Image glossary

HTML element for format fallbacks and art direction.

Loads the first `<source>` whose format the browser supports, falling back to the `<img>` element. Standard pattern: AVIF → WebP → JPG. Also used to deliver different crops per viewport (art direction).

Android icon system with foreground and background layers.

Since Android 8 (2017). The launcher masks the icon according to vendor style (round, squircle, Pixel). The foreground layer needs 25 % safe-zone padding — see Maskable Icon.

Adobe Illustrator's native vector format, internally PDF with Illustrator layers.

Since Illustrator 9 (1999) PDF-compatible — every AI file is also a PDF, plus Illustrator-specific binary editing information. Not web-suitable; export to SVG or PNG/WebP for web delivery.

Fourth image channel storing per-pixel transparency (0–255).

RGB + A. 0 = fully transparent, 255 = fully opaque. PNG and WebP support alpha; JPG does not.

Layering images while honoring transparency.

"Source Over" is the default operation: the new image partially covers the old. Premultiplied vs. straight alpha are the two variants — premultiplied avoids halo artifacts.

Image alternative text — required for accessibility and image SEO.

Screen readers announce the `alt` value. Search engines weight it as the most important image context. Descriptive, concise, without "image of …". Empty `alt=""` is allowed for purely decorative images.

Animated PNG — Mozilla's backwards-compatible PNG animation.

Proposed by Mozilla in 2004, rejected by the PNG committee, implemented anyway in Firefox, Safari, Chrome. Officially normalized as PNG-3 in 2025. Lossless animation with full 24-bit color and 8-bit alpha.

iOS home-screen icon — 180×180 px, mandatory for iPhone PWAs.

`<link rel="apple-touch-icon" href="…">`. Without it iOS shows a blurry screenshot of the page on the home screen instead of a clean app symbol.

Width-to-height ratio — e.g. 16:9, 4:3, 1:1, 9:16.

16:9 is YouTube and widescreen, 9:16 is Reels/Stories/TikTok, 1:1 is the Instagram feed square, 4:5 is Instagram portrait. When cropping, the aspect ratio drives composition.

Royalty-free video codec from the Alliance for Open Media — basis of AVIF.

Finalized as royalty-free in 2018. YouTube and Netflix deliver AV1 streams to supporting clients. Encoding is compute-heavy; SVT-AV1 is today's fastest open-source encoder.

Next-gen format after WebP (2019) — up to 50% smaller than WebP.

AV1 Image File Format, built on the AV1 video codec from the Alliance for Open Media. More efficient than WebP, with 12-bit color depth and HDR support. Browser support: Chrome 85+, Firefox 113+, Safari 16.4+.

Stepped transitions in gradients when color depth is too low.

Classically visible in 8-bit sky gradients. 10-bit encoding (HDR formats) eliminates it; dithering masks it.

Encoding of binary data as ASCII text — 33% size overhead.

Standard method for embedding binary data in text formats like HTML, CSS, JSON. Three bytes become four ASCII characters. Worth it only for very small assets (under 5 KB).

Color filter arrangement on camera sensors — RGGB mosaic.

Bryce Bayer, Kodak 1976. Each sensor pixel sees only one color; the missing two are interpolated via demosaicing. Fujifilm's X-Trans uses a different 6×6 pattern.

Data per second — central quantity for video and audio compression.

Measured in kbps or Mbps for video. Low bitrates produce block and banding artifacts; high bitrates inflate file size. Constant Bitrate (CBR) vs. Variable Bitrate (VBR) are the two main strategies.

Mixing mode between two layered images — multiply, screen, overlay, …

Mathematical operation per pixel pair. In Photoshop since 1994, in CSS since 2015 (`mix-blend-mode`). Important for photo composites, double exposures, UI effects.

Binary data representation in the browser — basis of every local image pipeline.

A `Blob` is an immutable byte stream with a MIME type. `File` is a `Blob` with a filename. `URL.createObjectURL(blob)` turns blobs into image sources.

