Best Video Compression Algorithm

What is the best video compression algorithm for video display on Ethernet, best image quality?

If your goal is the best image quality over Ethernet (especially for live security / technical video):

H.265 (HEVC) is currently the best practical choice

JPEG XS is the best quality-first, low-latency choice if bandwidth allows

H.264 is the safe, widely compatible fallback

Why this question matters (Ethernet ≠ Internet)

On Ethernet (LAN, fibre, PoE networks), you usually care about:

Visual fidelity (edges, thermal gradients, fine detail)

Low latency (live viewing, PTZ control, analytics)

Predictable bandwidth

Encoder/decoder stability (edge devices, NVRs, VMS)

So “best” depends on whether quality, latency, or compatibility is king.

H.265 / HEVC – Best overall balance

Winner for most professional Ethernet video systems

Why it’s excellent

• ~40–50% better quality than H.264 at the same bitrate

• Preserves fine textures, thermal gradients, edges

• Mature, stable, widely supported in:

  • Cameras

  • NVRs

  • VMS platforms

• Handles:

  • 4K / 8K

  • Thermal video

  • Low-light noise better than H.264

Trade-offs

• Higher CPU/GPU load

• Slightly higher latency than JPEG-based codecs

Best choice for:Security cameras, thermal cameras, long fibre runs, PoE networks, mixed vendor systems

JPEG XS – Near-lossless, ultra-low latency

Image-purist’s dream

Why it’s special

• Visually lossless (often indistinguishable from raw)

• Sub-frame latency (microseconds, not milliseconds)

• No motion artifacts, no smearing

• Ideal for:

  • Critical monitoring

  • Analytics validation

  • Control rooms

Trade-offs

• Much higher bandwidth than H.265

• Less common support in cameras/NVRs

• Usually found in:

  • Broadcast

  • Defence

  • High-end industrial systems

Best choice for:Mission-critical live viewing where quality > bandwidth

H.264 / AVC – Still very good

The workhorse

Why it’s still used

• Extremely stable

• Universal compatibility

• Lower processing cost

• Predictable behaviour

Downside

• Noticeable blocking and blur at lower bitrates

• Less efficient with:

  • Thermal video

  • Fog, sand, rain

  • Night noise

Best choice for:Legacy systems, low-power cameras, universal compatibility

AV1 – Future king, not yet practical

Technically superior, operationally immature

• Better compression than H.265

• Open and royalty-free

• BUT:

  • Very heavy processing

  • Limited real-time camera support

  • Rare in security hardware (today)

Best choice for:Future VMS platforms, not current perimeter deployments

What NOT to use if image quality matters

❌ MJPEG – wastes bandwidth

❌ MPEG-4 Part 2 – obsolete

❌ Aggressive GOP tuning without scene awareness

 

Practical recommendation (real-world Ethernet networks)

Scenario	Best codec
Perimeter security cameras	H.265
Thermal cameras	H.265 (low GOP, high bitrate)
Live control room / analytics	JPEG XS
Mixed vendor systems	H.265 fallback to H.264
Long fibre runs	H.265

Pro tip (often overlooked)

Codec choice matters less than encoder tuning

To truly maximize quality:

• Use longer GOP only if motion is slow

• Avoid aggressive bitrate caps

• Prefer CBR with headroom for analytics

• Disable over-aggressive noise reduction before encoding

• For thermal: prioritize bit-depth preservation over frame rate

Bottom line

If you want me to pick one single answer:

H.265 (HEVC) is the best video compression algorithm today for high-quality video display over Ethernet.

 H.265 PROFILE & PARAMETER TUNING

Visible vs Thermal cameras (Ethernet, best image quality)

Visible-light cameras (Day/Night)

Primary objective: preserve edges, textures, faces, plates without motion smear.

Recommended H.265 settings

Parameter	Best Practice
Profile	Main or Main10
Bit depth	10-bit (critical for dusk/dawn gradients)
GOP length	1–2 seconds (25–50 frames @25 fps)
Frame rate	12.5–25 fps (avoid 30 unless needed)
Rate control	CBR with headroom
Bitrate	Higher than marketing tables suggest
B-frames	Enabled (unless ultra-low latency needed)
Noise reduction	Low before encoding
Sharpening	Mild, never aggressive

Why this works

• H.265 handles motion prediction extremely well

• Longer GOP is fine because visible scenes change slowly

• Over-NR destroys information before compression (fatal mistake)

Thermal cameras (this is different)

Primary objective: preserve thermal gradients, NOT pretty pictures.

