What is CUDA-Based Color Manipulation in GStreamer?

CUDA-based color manipulation refers to performing real-time image and video color transformations using NVIDIA CUDA kernels inside a GStreamer pipeline. Instead of processing video frames on the CPU, video buffers remain in GPU memory (NVMM), and color operations are executed directly through parallel CUDA threads.

Under RAPIDSEA, the following optimized plugins are available:

  • gst_et_cuda_brightness
  • gst_et_cuda_contrast
  • gst_et_cuda_binarization
  • gst_et_cuda_whitebalance

These plugins are engineered for low latency, high throughput, and zero-copy GPU workflows, making them ideal for performance-critical embedded video applications.

The Zero-Copy Advantage

Traditional GStreamer pipelines often suffer from the "ping-pong" effect: copying data from the GPU to the CPU for processing and back again. RAPIDSEA utilizes NVMM (NVIDIA Memory Management) to ensure a Zero-Copy Workflow.

Performance Comparison:

Feature Standard CPU Plugins RAPIDSEA CUDA Plugins
Latency High (due to memory copying) Ultra-Low (Zero-copy)
Throughput Sequential/Limited Massive Parallelism
CPU Usage High (Heavily Taxed) Minimal (Idle for App Logic)

Color Adjustment: Brightness Plugin

The gst_et_cuda_brightness plugin is a CUDA-enabled GStreamer element designed to adjust the brightness of video frames in real-time. It provides high-performance pixel manipulation while keeping the CPU load minimal.

What the Brightness Plugin Does:

The plugin adjusts the overall luminance of video frames by applying a user-defined brightness factor to each pixel. The CUDA kernel updates the final image instantly, enabling smooth processing of high-resolution streams.

Common Operations:

  • Increasing/Decreasing Brightness: Shift luminance via positive or negative factors.
  • Uniform Shifts: Applies consistent brightness adjustments across all frames.
  • Real-time Configuration: Adjustable via a user-configurable property range.

Features:

  • Property: brightness (float).
  • Range: -1.0 to +1.0 (internally scaled to -255 to +255).
  • Default: 0.0 (no change).

Contrast Plugin (CUDA-Accelerated)

Designed to enhance or reduce the contrast of video frames, gst_et_cuda_contrast utilizes NVIDIA GPUs to deliver extremely fast adjustments for high-resolution video while keeping CPU usage to a minimum.

What the Contrast Plugin Does:

This plugin adjusts the dynamic range of pixel intensities in GPU-resident video frames. By scaling the difference between each pixel and the mid-range luminance point, it increases or decreases visual contrast across the entire image.

Features:

  • Property: contrast (float).
  • Range: 0.0 to 3.0.
  • Default: Default: 1.0.

Color Thresholding / Binarization - gst_et_cuda_binarization

The Thresholding/Binarization module is one of the RAPIDSEA CUDA Color Manipulation GStreamer plugins, designed for GPU-accelerated, real-time video processing on NVIDIA platforms. It converts grayscale or RGBA video frames into binary images, where each pixel is classified as either foreground (white) or background (black) based on a defined threshold, enabling efficient image segmentation in embedded vision and edge AI pipelines.

Color thresholding, also known as binarization, is a widely used image processing technique that simplifies a frame by converting it into two distinct pixel values (typically 0 and 255). This reduces visual complexity and improves performance for object detection, motion analysis, and computer vision applications within zero-copy GPU GStreamer pipelines.

The plugin compares pixel intensity to a configurable threshold value directly on CUDA memory for high-throughput, low-latency processing.

Features:

White Balancing

The gst_et_cuda_whitebalance plugin modifies the intensity of the red, green, and blue (RGB) channels to ensure neutral whites under varying lighting conditions.

Advanced Auto White Balance Engine

This integrated engine analyzes every frame in real-time to detect shifting lighting conditions, automatically correcting the color temperature for a natural look.

Manual RGB Channel Gain Control

For developers requiring total precision, the plugin supports direct manual gain adjustments for the Red, Green, and Blue channels, passed directly through your processing pipeline.

Custom Sensor Scene Calibration Support

This feature allows colorists and engineers to fine-tune color correction based on specific sensor characteristics, unique scene lighting, or custom calibration profiles.

Global Use Cases

From autonomous driving to medical diagnostics, RAPIDSEA’s plugins deliver the low-latency GPU performance required to solve the world's most complex visual challenges.

Contact sales to learn more
Autonomous Vehicles icon

Autonomous Vehicles

Real-time enhancement of sensor data in low-light environments for safer navigation.

Smart Surveillance icon

Smart Surveillance

Improving visibility in night-time footage and pre-processing for motion detection.

Medical Imaging icon

Medical Imaging

Sharpening contrast in diagnostic video feeds for better anomaly detection.

Broadcast & Media icon

Broadcast & Media

Color correction for live streaming events without expensive hardware encoders.

Industrial Inspection icon

Industrial Inspection

Using binarization to detect defects in high-speed manufacturing lines.

Unique Advantages of RAPIDSEA CUDA Plugins

icon

True Zero-Copy GPU Architecture

icon

CUDA-Based Parallel Pixel Processing

icon

Low CPU Resource Utilization Impact

icon

Seamless GStreamer Pipeline Integration

icon

Configurable and Flexible Control Parameters

icon

Designed for NVIDIA GPU-Based Systems

icon

Supports Object Detection & Segmentation

icon

CUDA-Powered GPU-Resident Video Processing

FAQs

RAPIDSEA GStreamer CUDA Color Manipulation plugins are GPU-accelerated video processing elements that perform brightness, contrast, binarization, and white balance adjustments directly on NVIDIA GPU memory using CUDA kernels within a GStreamer pipeline.

Struggling with CPU Bottlenecks in Video Processing?

Offload brightness, contrast, and segmentation tasks to RAPIDSEA’s CUDA-powered GPU plugins for efficient performance.

For further information on how your personal data is processed, please refer to the Rapidsea Privacy Policy.