Low-quality image placeholder — preview trick against layout shift.

An upscaled thumbnail (often 10 × 10 px as Base64) shows immediately and cross-fades to the real image. Eliminates perceived loading delay.

Windows bitmap (1987) — uncompressed legacy format.

Came out of the joint OS/2 project by IBM and Microsoft. Today only relevant in embedded systems, Windows system APIs and as an intermediate format. A 1920×1080 photo as BMP is 6 MB.

Google's HTTP successor to gzip — 15–25 % more efficient on text.

Brotli, specified as RFC 7932 in 2015. Universal browser support via `Accept-Encoding: br`. Higher CPU cost when encoding, identical decoding speed to gzip.

HTTP header that tells the browser how long assets may be cached.

For immutable images with hashed filenames: `Cache-Control: public, max-age=31536000, immutable`. Saves round-trips entirely for returning visitors.

Browser API for drawing and reading pixels in JavaScript.

Enables JNRT Pixel's client-side image processing: an image is drawn onto a `<canvas>`, manipulated and exported as a compressed blob — all in-browser, no server.

Green-screen technique — pixels of a defined color become transparent.

Classic compositing technique from film and live TV. Today done in OBS, DaVinci Resolve, After Effects via algorithm or ML models. Green is used because human skin contains little green and sensors have the highest resolution in the green channel.

Stores color information at lower resolution than luminance.

The eye is more sensitive to luminance than color. Standard notation: 4:4:4 (full color), 4:2:2 (half horizontal color), 4:2:0 (half horizontal + vertical — web default).

Measures how much layout elements shift during loading.

Images without explicit `width`/`height` attributes cause layout shifts and worsen the CLS value. Google recommends CLS below 0.1.

Subtractive print color space (cyan, magenta, yellow, key/black).

Not for web! Browsers misinterpret CMYK JPGs. If a PSD or TIFF is in CMYK, convert to RGB before web export. Print shops still want CMYK artwork as PDF/X.

COder/DECoder — algorithm for compression and decompression.

Image examples: JPEG, WebP, AV1, HEVC. Video: H.264, H.265, VP9, AV1. Codec is the algorithm, container the file format (MP4, WebM, HEIF).

Bits per pixel channel — determines available color steps.

8 bits = 256 values per channel (web standard), 10 bits = 1024 (HDR TV), 12 bits = 4096 (DCI cinema), 16 bits = 65,536 (photo RAW). Low color depth produces banding in gradients.

Defines how RGB values translate to actual displayed colors.

International Color Consortium Profile. Web standard is sRGB. Adobe RGB and Display-P3 cover wider gamuts but are ignored by older software.

File wrapper that bundles one or more codec data streams.

HEIF can carry HEVC (=HEIC) or AV1 (=AVIF). MP4 can hold H.264, H.265 or AV1. WebM is VP8/VP9/AV1. Container ≠ codec.

Google's metric set for user experience — LCP, INP, CLS.

Official Google ranking factor since 2021. Images directly affect LCP (load time) and CLS (layout stability).

Image visible in the initial viewport — usually the LCP element.

Critical images: never lazy-load, always mark with `fetchpriority="high"`, optionally preload. Below-the-fold images on the other hand should be lazy.

Image inline in HTML/CSS as `data:image/...;base64,...`.

Saves one HTTP request but bloats HTML/CSS. Sensible for tiny images (<2 KB) or LQIP placeholders. Larger images as data URI hurt caching and compression.

The mathematical transform that makes JPG compression work.

Decomposes an 8×8 pixel block into 64 frequency coefficients — low frequencies in the center, high frequencies at the edge. JPG quantizes high frequencies more aggressively because the eye perceives them less.

`<img decoding="async">` lets the browser decode the image non-blocking.

Prevents decoding from blocking the main thread. Often saves 50–200 ms on mobile devices. Default in modern browsers, but explicit setting doesn't hurt.

Lossless compression from LZ77 + Huffman — base of PNG, ZIP, gzip.