Recommended H.265 settings for thermal

Parameter	Best Practice
Profile	Main10 (mandatory)
Bit depth	10-bit (non-negotiable)
GOP length	Short: 0.5–1 sec
Frame rate	8–12.5 fps (thermal physics limit)
Rate control	CBR or capped VBR
Bitrate	Higher than visible for same resolution
B-frames	Minimal or off
Noise reduction	OFF or extremely low
AGC	Scene-adaptive, slow response

Why thermal fails with bad settings

• Long GOP = thermal smearing

• B-frames = false temperature blending

• 8-bit encoding = banding & lost contrast

• “Pretty” NR = lost detection probability

👉 This is where many integrators unknowingly cripple thermal systems.

 

WAVELET vs H.265 (MPEG-style)

Quality, latency, analytics reality

Compression philosophy difference

Aspect	Wavelet (JPEG2000 / proprietary)	H.265
Compression method	Spatial frequency decomposition	Temporal + spatial prediction
Motion artifacts	None	Possible
Latency	Very low	Low–medium
Bandwidth efficiency	❌ Poor	✅ Excellent
Hardware support	Limited	Universal
Analytics compatibility	Medium	Excellent

When Wavelet wins

✔ No blocking artifacts✔ No motion prediction errors✔ Stable thermal gradients✔ Great for short-range, critical feeds

But…

• Bandwidth heavy

• Vendor-locked

• Weak VMS support

• Poor scaling beyond a few cameras

When H.265 wins

✔ Long perimeter lines✔ Fibre/Ethernet efficiency✔ Multi-vendor systems✔ AI analytics pipelines✔ Storage cost control

Modern reality:A well-tuned H.265 stream now beats wavelet in real-world usable quality, especially over distance.

Verdict

Wavelet is theoretically “pure”, H.265 is operationally superior.

For Namib-style long perimeters, H.265 is the only scalable choice.

 

CODEC SELECTION BY CAMERA TYPE

What you actually deploy

Fixed visible cameras

Best codec:H.265 Main10

Why

• Static background

• Long dwell times

• Predictable motion

• Excellent compression efficiency

Tuning focus

• GOP optimisation

• Bitrate stability

• Analytics-friendly noise profile

Fixed thermal cameras

Best codec:

H.265 Main10 (thermal-specific tuning)

Key focus

• Bit-depth preservation

• Short GOP

• No aggressive temporal filtering

Avoid

• MJPEG (wasteful)

• Over-compressed H.264

• Vendor “beautification” modes

PTZ cameras

Best codec:H.265 with adaptive GOP

Why

• Scene changes constantly

• Motion prediction must reset quickly

Tuning

• Short GOP during movement

• Longer GOP when stationary

• Dynamic bitrate scaling

Mission-critical / control room

Best codec (if bandwidth allows):

JPEG XS

Why

• Near-lossless

• Ultra-low latency

• Perfect for human verification

Reality check

• Rare in standard security ecosystems

• Best used selectively (not everywhere)

FINAL EXECUTIVE CONCLUSION

If you want one professional, defendable recommendation:

🏆 H.265 (HEVC), Main10 profile, correctly tuned per camera type, is the best video compression solution today for high-quality video over Ethernet.

JPEG XS is superior only where:

• Bandwidth is abundant

• Latency must be microscopic

• Camera count is limited

Video compression for high-quality video display over Ethernet (LAN)

For Ethernet-based video distribution (e.g., IP CCTV, industrial monitoring, control rooms), the dominant design constraint is not “internet streaming,” but predictable bandwidth, low latency, decoder reliability, and preservation of diagnostically relevant detail (edges, textures, gradients, and motion cues). In this context, modern inter-frame codecs such as H.265/HEVC generally provide the best operational balance between compression efficiency and perceptual quality, typically delivering substantial bitrate reductions versus H.264/AVC at similar subjective quality, while remaining widely supported in camera SoCs, NVRs, and VMS ecosystems (Dumić et al., 2022). 

