GStreamer Webcam Capture

A C++ application using GStreamer to capture webcam video with MJPEG input, apply various transformations, and encode/decode using NVIDIA hardware acceleration.

Features

• MJPEG Input: Captures from /dev/video0 at 1280x720@30fps in MJPEG format
• Video Processing Pipeline:
  • Decode MJPEG stream
  • Upscale to 1920x1080
  • Apply color inversion filter (frei0r-filter-invert0r)
  • Horizontal flip
  • NV12 format optimization for NVIDIA hardware
  • NVIDIA H.264 hardware encoding (10Mbps, low-latency)
  • NVIDIA H.264 hardware decoding
• Auto Video Display: Automatically selects appropriate video sink

Prerequisites

System Requirements

• Linux operating system
• Webcam device at /dev/video0 supporting MJPEG format
• NVIDIA GPU with hardware encoding/decoding support (for nvh264enc/nvh264dec)
• GStreamer 1.0 or later

Dependencies

sudo apt-get update
sudo apt-get install -y \
    build-essential \
    pkg-config \
    libgstreamer1.0-dev \
    libgstreamer-plugins-base1.0-dev \
    gstreamer1.0-plugins-base \
    gstreamer1.0-plugins-good \
    gstreamer1.0-plugins-bad \
    gstreamer1.0-tools \
    gstreamer1.0-plugins-ugly \
    frei0r-plugins \
    gstreamer1.0-plugins-nvcodec

Or use the Makefile target:

make install-deps
make install-nvidia-deps

Building

Compile the application:

make

Clean build artifacts:

make clean

Usage

Run the application:

./webcam_capture

Or use the Makefile:

make run

The application will:

1. Initialize the GStreamer pipeline
2. Start capturing from the webcam
3. Apply transformations and display the video
4. Run for 10,000 seconds (configurable in source)
5. Press Ctrl+C to stop early

Pipeline Details

The GStreamer pipeline performs the following steps:

1. Capture: v4l2src device=/dev/video0 - Captures from webcam
2. Input Format: MJPEG 1280x720@30fps
3. Decode: jpegdec - Decodes MJPEG
4. Convert: videoconvert - Converts to raw video
5. Upscale: videoscale - Scales to 1920x1080
6. Invert: frei0r-filter-invert0r - Inverts colors
7. Flip: videoflip method=horizontal-flip - Mirrors horizontally
8. Format: video/x-raw,format=NV12 - Optimized format for NVIDIA hardware
9. Encode: nvh264enc - NVIDIA H.264 encoding at 10Mbps (low-latency, CBR)
10. Parse: h264parse - Parses H.264 stream
11. Decode: nvh264dec - NVIDIA H.264 decoding
12. Display: autovideosink - Displays video

Why NV12?

NV12 is the native format for NVIDIA NVENC/NVDEC hardware, providing:

• Zero-copy operations between videoconvert and nvh264enc
• Optimal GPU performance
• Lower CPU overhead compared to I420 format

Customization

Edit main.cpp to customize:

• Input device path
• Resolution and framerate
• Bitrate and encoding settings
• Video filters and effects
• Display duration

Then rebuild with make.