Phil Katz, 1993. Dictionary-based LZ77 finds repeating byte sequences; Huffman coding compresses the frequencies. Built into practically every lossless format and HTTP content encoding.

Reconstructing missing color values from Bayer sensor data.

Turns a RAW with one color per pixel into a complete RGB image via interpolation. Algorithms range from simple bilinear to adaptive edge-aware methods like AHD.

Apple's wide-gamut color space — ~25 % larger than sRGB.

Apple uses Display P3 since iPhone 7, iPad Pro and Retina MacBook Pro. Stronger red and green saturation than sRGB. PNG and WebP can carry P3 as an embedded ICC profile.

Deliberate noise to avoid color banding.

With reduced color palettes (GIF, indexed PNG) dot patterns simulate what the palette cannot show directly. Floyd-Steinberg is the most famous algorithm.

Adobe's open RAW format (2004) — Digital Negative.

ISO 12234-2 since 2009. Aims to replace the proprietary RAW chaos. Leica, Hasselblad and Pentax adopted; Canon, Nikon, Sony did not. Lightroom can convert existing RAWs to DNG on import.

Print dots vs. screen pixels per inch.

DPI (Dots Per Inch) is a print metric, PPI (Pixels Per Inch) a screen metric. Often used interchangeably in web contexts; technically PPI is correct. Print standard quality: 300 DPI at final print size.

Encapsulated PostScript — legacy vector exchange of the DTP era.

Defined by Adobe in 1987. Until the early 2000s the standard for vector exchange in print workflows. Marked deprecated by Adobe in 2017. For modern pipelines replaced by SVG (web) or PDF/X (print).

Metadata block in images: camera, lens, exposure, GPS.

Exchangeable Image File Format, standardized in 1995 by the Japanese JEIDA. A modern smartphone camera embeds 100–500 KB of EXIF in every photo.

Tiny icon in the browser tab — `/favicon.ico` plus modern PNG link tags.

De-facto standard since Internet Explorer 5 (1999). Today's best practice: one 32×32 PNG plus 180×180 Apple touch icon plus 512×512 PWA icon plus classic `favicon.ico` as a fallback.

First Contentful Paint — when does the first text or image content appear?

Web vital that shapes perceived loading. Good values: <1.8 s. Improved by critical CSS, preloads and a fast TTFB.

HTML attribute for loading priority — `high` for LCP images, `low` for below-the-fold.

Available in Chrome, Edge, Firefox since 2023. `fetchpriority="high"` on the hero image measurably improves LCP. `low` on carousel images prevents bandwidth contention.

Fabrice Bellard's universal codec suite for video, audio, image — since 2000.

Converts between practically every format. Inside YouTube, Netflix, every browser video decoder and countless toolchains. Nicknamed the "Swiss Army knife of media".

Non-linear brightness encoding — sRGB uses gamma ~2.2.

Humans perceive brightness non-linearly. Gamma encoding allocates bit depth so more values describe dark tones — precisely where the eye is more sensitive.

Animation format from 1987, max 256 colors, binary transparency.

Graphics Interchange Format, developed by CompuServe. Outdated LZW compression and 8-bit color space make it markedly weaker for animation than animated WebP today — typically 3–4× larger.

Location coordinates in photo metadata — a privacy risk when sharing.

Smartphones store the capture location to the meter. Sharing photos on Instagram, Twitter or via email often unintentionally exposes the location. Strip before sharing.

HTTP standard compression since 1992 — Deflate plus headers.

GNU zip. Negotiated by browsers via `Accept-Encoding: gzip`; saves 60–80 % on text (HTML, CSS, JS, JSON). Pointless for already-compressed images.

Web video codec standard since 2003 — supported everywhere.

ITU-T H.264 / MPEG-4 Part 10. Patent-encumbered (MPEG LA) but universally hardware-accelerated. YouTube, Netflix, Zoom — almost every non-AV1 video stream is H.264.

H.264 successor — 50 % more efficient but patent-encumbered.