However, “best image quality” is not solely a function of codec name; it is also determined by bit-depth, GOP structure, rate control, and pre-processing. For scenes with smooth gradients (dusk/dawn skies, fog layers, thermal palettes), higher bit-depth (e.g., HEVC Main10) can reduce visible banding and preserve gradation detail that is otherwise quantised away, improving perceived fidelity and post-processing headroom (Intel, 2017). 

For ultra-low latency and near-uncompressed visual fidelity, JPEG XS (ISO/IEC 21122) is designed specifically for visually lossless quality with minimal additional latency (often described in terms of added lines rather than full frames) and relatively lightweight implementations, making it attractive for control-room grade “glass-to-glass” monitoring over IP where bandwidth is available (JPEG Committee, n.d.; ISO, n.d.). 

Finally, codec selection must consider downstream video analytics. Research on surveillance-style object detection indicates that detection performance can remain robust under moderate compression, but degrades meaningfully at higher compression levels, particularly in difficult conditions such as poor lighting, complex scenes, or fast-moving targets; this makes encoder tuning (bitrate caps, GOP length, quantisation) a first-order design lever for systems that depend on reliable detection rather than “pretty pictures” (O’Byrne et al., 2022). 

References (Harvard)Dumić, E. et al. (2022) Subjective Quality Assessment of H.265 versus H.264 Video Coding for High-Definition Video Systems.Intel (2017) Enable Billions of Colors with 10-Bit HEVC.ISO (n.d.) ISO/IEC 21122 (JPEG XS) overview information.JPEG Committee (n.d.) JPEG XS White Paper.O’Byrne, M. et al. (2022) Impact of Video Compression on the Performance of Object Detection Systems for Surveillance Applications. arXiv.

ExCo decision matrix (quick, defensible)

Decision lens	H.265 (HEVC Main/Main10)	H.264 (AVC)	JPEG XS (ISO/IEC 21122)	AV1
Best image quality per Mbps	✅ Best practical	◻ Good	❌ Needs more Mbps	✅ Excellent (theory)
Latency	Low–Medium	Low	✅ Ultra-low	Medium–High
Ecosystem support (CCTV/VMS/NVR)	✅ Very strong	✅ Universal	◻ Limited / niche	◻ Emerging
Analytics friendliness	✅ Strong if tuned	✅ Strong if tuned	✅ Excellent	◻ Depends on deployment
Compute load (edge devices)	Medium–High	Low–Medium	Medium	High
Best-fit use case	Default for LAN CCTV + thermal	Legacy / compatibility	Control-room, mission-critical feeds	Future-facing platforms

Executive call:

• Default: H.265 Main10 (best quality over Ethernet at scale).

• Control-room “gold feed” (few cameras): JPEG XS when bandwidth allows. 

Practical bitrate tuning guide (desert / fog / night / sand)

(Rule-of-thumb starting points for Ethernet LAN; adjust after scene tests.)

Assumptions: 1080p or 4MP, 12.5–25 fps visible, 8–12.5 fps thermal, continuous streaming, analytics on.

A) Visible cameras (H.265 Main10)

• Clear day (stable scene):

  • 1080p @ 12.5–15 fps: 2–4 Mbps

  • 1080p @ 25 fps: 4–8 Mbps

• Night / low light (sensor noise rises):

  • 1080p @ 12.5–15 fps: 4–8 Mbps

  • 1080p @ 25 fps: 8–12+ Mbps

• Fog / mist (lots of low-contrast “moving texture”):

  • Add +30–60% bitrate vs clear day OR shorten GOP to protect detail (fog is compression-hostile).

• Sand/dust/wind-blown clutter (high motion micro-texture):

  • Add +40–100% bitrate vs clear day (worst-case), and keep noise reduction conservative.

    
    

Key tuning knobs

• Shorten GOP when motion/clutter spikes (PTZ moves, wind-blown vegetation, dust).

• Prefer CBR with headroom for analytics stability.

