor manages GPU resources and provides synchronous operations.

3. Image Creation

HostImage stores data in CPU RAM with convenient pixel accessors.

4. GPU Upload

loadFromHost() copies data to GPU memory via PCIe.

5. Processing

Operations run on GPU and synchronize automatically.

6. Download

downloadImage() copies results back to CPU memory.

Compilation

# With CMake
target_link_libraries(my_app gpu_image::gpu_image_processing)

# Direct compilation
nvcc -std=c++17 example.cpp -I/path/to/mini-opencv/include \
    -L/path/to/mini-opencv/build/lib -lgpu_image_processing -o example