High Efficiency Video Coding. Three patent pools (MPEG LA, HEVC Advance, Velos Media) made HEVC unattractive for open-source software. Never established on the web; AVIF and AV1 video are the royalty-free alternatives.

High Dynamic Range — brighter highlights and deeper shadows than SDR.

HDR requires at least 10 bits per channel, a wide-gamut color space (Rec. 2020) and an OETF such as PQ (HDR10) or HLG. Among image formats AVIF and JPEG XL support HDR; WebP and classic JPG do not.

Apple's default since iOS 11 — 30–50% smaller than JPG.

High Efficiency Image Container, based on the HEVC/H.265 video codec. Default on iPhones since 2017. Often incompatible outside the Apple ecosystem — Windows 10 needs a paid extension.

Distribution chart of pixel brightnesses from 0 to 255.

Critical for exposure assessment. Spikes at the left edge (underexposure) lose shadow detail; spikes at the right (overexposure) blow out highlights. Lightroom and all RAW tools show it prominently.

Perceptually intuitive color models: Hue, Saturation, Lightness/Value.

More intuitive for designers than RGB. CSS supports `hsl()` and `hsla()` natively. HSV (also HSB) is used in many color pickers, including our color palette extractor.

Multiplexing protocol — many images in parallel over one TCP connection.

In all major browsers since 2015. Eliminates the HTTP/1.1-era domain-sharding anti-pattern. HTTP/3 with QUIC is also widespread in 2026.

Variable-length coding — frequent symbols short, rare symbols long.

David Huffman, 1952. Used in JPEG, Deflate and many other formats as the entropy coding stage after the actual data transform. Mozjpeg optimizes Huffman tables per image.

Embedded color-space description — required for correct rendering.

International Color Consortium spec since 1993. Tells the decoder "the RGB values in this file mean Display P3" (or sRGB, Rec. 2020 etc.). Without a profile browsers assume sRGB.

Apple's multi-resolution icon container for macOS apps.

Originated in the resource forks of classic Mac System 7, became a standalone file with Mac OS X 10.0. Holds PNG and JPEG 2000 sublayers for every resolution from tab icon to 1024×1024.

Windows icon container (1985) — still the legacy favicon fallback.

Multi-resolution container: a single file can hold several icon sizes and bit depths. Established on the web as /favicon.ico since Internet Explorer 5 (1999).

On-the-fly image transformation at the edge — Cloudinary, Imgix, Vercel Image.

Serves the optimal size, format (AVIF/WebP/JPG) and quality per request. Saves storage compared to manual variant generation. Vercel/Next.js ships it built-in.

XML sitemap extension — tells Google which images belong to which pages.

Listed inside the `<image:image>` block of the sitemap. Especially helpful for lazy-loaded and CMS-generated images that are hard to crawl in the HTML.

GPU-backed image buffer for fast canvas drawing.

`createImageBitmap(blob)` parses and decodes the image asynchronously on a worker thread and returns a GPU-ready handle. Significantly faster than classic `<img>` decoding for bulk operations.

Classic command-line suite for image processing since 1987.

Can read and write practically every format. Unbeatable for script workflows and batch conversions. Somewhat slower than Sharp/libvips for large pipelines.

Interaction to Next Paint — Core Web Vital since March 2024.

Replaces FID. Measures the longest delay between user interaction and visible UI response. Good values: <200 ms. Improved by reducing long main-thread tasks.

Browser API for visibility detection — basis of modern lazy loading.

Fires events when an element enters or leaves the viewport. Replaces the old scroll-listener hacks. Today's standard pattern for image lazy-loading polyfills.

Press photo metadata — caption, author, keywords, location.

International Press Telecommunications Council, standardized since 1990. Publishers, photo agencies and CMSes read IPTC headlines, captions and keywords automatically. Embedded in JPG and TIFF.

JPEG File Interchange Format — the actual container behind "JPG".

Eric Hamilton, C-Cube Microsystems 1992. JPEG standardizes the compression algorithm, JFIF the file structure. Today usually combined with embedded EXIF.

