Image compression in photography is the system used to reduce file size through JPEG compression, RAW compression, HEIF/HEIC, lossless encoding, lossy encoding, chroma subsampling, quantisation and metadata choices. Search for “JPEG quality,” “RAW vs JPEG,” “lossless compression,” “photo compression,” “why JPEG loses quality,” or “best image export settings,” and the discussion often treats compression as a storage decision. It is also an information decision.
A lossless file can be reconstructed exactly from its compressed representation. A lossy file deliberately discards or approximates information judged less important to perception. JPEG can reduce fine chroma detail, quantise transform coefficients and create blocking or ringing at aggressive settings. Re-saving a JPEG can compound loss. RAW compression may be mathematically lossless, visually lossless or lossy depending on camera and mode. Two files with the same pixel dimensions can therefore contain different amounts and kinds of recoverable image information.
This is why JPEG, RAW, HEIF, photo compression, image quality, lossless compression, lossy compression, chroma subsampling, bit depth, export settings and compression artefacts belong to one mechanism. File size is not image quality, and pixel dimensions are not information content. The correct compression depends on whether the photograph is an archive master, web delivery, client proof, scientific record, print source or disposable preview.
Central proposition: compression changes representation. Lossless compression changes storage efficiency; lossy compression changes which measured or rendered information survives.
1. Compression Exploits Redundancy
Photographs contain repeated and predictable structure. Compression stores that structure more efficiently rather than representing every value independently.
2. Lossless Means Reversible
A lossless algorithm allows the original encoded data to be reconstructed exactly. File size falls without deliberate information deletion.
3. Lossy Means Approximation
A lossy algorithm discards or coarsens information to achieve greater reduction. The decoded image approximates the original rather than reproducing every encoded value.
4. JPEG Separates Brightness and Colour Information
JPEG commonly transforms RGB-like data into luminance and chroma components because human vision is often more sensitive to fine brightness structure than fine colour structure.
5. Chroma Subsampling Reduces Colour Resolution
Formats such as 4:2:0 or 4:2:2 store colour at lower spatial resolution than luminance. This can be efficient for photographs but problematic for coloured text, graphics and some keying workflows.
6. JPEG Works in Blocks
Classic JPEG processes small image blocks using frequency transforms. Heavy quantisation can make block boundaries visible.
7. Ringing Appears Near Strong Edges
Aggressive frequency compression can create halos or oscillating artefacts around high-contrast transitions. The edge remains recognisable while its local structure becomes artificial.
8. Mosquito Noise Is a Compression Artefact
Fine shimmering or speckled patterns around edges can come from compression rather than sensor noise. The repair is a better source or gentler encoding, not lower ISO.
9. Re-Saving JPEG Can Compound Loss
Decode a JPEG, edit it and compress it again, and new quantisation occurs. Repeated generations can progressively damage fine structure.
10. Editing Masters Should Avoid Unnecessary Generational Loss
Keep RAW, TIFF, PSD or another appropriate high-quality master and create JPEG delivery files from it rather than repeatedly editing the delivery copy.
11. JPEG Quality Numbers Are Not Universal
A quality value of 80 in one application need not correspond to 80 in another. Encoders use different quantisation tables and optimisation strategies.
12. File Size Depends on Image Complexity
A smooth sky compresses efficiently. Fine foliage, hair, grain and noise require more data at comparable quality because they contain less predictable high-frequency structure.
13. Noise Is Expensive to Compress
Random variation is difficult to predict. Denoising before export can reduce file size substantially because the encoder sees simpler structure.
14. Denoising Before Compression Changes Information Twice
The denoiser decides what variation is signal; the compressor then decides which remaining information can be approximated. A small web file may therefore be several inferential steps away from the RAW capture.
15. RAW Compression Is Not One Thing
Cameras may offer uncompressed RAW, lossless-compressed RAW and smaller lossy RAW variants. The exact trade-offs are manufacturer-specific.
16. Lossless RAW Can Save Space Without Sensor-Data Loss
Predictable numerical relationships can be encoded more efficiently while preserving exact decoded values.
17. Lossy RAW Can Be Operationally Excellent
Well-designed lossy RAW may preserve visually important latitude while reducing storage and increasing burst depth. Whether the compromise is acceptable depends on extreme recovery needs and archival requirements.
18. Bit Depth and Compression Are Different Axes
A 14-bit source can be compressed losslessly. An 8-bit JPEG can be lightly or heavily compressed. Numerical precision and compression ratio should not be conflated.
