AllOnline Stories · Vol. II · The Photographic Language
JPG
Photography learned how to forget. How human perception, 8×8 blocks and frequency mathematics let millions of colours travel around the world at a fraction of their original size.
Scroll into the photograph
01 · THE PROBLEM
Before JPG, a digital photograph was simply too expensive.
By the late 1980s, storing and transmitting continuous-tone images meant moving enormous amounts of information. The challenge was not drawing more pixels: it was deciding which visual information could disappear without the eye missing what mattered.
1986 JPEG group work begins1992 the baseline standard is publishedToday it remains one of the universal languages of digital photography
02 · PERCEPTION
JPG begins by compressing the observer.
Human vision notices changes in luminance much more precisely than small variations in colour. JPEG exploits that asymmetry: it separates brightness from chroma so it can spend more information on what we tend to notice first.
Y · LUMINANCECb + Cr · COLOUR
CHROMA SUBSAMPLING4:2:0
Luminance keeps the full sample resolution; colour information is shared across groups of pixels.
KEY IDEAJPG does not simply ‘see worse’. It decides where to spend precision according to how we usually perceive a photograph.
03 · 8 × 8
The photograph breaks into frequencies.
In the most common DCT-based JPEG, the image is processed in blocks. A discrete cosine transform stops asking ‘what colour is this pixel?’ and starts asking ‘which frequency patterns explain this block?’
SPACE8 × 8 pixels
Light and colour still live as positions inside a small region.
FREQUENCY64 coefficients
Slow variations dominate the upper-left; toward the opposite corner, increasingly fine detail appears.
04 · QUANTIZATION
This is where the loss that makes JPG small happens.
Frequency coefficients are divided by quantization values and rounded. Much fine detail becomes zero. Those zeros are easy to compress, but the discarded information can no longer be reconstructed exactly.
DC123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263
42 / 64coefficients survive
The more aggressive the quantization, the more high-frequency detail disappears and the more repetitive the remaining information becomes.
05 · QUALITY
Quality is not a universal number. It is a negotiation.
The quality control in an exporter is not part of JPEG as an absolute scale. Each encoder maps that number to its own tables and decisions. What matters is the result: how much detail survives and what it costs to keep it.
JPEG QUALITY82
high detail≈ 38% of visual budget
1050100
This visualization is educational: it recreates blocking, smoothing and detail loss to explain the concept, not the exact output of a specific encoder.
The best setting is the point where lowering quality any further starts to cost more visually than it saves in bytes.
06 · ORDERING THE SILENCE
After removing detail, JPG organizes the zeros.
Quantized coefficients are traversed in zig-zag order from low to high frequencies. This tends to group long runs of zeros, which can then be encoded compactly together with DC differences and entropy coding.
AC →42, 34, 28, 18, 12, 8, 5, 3, 2, 1, 0, 0, 0, 0…
When fine detail disappears, the end of the sequence fills with zeros. Compression takes advantage of that statistical silence.
07 · FILE ANATOMY
A .jpg is not a bag of pixels. It is a structured sequence.
A JPEG stream is recognized by markers. Between Start Of Image and End Of Image, instructions for the decoder, file context and one or more scans carry the compressed image.
→→→→→→
FF D8 · SOIThe sequence begins.
The Start Of Image marker opens the stream. From here, the decoder walks through segments that explain how to interpret what comes next.
01 · CONTEXTJFIF · EXIF · ICC
Information that helps describe, exchange or interpret the file.
02 · INSTRUCTIONSDQT · SOF · DHT
Tables and parameters that prepare the decoder to reconstruct the image.
03 · DATASOS → SCAN
The compressed information that ultimately becomes visible pixels again.
Think of a JPG as a conversation: first it identifies itself, then explains how to read its data and finally delivers the compressed image.
08 · ARRIVING IN PASSES
A photograph can also reveal itself gradually.
A baseline JPEG usually completes the image information in a single encoded pass. A progressive JPEG organizes multiple scans: first a global approximation can appear, then later refinements add detail.
BASELINEline by line
The visible version grows like a window completing itself from the top.
PROGRESSIVEan idea → more detail
The full composition appears earlier and gains precision in successive passes.
09 · THE TRACE OF LOSS
Compress too far and the mathematics starts to show.
Blocking, ringing and texture loss are not random mistakes. They are clues that too much information has been removed, or that the image has gone through too many cycles of decoding, editing and recompressing.
GENERATION 01
MASTER8TH RECOMPRESSION
BLOCKINGRINGINGTEXTURE LOSS
Generation loss depends on settings and content; this scene exaggerates the effect so it stays readable during the scroll.
11 · WHERE IT FITS
JPG is brilliant when the image behaves like a photograph.
Its strength appears in scenes with continuous tones, natural textures, light, skin, landscape and product photography. In graphics with transparency, small text or extremely hard edges, its compression decisions stop working in your favour.
The right question is not ‘is JPG old?’It is ‘does this image benefit from what JPG knows how to sacrifice?’
12 · PROFESSIONAL WORKFLOW
Compress at the end. Keep the original first.
Think of JPG as the last part of a chain: first preserve a useful source, then adapt dimensions and quality to the destination, remove what should not travel and validate the result where it will actually be seen.
13 · AFTER JPG
Thirty years later, web photography speaks more than one language.
WebP and AVIF can reduce weight further or add capabilities everyday JPEG does not have. JPEG XL aims for a much broader evolution. But JPEG keeps one advantage that is hard to overstate: decades of compatibility, tools, cameras, workflows and accumulated knowledge.
Static photography, near-universal support and a pipeline understood by almost every tool.
Modern alternatives when the main goal is reducing transfer or expanding delivery capabilities.
A modern format with advanced features, but without the classic JPEG’s universal web adoption yet.
A format does not die when another one is better on a single metric.It disappears when it stops solving a problem better than the cost of replacing it.
From perception to action
Keep what matters. Compress what does not.
Adjust quality, remove unnecessary metadata, resize for the destination and convert your images with AllOnline JPG tools directly in the browser.
AllOnline Stories · Understanding the format changes how you export it.