Wavelet-based JPEG successor (2000), dominant in cinema and medicine.

ISO/IEC 15444. Uses Discrete Wavelet Transform instead of DCT. Never adopted on the web, but mandatory for digital cinema projection (DCI/DCP) and DICOM medical workflows.

Modern JPEG successor (ISO 18181, 2021) with lossless JPG re-compression.

Merger of the research codecs PIK (Google) and FUIF (Cloudinary). Can shrink existing JPGs by ~20 % without pixel loss, supports HDR and progressive loading. Removed from Chrome in 2022, reinstated in 2025.

Microsoft's HD Photo format (2006), ISO 29199, never made it to the web.

Evolved from Windows Media Photo. Internet Explorer 9 supported it, no other browser followed. Definitively gone from the mainstream with Edge's Chromium switch in 2020.

Lossy image format (1992), the photo standard on the web.

Joint Photographic Experts Group, 1992. Splits images into 8×8 blocks and applies Discrete Cosine Transform (DCT) plus quantization to drastically reduce file size at the cost of invisible detail. Web sweet-spot quality: 78. No alpha channel.

Visible compression traces from overly aggressive JPG.

Block artifacts (8×8 squares), mosquito noise (ringing around edges) and banding (steps in gradients) appear when quality is set too low or when an image is re-encoded as JPG multiple times in a row (generation loss).

Fully encoded video frame anchoring the delta frames that follow.

Called an I-frame in H.264, H.265, AV1. P- and B-frames only store the difference to the nearest keyframe. Same principle in animated WebP/AVIF/GIF; more keyframes = better seek, larger file.

Khronos texture container for GPU textures — loaded straight into VRAM.

Holds GPU-native compressed textures (BC, ETC, ASTC) without runtime re-encoding. KTX2 (2020) adds Basis Universal supercompression with ZSTD. The preferred texture container in game engines, glTF and WebGPU.

Load images only when they scroll into the viewport.

Natively supported via `loading="lazy"` since 2019. Saves initial bandwidth and improves LCP. Hero images should explicitly use `loading="eager"`.

Most important Core Web Vitals metric — time to largest visible element.

Measures the render time of the largest content element in the viewport. Target: under 2.5 seconds. Images often cause the LCP and need to be loaded with appropriate priority.

Reference PNG library since 1995 — built into virtually every system.

In every browser, every OS, every image processing program. Together with zlib the basis of the entire PNG world.

Streaming image processing library with low memory footprint.

Processes images as pipelined streams rather than full in-memory buffers. On 100-MP images often 10× faster and 30× more memory-efficient than ImageMagick. The basis of Sharp.

Google's reference implementation of the WebP codec.

Built into every modern browser. CVE-2023-4863 (heap buffer overflow) caused the broadest emergency browser patch in web history in 2023.

Pixels stay byte-exact, only storage becomes more efficient.

PNG, WebP-Lossless and AVIF-Lossless work losslessly — no pixel is altered. Achieved through entropy coding (Deflate, Huffman) and prediction models.

Deliberately discards information the eye perceives poorly.

Uses psychovisual models to reduce data the human eye barely registers. JPG, WebP-Lossy and AVIF work this way. Invisible at the right quality setting (78–82 for JPG).

JSON-based vector animation format — by Airbnb 2015.

Exported from After Effects. Rendered as SVG/Canvas at runtime in the browser. Far more efficient than animated GIFs or videos for complex vector animations, often <30 KB.

Lempel-Ziv-Welch — the GIF algorithm with a long patent history.

Terry Welch, 1984. Patent-encumbered (Unisys) until 2004, which sparked PNG's creation in 1994 as a patent-free replacement. Free today and used in GIF, TIFF and PostScript.

Grayscale map that controls per-pixel visibility — white visible, black hidden.

Standard for non-destructive editing in Photoshop, Affinity and GIMP. CSS has `mask-image` as the web equivalent. With PWA maskable icons the mask is set by the OS.

