Gaussian

Overview

The Gaussian Blur sample demonstrates how to apply Gaussian smoothing to images using CV-CUDA.

Usage

Basic Usage

Apply Gaussian blur with default parameters:

python3 gaussian.py -i input.jpg

The output will be saved as cvcuda/.cache/cat_blurred.jpg with a 9×9 kernel and sigma=1.5.

Custom Output Path

Specify the output file:

python3 gaussian.py -i image.jpg -o blurred.jpg

Command-Line Arguments

Argument	Short Form	Default	Description
--input	-i	tabby_tiger_cat.jpg	Input image file path
--output	-o	cvcuda/.cache/cat_blurred.jpg	Output image file path

Implementation

Complete Code

The entire sample is remarkably concise:

args: argparse.Namespace = parse_image_args("cat_blurred.jpg")
# docs_tag: begin_read_image
input_image: cvcuda.Tensor = read_image(args.input)
# docs_tag: end_read_image

# docs_tag: begin_gaussian_blur
# 1. Perform a gaussian blur on the image
output_image: cvcuda.Tensor = cvcuda.gaussian(
    input_image,
    (9, 9),
    (1.5, 1.5),
)
write_image(output_image, args.output)
# docs_tag: end_gaussian_blur

That’s it! Just three key steps:

1. Load the image from disk

2. Apply Gaussian blur

3. Save the result

Expected Output

The output shows the image with Gaussian blur applied, smoothing the image while preserving overall structure:

Original Input Image

Output: Gaussian Blurred

CV-CUDA Operators Used

Operator	Purpose
cvcuda.gaussian()	Apply Gaussian blur with specified kernel and sigma

Common Utilities Used

• read_image() - Load image into CV-CUDA tensor

• write_image() - Save CV-CUDA tensor as image

See Also

• Hello World Sample - Uses Gaussian blur in pipeline

• Segmentation Sample - Uses blur for background effects

• cvcuda.bilateral_filter() - Edge-preserving alternative

• Common Utilities - Helper functions