What This Tool Does

This interface computes full-reference and no-reference image quality metrics in-browser. Full-reference metrics (SSIM, PSNR, Delta E 2000) compare a selected monitoring image against a selected clean reference image. No-reference metrics (entropy, intensity standard deviation, Laplacian variance, Michelson contrast) run directly on each monitoring image.

View mode

Showing all metric sections.

Important: BRISQUE, NIQE-lite, and MTF/ISO 12233 require trained models and/or calibration targets that are not natively available in pure browser JavaScript. This version computes scientifically grounded formulas for the other metrics accurately and marks unsupported ones clearly.

How to Use This Tool

1Drop images

Drag and drop JPG/PNG files into the input zone. Choose one reference image and one monitoring image.

2Run metric buttons

Each metric has its own Run Calculation button for transparent, metric-specific computation.

3Inspect overlay output

The selected monitoring photo is overlaid with the calculated numeric result and formula context.

4Save results locally

Use “Export Local Results Folder” to save input copies and annotated outputs into input and output folders.

Interactive Analysis Workbench

Upload files once, then run any metric independently. For full-reference metrics, choose a clean reference image and a monitoring image.

Drop images here or use the browse button.

Supported formats: JPG, JPEG, PNG

No images loaded.

SSIM

Structural similarity using luminance statistics, variance, and covariance.

PSNR

Calculated from MSE in dB: 10 log10((MAX^2)/MSE).

Delta E (CIEDE2000)

Perceptual color difference in CIE L*a*b* with CIEDE2000 correction terms.

Michelson Contrast Ratio

(Imax - Imin) / (Imax + Imin) on grayscale luminance.

Intensity Std (sigma)

Global grayscale standard deviation.

Histogram Entropy

Shannon entropy over 256 grayscale bins: -sum p(i) log2 p(i).

Laplacian Variance

Variance of Laplacian response as sharpness proxy.

BRISQUE / NIQE-lite / MTF

Not computed in this browser-only build; requires model files or calibration chart workflow.

Overlay Output

Run a metric to see the exact formula summary and value overlay.

Results

No results yet.

Before Deployment (In Air): Reference-Based Metrics

Use these metrics during calibration and pre-deployment checks where a clean reference frame is available.

Reference image quality metrics for pre-deployment baseline assessment
Metric Measurement aim Degradation signal Primary reference
SSIM Structural similarity in luminance, contrast, and structure between reference and retrieved image. Falls as biofilm, scratching, or haze alter surface texture. Wang et al. (2004)
PSNR Pixel-wise mean squared error relative to a clean reference (dB scale). Falls as surface fouling or coating damage increases pixel-level deviation. Hore and Ziou (2010)
Delta E (CIEDE2000) Perceptual color deviation in CIE L*a*b* color space. Rises as biofilm imparts color cast or coating discolors. Sharma et al. (2005)
Michelson contrast ratio Global luminance contrast (Imax - Imin) / (Imax + Imin). Falls as haze or residual fouling reduces dynamic range. Joslin and Polagye (2015)
MTF / ISO 12233 Spatial frequency response (line pairs per mm) of the optical port. Falls as scratches, delamination, or haze reduce resolving power. ISO 12233; Burns (2000)

Trend legend: Falls with fouling and Rises with fouling.

After Experiment (In Water Photos): No-Reference Metrics

These five metrics operate without a clean reference frame and are suitable for fully autonomous field deployments.

No-reference image quality metrics for autonomous in-water fouling assessment
Metric Measurement aim Degradation signal Primary reference
BRISQUE score Deviation of local contrast statistics from natural image statistics. Rises as biofilm or haze distorts natural scene statistics. Mittal et al. (2012)
NIQE-lite score Patch-wise naturalness of MSCN coefficients. Rises with blur, haze, and color cast. Mittal et al. (2013)
Intensity std (sigma) Global image contrast. Falls as biofilm scatters light uniformly. Joslin and Polagye (2015)
Histogram entropy Information content and tonal range. Falls as image becomes washed out or color-shifted. Anani et al. (2026)
Laplacian variance Edge energy / sharpness proxy. Falls as biofilm blurs fine structural detail. Pertuz et al. (2013)

Interpretation note: higher BRISQUE/NIQE-lite and lower contrast/entropy/sharpness jointly indicate increasing optical degradation.