Sharpness level

The Sharpness Level Check in SUPERWISE® provides an automated safeguard against out-of-focus or motion-blurred images that can compromise the accuracy of computer vision models. To quantify blurriness, this check utilizes the Variance of the Laplacian method, a high-performance image processing technique that measures the second derivative of image intensity. In practical terms, this algorithm detects the presence of edges and fine details: sharp images produce a high variance (higher scores), while blurry images exhibit low variance (lower scores). By setting a Minimal Sharpness Level threshold on a scale from 0 (Blurry) to 1000 (Sharp), users can programmatically reject low-quality inputs before they reach the model or validate that generated outputs meet professional standards. This rule is essential for maintaining consistent performance in mission-critical tasks—such as document processing, medical imaging, or autonomous navigation—where even slight blurring can lead to significant errors. The integrated testing pane allows for immediate validation, enabling users to upload samples and calibrate the exact threshold required for their specific use case.

Using the SDK

from superwise_api.models.guardrails.guardrails import BlurrinessGuard

image_sharpness_rule = BlurrinessGuard(name="Image Sharpness Guard", min_variance=100.0)