• Keep camera-side NR low; excessive NR deletes real detail before encoding.

B) Thermal cameras (H.265 Main10, thermal-specific)

• Treat thermal as “gradient-critical.” Use Main10 and avoid aggressive temporal tricks.

• Typical starting points (common thermal resolutions):

  • 640×512 @ 8–12.5 fps: 1.5–4 Mbps

  • 1280×1024 @ 8–12.5 fps: 4–10 Mbps

• In fog/sea mist (coastal) thermal contrast may drop; paradoxically you often need more bitrate to preserve subtle gradients.

4) Codec choice mapped to AI detection accuracy (what to expect)

What research suggests (surveillance context):

• Detection can be robust under moderate compression, enabling large bitrate reductions without major loss in detection performance. 

• Detection degrades sharply at high compression (strong quantisation / low bitrate), and degradation is worse in poor lighting, complex scenes, and fast motion. 

Practical mapping (field reality)

• H.265 (well-tuned, moderate compression):

  ✅ Best “scale + analytics” choice.

  Risk appears when you push bitrates too low: blocking, smear, edge loss → missed/late detections.

• H.264 (moderate compression):

  ✅ Often very stable for analytics, but needs more bitrate for same visual quality.

• JPEG XS:

  ✅ Best for analytics validation and critical detection zones (near-uncompressed-like). 

• AV1:

  Potentially excellent compression, but current real-time CCTV hardware support is the constraint (practical deployment factor).

  
  

“Analytics-safe” encoder rules

1. Don’t chase the lowest Mbps—chase the lowest Mbps that preserves edges.

2. For night + fog + motion, budget bitrate headroom and shorten GOP.

3. If analytics is mission-critical, consider a two-stream strategy:

   • Stream 1 (recording/operator): efficient H.265

   • Stream 2 (analytics): higher bitrate / shorter GOP (or JPEG XS for limited channels)

 

Video Compression for High-Quality Video over Ethernet (LAN)

Executive Summary

For Ethernet-based video systems (security, perimeter, control-room environments), H.265 (HEVC), preferably Main10, delivers the best overall image quality per Mbps, with mature ecosystem support and proven performance in both visible and thermal imaging. When correctly tuned, H.265 preserves diagnostically relevant detail while enabling scalable deployment across long fibre or PoE networks.

For mission-critical, ultra-low-latency viewing, JPEG XS provides near-lossless quality but requires significantly higher bandwidth and is best reserved for limited, high-value channels.

Recommended Codec Strategy

Default standard (all cameras):

• H.265 (HEVC), Main10

• Tuned per camera type and environment

Selective premium use (control room / validation feeds):

• JPEG XS (where bandwidth allows)

Why H.265 Main10

• ~40–50% bitrate saving vs H.264 at equivalent quality

• Better preservation of:

  • Fine edges

  • Low-contrast detail

  • Thermal gradients

• Strong compatibility across:

  • Cameras

  • NVRs

  • VMS platforms

• Proven analytics performance when not over-compressed

Risk if Poorly Tuned

• Over-compression → missed detections

• Long GOPs → motion smear (especially PTZ & thermal)

• Aggressive noise reduction → permanent information loss

Mitigation: enforce encoder profiles, bitrate floors, and acceptance testing.

Executive Decision (Plain Language)

Adopt H.265 Main10 as the standard codec across the Ethernet video network, with JPEG XS reserved for limited, mission-critical live feeds.

SITE ACCEPTANCE & COMMISSIONING CHECKLIST

Codec & Image Quality Validation

1️⃣ Camera Configuration (Before Go-Live)

☐ Codec = H.265

☐ Profile = Main10

☐ Bit depth = 10-bit

☐ Frame rate appropriate to sensor (thermal ≤12.5 fps)

☐ GOP length documented

☐ Noise reduction = low / conservative

☐ Sharpening = mild only

2️⃣ Bitrate & Scene Stress Testing

Test each camera under:

☐ Clear day

☐ Night (low light / high gain)

☐ Fog / mist / dust / sand

☐ High motion (people + vehicles)

☐ PTZ movement (if applicable)