PWA icon with safe-zone padding for Android launcher masks.

Standard since Android 8 (2017). The icon fills the entire 108-dp canvas; the OS masks it round, square or squircle depending on the launcher.

Mozilla encoder that produces JPG files 5–15% smaller than libjpeg.

Open-source fork of libjpeg-turbo (2014). Adds trellis quantization, optimized Huffman tables and a better default quantization matrix — fully compatible with any JPG decoder.

Mozilla's optimization encoder versus the standard JPEG encoder.

At identical decode compatibility, mozjpeg delivers 5–15 % smaller files via trellis quantization, optimized Huffman tables and a better default matrix. libjpeg-turbo is faster, mozjpeg more efficient.

MPEG-4 Part 14 — the dominant web video container.

Holds H.264, H.265 or AV1 plus audio (AAC, Opus). MOV (Apple's QuickTime container) is binary near-identical. For web delivery of looping videos the right choice instead of GIF.

Smoothing of sensor noise — important for high-ISO captures.

Classic via median filter, modern via AI models such as DeepPRIME (DxO), Denoise AI (Topaz) or Lightroom's own AI denoise. Trade-off: less noise vs. preserved detail.

Canvas rendering in a web worker — prevents UI blocking.

Allows image processing in a worker thread. Important for browser tools like our compression pipeline so the UI stays interactive during computation.

Preview image for social-media cards — typically 1200 × 630.

Facebook standard, adopted by Twitter, LinkedIn, Slack, Discord. Referenced via `<meta property="og:image">`. Should be 1.91:1, JPG or PNG, under 2 MB.

HDR industry standard for VFX and 3D rendering — by Industrial Light & Magic 1999.

Half-float 16-bit per channel (IEEE 754) stores physical light values. Multi-layer container for render passes (diffuse, specular, z-depth). Today standard at Pixar, Disney, Netflix and in Unreal/Unity.

EXIF field that controls display rotation — culprit of the portrait-photo gotcha.

Value 1 = normal, 6 = rotate 90° CW, 8 = rotate 90° CCW. Ignoring the tag causes iPhone photos to appear sideways.

Portable Document Format (1993) — universal document container.

John Warnock's Camelot project. ISO standard 32000 since 2008. Combines vector, raster, text with embedded fonts and metadata. PDF/A is the archive variant, PDF/X the print variant.

Lossless image format (1996) with alpha-channel transparency.

Portable Network Graphics, 1996, designed as a patent-free GIF successor. Lossless compression via Deflate, full 24-bit color, partial transparency. Ideal for logos, diagrams, screenshots, pixel art. For photos, three to ten times larger than JPG.

`<link rel="preload" as="image">` loads hero images before the HTML parse completes.

Important for LCP images embedded via CSS background — otherwise the browser discovers them late. Also works for responsive sets via `imagesrcset` and `imagesizes`.

Adobe Photoshop Document — master format with layers, masks, smart objects.

Thomas Knoll, first version 1990. Keeps the entire editing session: pixel data, adjustment layers, smart objects, effects. PSB is the large-format variant (up to 300,000 × 300,000 px).

Web app installable like a native application.

Coined by Alex Russell in 2015. Combines Service Workers (offline cache), Web App Manifest (icons + theme) and push notifications. JNRT Pixel is installable as a PWA.

Rounding of DCT coefficients — the actual loss in JPG.

Each frequency coefficient is divided by a quantization table and rounded to an integer. High frequencies are quantized aggressively; low frequencies stay nearly intact. This is where the irretrievable loss happens.

Unprocessed sensor data directly from the camera.

Vendor-specific: Canon CR2/CR3, Nikon NEF, Sony ARW, Olympus ORF, Fujifilm RAF, Panasonic RW2. 12–16 bits per channel, maximum dynamic range, room for non-destructive exposure and white-balance edits.

UHD/HDR color space — the target of modern HDR pipelines.

