The Universal Image Quality Checker & Forensic Fidelity Studio is an advanced, privacy-first computer vision utility designed for photographers, e-commerce managers, and web performance engineers. Executing 100% inside your browser using HTML5 Canvas pixel analysis, the tool quantifies image acutance using discrete Laplacian variance ($\sigma^2$), detects JPEG 8×8 DCT macroblock boundary discontinuities and chroma ringing, calculates dynamic information richness via Shannon entropy ($H$), and flags blown specular highlights and crushed shadows. With multi-image batch ingestion of up to 250+ files, interactive high-pass edge heatmaps, a 200%/400% zoom loupe lens, Core Web Vitals LCP readiness grades, and 1-click CSV/JSON/clipboard audit reporting, it provides an objective, client-side standard for certifying digital image fidelity with zero file uploads.
The discrete Laplacian operator convolves the luminance channel with the 3x3 kernel:
Kernel = [[0, 1, 0], [1, -4, 1], [0, 1, 0]]
This acts as an isotropic 2nd-order derivative, amplifying high-frequency spatial gradients (edges).
In in-focus images, sharp transitions produce wide gradient swings yielding high variance ($\sigma^2 > 500$). Defocused or motion-blurred captures lack high frequencies, dropping variance below 50.
Standard JPEG compression splits the raster into periodic $8 \times 8$ pixel macroblocks and applies Discrete Cosine Transforms before quantizing coefficients. Under heavy compression, high frequencies are zeroed out, creating abrupt stepped intensity discontinuities at block boundaries $x \equiv 0 \pmod 8$.
Our algorithm computes the ratio of inter-block boundary luminance steps against intra-block pixel variation to quantify visible grid blocking and chroma ringing.
Shannon information entropy measures the average information density in the grayscale histogram:
H = -∑ P(i) log2 P(i) (for i = 0 to 255)
An 8-bit image with rich textures, smooth gradients, and wide dynamic range approaches $7.2 - 7.8$ bits. Heavily posterized, over-compressed, or flat solid images have narrow histogram distributions resulting in low entropy ($< 5.5$ bits).
All raw pixel conversions, matrix convolutions, and forensic calculations execute 100% locally on your machine via JavaScript HTML5 Canvas `ImageData` pixel buffers. No pictures, metadata, or diagnostic logs are ever transmitted across network sockets or saved on external servers.