Pass criteria

• No blockiness on edges

• No thermal smearing

• No banding in gradients

• Stable bitrate (no starvation)

• Analytics detections unchanged vs baseline

3️⃣ Thermal-Specific Validation

☐ No long-GOP ghosting

☐ No temperature “blending” on moving targets

☐ Detection distance unchanged when bitrate reduced slightly

☐ AGC response not oscillating

4️⃣ Analytics Compatibility Check

☐ Person detection stable

☐ Vehicle detection stable

☐ False positives unchanged or reduced

☐ No detection delay during scene change

5️⃣ Control Room / Operator Review

☐ Live latency acceptable

☐ PTZ response acceptable

☐ Operator confirms visual clarity at decision distance

☐ Optional: JPEG XS feed evaluated (if deployed)

6️⃣ Documentation & Sign-Off

☐ Codec settings captured per camera

☐ Bitrate baselines recorded

☐ Environmental test results logged

☐ ExCo / client sign-off obtained

Final Operational Rule (Worth Framing)

Never reduce bitrate to the point where analytics or human judgment is compromised storage is cheaper than missed detections.

Video Compression Algorithms for High-Quality Video Transmission over Ethernet Networks

1. Introduction

Ethernet-based video transport has become the dominant medium for security, industrial, and critical-infrastructure surveillance systems due to its scalability, determinism, and integration with IP-based analytics platforms. Unlike internet streaming, where adaptive delivery and buffering dominate design considerations, Ethernet video systems prioritise image fidelity, low and deterministic latency, analytical integrity, and predictable bandwidth utilisation. Consequently, the selection and configuration of video compression algorithms play a decisive role in operational effectiveness rather than mere bandwidth reduction.

Modern inter-frame compression standards, most notably H.265/HEVC, have emerged as the preferred solution for high-quality video over Ethernet, offering substantial efficiency gains over legacy codecs such as H.264/AVC while maintaining compatibility with contemporary cameras, network video recorders (NVRs), and video management systems (VMS). However, alternative codecs, including wavelet-based compression and JPEG XS, retain relevance in specific low-latency or near-lossless scenarios.

2. Compression Paradigms

2.1 Inter-frame (Temporal) Compression

H.264 and H.265 exploit temporal redundancy by predicting motion between frames using block-based motion estimation. H.265 introduces larger coding tree units, improved motion vectors, and enhanced entropy coding, resulting in superior compression efficiency and reduced artefacts at comparable bitrates.

2.2 Intra-frame (Spatial / Wavelet) Compression

Wavelet-based codecs (e.g., JPEG2000 derivatives) compress each frame independently, avoiding motion artefacts and providing visually stable output. This approach offers excellent image consistency but at the expense of significantly higher bandwidth requirements and limited ecosystem support.

2.3 Visually Lossless Low-Latency Compression

JPEG XS was designed specifically for professional video transport over IP, providing visually lossless quality with sub-frame latency. Unlike traditional mezzanine codecs, it balances compression efficiency with extremely low delay, making it suitable for control rooms and critical decision environments.

3. Image Quality Considerations

Image quality in Ethernet video systems is influenced by:

• Bit depth (8-bit vs 10-bit)

• GOP (Group of Pictures) length

• Rate control strategy

• Pre-encoding image processing (noise reduction, sharpening)

• Scene dynamics and environmental conditions

For thermal imaging and low-contrast environments (fog, dusk, desert haze), 10-bit encoding is critical to preserve subtle gradients and detection cues. Over-aggressive noise reduction or bitrate limitation irreversibly removes information prior to compression, negatively affecting both human interpretation and AI analytics.

4. Impact on Video Analytics

Empirical research demonstrates that AI-based object detection systems tolerate moderate compression but experience nonlinear degradation once quantisation thresholds remove edge and contrast information. This effect is amplified in low-light, high-noise, or cluttered scenes. Consequently, encoder configuration must be treated as part of the analytics pipeline rather than a downstream optimisation.

5. Conclusion

H.265 (HEVC), when configured using Main10 profiles and environment-appropriate parameters, represents the optimal balance between image quality, bandwidth efficiency, and system compatibility for Ethernet-based surveillance systems. JPEG XS provides superior quality and latency characteristics in limited, high-value scenarios, while wavelet codecs remain niche due to scalability constraints.