Specified by the ITU for 4K and 8K television. Used in HDR10, Dolby Vision and AVIF HDR workflows. Only wide-gamut display hardware can show its full range.

Recomputing pixels when scaling — Lanczos, bicubic, bilinear.

Lanczos gives the sharpest results (Photoshop's default), bicubic is a good all-rounder, bilinear is fast but soft, nearest neighbor preserves pixel-art look.

Resolution-aware delivery via `srcset`, `sizes`, `<picture>`.

Goal: every browser only receives the pixels it actually needs. Typically saves 50–70 % of bytes on mobile. Combined with format fallback for maximum efficiency gain.

Displays with double (or higher) pixel density.

Apple's marketing term for displays with 2× or 3× DPR. For sharp rendering, a 600 px container must load a 1200 px image, otherwise the result looks soft.

Additive color space of red, green, blue — the web's standard model.

Each pixel as three values: red, green, blue. With an alpha channel it becomes RGBA. CSS notation `rgb(255, 128, 0)` or `#FF8000`. For wide-gamut images the ICC profile decides which concrete colors are meant.

Run-Length Encoding — primitive lossless compression for solid areas.

Encodes repetitions compactly: "20× white" instead of 20 white bytes. Built into BMP, TGA, ICO and many 90s games. Almost always replaced by Deflate today.

Classic composition: place main subjects on the third lines.

The image is divided into 3×3 mentally. Subjects at the intersections (power points) feel more dynamic than centered ones. Our cropper overlays the grid automatically.

Structured image description as JSON-LD or microdata.

`ImageObject` markup tells search engines which image belongs to a page, its resolution and creator. Improves Google image search and featured-snippet selection.

Browser proxy between web app and network — enables offline.

Introduced from 2014. Intercepts network requests and can serve responses from cache. Prerequisite for PWA offline functionality.

Node.js image library built on libvips — the de-facto standard.

Fast (40–60 % faster than ImageMagick), produces WebP, AVIF, JPG, PNG from pipelines. Used by Vercel/Next.js Image Optimization, Cloudinary workers and many CMS platforms.

Tells the browser which image width to expect per viewport.

Goes hand in hand with `srcset`. Example: `sizes="(max-width: 600px) 100vw, 50vw"`. Without `sizes` the browser often picks unnecessarily large variants and wastes bandwidth.

Launch image when opening an installed PWA.

Android generates it automatically from the manifest icon and `background_color`. iOS additionally requires `apple-touch-startup-image` links per screen size — the PWA Generator covers all of them.

HTML attributes for serving correctly sized images per device.

With `srcset` you list image variants at different resolutions; the browser auto-picks the optimal one for viewport and DPR. `<picture>` adds format switching for AVIF/WebP fallbacks.

The web's standard color space since 1996 — default for every browser.

HP and Microsoft defined sRGB as the safe common denominator: a relatively narrow gamut that displayed reasonably consistently on every monitor of the time. Today the implicit default of every web image.

Vector image format — infinitely scalable, described in XML.

Scalable Vector Graphics. Defines images via paths and shapes instead of pixel rasters. Perfect for logos and icons (often under 5 KB), stylable via CSS, animatable, accessible. Not suitable for photos.

Truevision Targa (1984) — game industry standard for textures.

32-bit RGBA and a simple RLE decoder made TGA the choice of classic game engines (Wing Commander, Diablo, Quake, Half-Life). Still active in Substance, Quixel and many game engines in 2026.

Color of the browser status bar on mobile tabs and PWAs.

`<meta name="theme-color" content="#d4a017">`. Makes the browser header match the brand color. Optionally per light/dark mode via `media` queries.

Tag-based archive standard since 1986 — libraries, print, microscopy.

Tagged Image File Format, originally at Aldus for scanner workflows, with Adobe since 1994. Lossless, with BigTIFF can exceed 4 GB. The Library of Congress and many government archives use TIFF as their master format.

Mapping HDR data into SDR display-capable values.

Necessary when an HDR image plays back on a classic 8-bit sRGB display or on the web. Algorithms like Reinhard, ACES or Hable balance highlights, mid-tones and shadows.