19. HEIF Can Store More Efficiently Than JPEG
Modern codecs can achieve comparable visible quality at smaller sizes or support higher bit depth and advanced metadata. Compatibility remains a practical constraint.
20. PNG Is Lossless but Not Always Efficient for Photographs
PNG excels with graphics, transparency and repeated flat regions. Natural photographic texture can produce much larger files than perceptually compressed formats.
21. TIFF Is a Container With Several Possibilities
TIFF can store high-bit-depth images and use different compression schemes. The extension alone does not tell you the exact encoding.
22. Metadata Can Be Lost Even When Pixels Look Fine
Export tools may strip EXIF, copyright, GPS, captions or colour profiles. Compression workflows therefore affect contextual information as well as visible pixels.
23. Removing the Colour Profile Can Change Appearance
An RGB file without its intended profile can be interpreted incorrectly by another application. The numerical pixels survive while their colour meaning becomes ambiguous.
24. Resizing Is Not Compression
Reducing pixel dimensions discards spatial samples. Compression encodes the remaining samples more efficiently. Web export often performs both, but they are different operations.
25. Cropping Is Not Compression Either
Cropping removes image area. It may reduce file size because fewer pixels remain, but its primary operation is selection rather than coding efficiency.
26. Sharpening Before Export Changes Compressibility
Strong sharpening adds high-frequency edge contrast, which can require more bits or create more visible ringing under heavy JPEG compression.
27. Web Images Need Delivery Optimisation
Huge camera files waste bandwidth when displayed small. Appropriate resizing, modern codecs and sensible quality settings improve page speed without visibly harming normal viewing.
28. Archive Masters Need Different Priorities
An archive values recoverability, metadata and future processing more than immediate download speed. RAW plus robust backups often serves this job better than a heavily compressed delivery JPEG.
29. Social Platforms Recompress
Uploading a carefully exported file does not guarantee the platform will preserve it. Services may resize, strip metadata and recompress for delivery.
30. Messaging Apps Can Recompress Again
A photograph forwarded through several services can become a later-generation derivative with less detail and metadata than the original.
31. Screenshots Create Another Derivative
A screenshot captures a rendered display representation, not the original file. Resolution, colour management and metadata can all change.
32. Compression Can Hide Editing Evidence
Heavy recompression can obscure subtle pixel-level traces used in image analysis. Forensic workflows therefore prefer original files and documented provenance.
33. Compression Can Also Create False Editing Clues
Block boundaries and ringing can resemble local manipulation artefacts. Analysts must distinguish codec behaviour from editing operations.
34. Scientific Images Need Controlled Export
If pixel values are measurements, lossy compression can change them. Scientific workflows should preserve appropriate lossless data and document any display derivatives separately.
35. Medical and Technical Images Have Their Own Standards
Diagnostic or technical imaging can use specialised formats and validated compression rules because visual information may carry high-stakes meaning. Ordinary social-media export habits do not belong there.
36. Printing Does Not Need Every Camera Pixel
Once sufficient resolution exists for the print size and viewing distance, additional pixels may not improve visible output. Compression should preserve the detail the printer and viewer can actually use.
37. Compression Quality Should Be Judged at Intended Size
A JPEG artefact obvious at 400 percent may be invisible in normal web display. Conversely, coloured text can show chroma damage at ordinary size even when a photograph looks fine.
38. There Is No Universal Best JPEG Quality Number
The optimal setting depends on encoder, image complexity, delivery size and tolerance for artefacts. Test the actual workflow rather than importing one magic number.
39. Smaller Is Not Automatically Worse
A well-encoded, correctly resized modern image can be far smaller than an oversized source while looking identical at intended display size.
40. Larger Is Not Automatically Better
A huge uncompressed file can contain blur, noise or weak colour. Storage volume is not photographic information.
41. Compression Is a Reader-Job Decision
A website needs speed. A client proof needs convenience. A gallery master needs high-quality output. An archive needs recoverability. A scientific record needs controlled values. One compression policy cannot optimise all five.
42. Keep the Original When the Future Is Unknown
Future displays, algorithms and printing methods may extract value from information that seems unnecessary today. Preserving the highest-value original protects optionality.
43. Delivery Copies Should Be Disposable
A web JPEG or messaging copy can be regenerated from the master. Treating the derivative as the archive reverses the information hierarchy.