EXCO / BOARD-LEVEL DECISION MATRIX

Criterion	H.265 (Main10)	H.264	Wavelet	JPEG XS	AV1
Image quality per Mbps	⭐⭐⭐⭐⭐	⭐⭐⭐	⭐⭐	⭐⭐⭐⭐⭐	⭐⭐⭐⭐⭐
Latency	Low	Low	Very Low	Ultra-Low	Medium
Thermal suitability	Excellent	Moderate	Good	Excellent	Good
AI analytics support	Excellent	Good	Moderate	Excellent	Emerging
Bandwidth efficiency	Excellent	Moderate	Poor	Moderate	Excellent
Ecosystem maturity	Very High	Universal	Low	Low–Medium	Low
Operational scalability	High	High	Low	Low	Low
Recommended use	Default	Legacy	Niche	Premium feeds	Future

Executive Resolution:

Adopt H.265 Main10 as the standard video compression codec across all Ethernet-based camera systems, with JPEG XS reserved for limited mission-critical live feeds.

PRACTICAL BITRATE & ENCODER TUNING GUIDE

(Field-tested starting values – adjust after site testing)

1. Visible Cameras (1080p–4MP)

Clear Day

• 12.5–15 fps: 2–4 Mbps

• 25 fps: 4–8 Mbps

Night / Low Light

• 12.5–15 fps: 4–8 Mbps

• 25 fps: 8–12+ Mbps

Fog / Mist / Coastal Haze

• Increase bitrate by +30–60%

• Reduce GOP length

• Avoid aggressive NR

Sand / Dust / Vegetation Movement

• Increase bitrate by +40–100%

• Prefer CBR with headroom

2. Thermal Cameras

Resolution	FPS	Recommended Bitrate
640×512	8–12.5	1.5–4 Mbps
1024×768	8–12.5	3–7 Mbps
1280×1024	8–12.5	4–10 Mbps

Mandatory settings

• H.265 Main10

• Short GOP (0.5–1 s)

• Minimal B-frames

• NR OFF or very low 

CODEC vs AI DETECTION ACCURACY MAPPING

Key Observations

• AI performance remains stable under moderate compression

• Sharp degradation occurs once:

  • Edges blur

  • Contrast collapses

  • Motion smear appears

• Thermal analytics are especially sensitive to:

  • Bit depth reduction

  • Long GOPs

  • Temporal smoothing

Codec Impact Summary

Codec	Analytics Reliability	Notes
H.265 (well tuned)	⭐⭐⭐⭐⭐	Best balance for scale
H.264	⭐⭐⭐⭐	Needs more bitrate
Wavelet	⭐⭐⭐	Stable but inefficient
JPEG XS	⭐⭐⭐⭐⭐	Ideal but bandwidth-heavy
AV1	⭐⭐⭐⭐	Limited real-time support

Best-Practice Analytics Rules

1. Preserve edges > beauty

2. Shorten GOP in dynamic scenes

3. Budget bitrate for worst-case night/fog

4. Consider dual-stream strategy:

   • Stream A: efficient recording

   • Stream B: analytics-optimised

SITE ACCEPTANCE & COMMISSIONING CHECKLIST

Configuration

☐ Codec = H.265

☐ Profile = Main10

☐ Bit depth = 10-bit

☐ GOP documented

☐ NR conservative

Environmental Testing

☐ Day

☐ Night

☐ Fog / dust / wind

☐ High motion

☐ PTZ movement

Performance Validation

☐ No blocking

☐ No thermal smear

☐ Detection distance unchanged

☐ Analytics stable

Documentation

☐ Bitrates logged

☐ Profiles recorded

☐ Acceptance signed

FINAL PROFESSIONAL CONCLUSION

H.265 (HEVC), Main10 profile, correctly tuned per environment and sensor type, is the best video compression solution currently available for high-quality video transmission over Ethernet networks.

JPEG XS should be treated as a precision instrument, not a default invaluable where latency and fidelity trump bandwidth, but impractical at scale.

Tinus Diedericks

CEO of Timeless Technologies