Time to First Byte — server response time to the first HTML byte.

An early indicator for LCP. Good values: <200 ms. If TTFB is slow, no amount of image shrinking helps — the server is the bottleneck.

X's (Twitter) own link-preview variant — Summary vs. Summary Large Image.

Own `<meta name="twitter:*">` tags; falls back to Open Graph images by default. Summary Large Image requires at least 300 × 157 px, recommended 1200 × 628.

Classic sharpening technique — difference to a blurred copy.

Borrowed from analog enlarger workflows. Looks more natural than simple edge enhancement because it emphasizes mid-tone contrast. Standard in Photoshop, Lightroom and Capture One.

Google's royalty-free video codec — predecessor of AV1.

Replaced VP8 for YouTube streaming. Comparable efficiency to HEVC without patent issues. Never extended into a still-image format — Google went straight from VP9 to AV1/AVIF.

Visible brand name or logo overlay as a branding and anti-theft measure.

Classically bottom-right with reduced opacity. Press photos often place it centered so cropping cannot remove it. Invisible digital watermarks (steganography) are a separate class.

Mathematical decomposition into frequency and spatial parts — basis of JPEG 2000.

Unlike DCT with its fixed block size, the Discrete Wavelet Transform (DWT) operates on the whole image with multiple detail layers. Avoids the typical JPEG block artifacts but is more compute-intensive.

JSON file that registers your PWA as an installable app.

`<link rel="manifest" href="/manifest.json">`. Defines name, icons, theme and background color, display mode, start URL. Our Icon Studio and PWA Generator emit the snippet for you.

Browser runtime for C/C++/Rust code at near-native speed.

Lets complex encoders like libwebp, Mozjpeg, libavif run in the browser at performance comparable to server tools. The WASM file loads once per session.

Low-level browser API for video and image encoder/decoder access.

Available since Chrome 94 (2021). Allows programmatic access to VideoFrame streams and hardware-accelerated codecs. Foundation of modern browser video editors and fast frame extraction.

GPU-accelerated 3D and 2D graphics in the browser via OpenGL ES.

In all major browsers since 2011. Basis for Three.js, A-Frame, Google Maps. WebGL 2 with modern features is widespread since 2017. Successor: WebGPU.

Modern GPU API for the browser — successor to WebGL.

Shipped in Chrome 113 (2023), Safari Tech Preview (2024), Firefox (2025). Offers compute shaders and lower overhead than WebGL. Used for browser LLMs and fast image processing.

Google's open video container — VP8/VP9/AV1 + Vorbis/Opus.

Introduced together with WebP in 2010. Available in Chrome, Firefox, Edge and all Chromium browsers. Safari also supports WebM since Big Sur (2020).

Modern image format by Google (2010) — 25–35% smaller than JPG.

Built on VP8 video codec technology. Supports lossy + lossless + animation + alpha channel in one format. Universal browser support since 2020.

Adobe standard for extended image metadata in XML.

Extensible Metadata Platform, introduced 2001. Structured storage of editing history, licensing info, tags, copyright information.

Luma/chroma color space — basis of nearly every video and JPEG codec.

Y carries luminance, Cb and Cr the color differences. Because the eye is more sensitive to brightness than color, chroma subsampling (4:2:0 etc.) can heavily compress chroma channels without visible loss. JPEG, MPEG, H.26x and AV1 all use YCbCr internally.

Grayscale channel storing the distance of each pixel to the camera.

From 3D render pipelines (multi-layer OpenEXR) and modern smartphones (iPhone portrait mode, Google Pixel). Enables post-shot bokeh, AR compositing and 3D effects. HEIC can carry a depth map as a separate auxiliary layer.

Open-source library for Deflate — the software basis of PNG, gzip, ZIP.

Jean-loup Gailly and Mark Adler, 1995. Compact, BSD-style license, available in practically every OS and language. Stable for over three decades and has carried the entire web compression ecosystem.