Sony FCB camera block
For extreme low‑light scenarios such as night‑time FPV UAV inspection, vehicle‑mounted night‑vision assistance and outdoor low‑light security surveillance, a qualified starlight full‑color USB camera can output colour footage free of prominent noise under near‑total darkness as low as 0.0001 Lux. It faithfully restores colour and textural details that ordinary infrared night‑vision systems tend to lose. Even so, many developers focus solely on image quality after receiving the camera, overlooking video compression formats — a core element in component selection.
Drawing on extensive real‑world test data collected in outdoor low‑light environments, this article delivers an in‑depth, multi‑dimensional analysis of five widely‑used formats for FPV starlight full‑color USB cameras: H.265, H.264, MJPEG, YUY2 and N12. It helps developers avoid parameter‑related misconceptions and select optimal solutions tailored to project requirements.
I. Maximised Performance Depends on Proper Video‑Format Selection
Choosing a format with an excessively high bitrate will saturate the wireless transmission bandwidth of FPV UAVs. This results in frequent video stuttering and stream disconnection during long‑range flight, while on‑board storage fills up within less than three days.
Over‑aggressive compression, on the other hand, smears fine low‑light noise into blocky mosaics, erasing distinguishable human silhouettes and vehicle licence plates. The accuracy of AI facial recognition may drop by more than 30 %.
Formats with poor platform compatibility can drive CPU usage on embedded on‑board boards to 100 %, triggering frame drops and even full failure of the vision system.
Designed to address such diverse low‑light‑scene requirements, Ruiloog Electronic’s FPV starlight full‑color USB camera natively supports all five mainstream formats: H.265, H.264, MJPEG, YUY2 and N12. It offers substantial flexibility for developers working on remote video transmission as well as local AI inference, eliminating the need for extra transcoding hardware for format adaptation.
II. Breakdown of the Five Mainstream Video Compression Formats
1. H.265 (HEVC) — High Compression, Saving Both Bandwidth and Storage
As the benchmark for modern high‑efficiency compression standards, H.265 delivers roughly 50 % higher compression ratio than H.264 while maintaining comparable subjective image quality. It is optimised for high‑definition high‑bitrate video transmission.
Advantages: Extremely high compression ratio and low bandwidth footprint. Recorded file sizes are approximately half those produced by H.264. Well‑suited for long‑range low‑bandwidth FPV UAV video transmission and 7‑by‑24‑hour continuous recording.
Disadvantages: Encoding demands substantial hardware computing power. Older low‑end embedded boards are prone to frame drops and stuttering when decoding 4K H.265 streams.
Use‑cases: Night‑time long‑range FPV UAV inspection, long‑duration outdoor‑security recording, real‑time remote preview over limited‑bandwidth networks, and workflows involving NVR storage and cloud‑based video upload.
2. H.264 (AVC) — Maximum Compatibility, the Universal Choice
A mature lossy compression standard deployed across the security industry for nearly two decades, H.264 enjoys native support across nearly all platforms, media players and embedded motherboards. It offers the broadest compatibility among video‑coding standards.
Note: Fine‑grained random noise in low‑light full‑colour footage pushes H.264 bitrates significantly higher. Reusing daytime bitrate settings often causes severe blocking artefacts at night; fine textures degrade into indistinct colour blocks due to insufficient bitrate allocation.
Advantages: Abundant hardware‑ and software‑decoding resources, low debugging overhead and rare platform‑compatibility failures, enabling fast project deployment.
Disadvantages: For equivalent perceived quality, bitrate and storage consumption are double those of H.265, consuming more valuable bandwidth during long‑range FPV transmission.
Use‑cases: General‑purpose video preview, routine recording, integration with legacy third‑party security platforms and projects prioritising rapid deployment over bandwidth constraints [1].
3. MJPEG (JPEG Streaming) — Simple, Reliable and Frame‑Independent
MJPEG encodes each frame independently as a JPEG image and outputs successive frames to form a video stream. No inter‑frame predictive information exists, making it the structurally simplest coding format.
Advantages: During high‑speed flight or heavy vibration of FPV UAVs, brief packet loss only corrupts the affected frame. Unlike H.264 / H.265, it avoids cascading corruption spanning dozens of consecutive frames. Video recovery is instantaneous, so critical snapshot opportunities are never missed.
Disadvantages: Very low compression efficiency. At 2K resolution, bandwidth usage exceeds three times that of H.265. High‑resolution output can saturate the USB bus and trigger packet loss.
Use‑cases: Point‑in‑time snapshots by FPV UAVs, AI image capture and workflows requiring frequent single‑frame extraction. Not recommended for sustained high‑bitrate streaming [2].
4. YUY2 — Raw Uncompressed Pixel Data, Preferred for AI Algorithms
A lossless raw pixel format preserving complete colour and luminance data straight from the image sensor without compression or transcoding. It represents one of the closest representations of native sensor output.
Advantages: Subtle edge details and shadow textures in low‑light footage remain intact without compression‑induced artefacts or noise. When fed directly into AI pipelines, it yields higher inference accuracy for night‑time facial recognition and obstacle detection compared with compressed H.264 streams.
Disadvantages: Massive raw data volume — nearly 300 MB per second at 2K resolution — imposes heavy requirements on USB bus bandwidth. It cannot be used directly for long‑term recording without compression.
Use‑cases: Machine‑vision development on Windows, edge‑gateway AI recognition and post‑processing workflows. Not intended for direct network streaming.
5. N12 (NV12) — Dominant Raw Format in Embedded Systems
A widely‑adopted YUV planar raw stream for industrial vision and embedded development, N12 is the native output format for most image ISP chips. Luminance and chrominance components are stored separately, retaining full image information while reducing overall data size.
Advantages: Data volume is one‑third smaller than YUY2. Hardware ISPs on mainstream embedded platforms such as Rockchip and HiSilicon provide native acceleration. On FPV on‑board embedded boards, CPU utilisation can drop by 40 %. Even low‑compute boards can smoothly execute low‑light‑image enhancement.
Disadvantages: As raw bitstream data, it cannot be opened by generic media players and requires decoding and rendering by application‑layer software. New developers need to invest time in adaptation and debugging.
Use‑cases: Linux‑based embedded development, SDK secondary‑programming and local real‑time image processing. Commonly deployed for on‑board AI devices on FPV UAVs and industrial low‑light visual inspection.
> Important Reminder: Fine random noise present in starlight‑grade low‑light footage raises the actual bitrate of all compression codecs. When using H.264 or H.265, increase the bitrate setting by approximately 30 % relative to daytime values. This eliminates night‑time blocking mosaics and preserves full‑colour clarity at minimal bandwidth cost.
FAQ
Q: For night‑time on‑board AI recognition with an FPV starlight full‑colour camera, should I choose YUY2 or N12?
A: Both are lossless raw‑image formats preserving shadow details in low‑light footage and well‑suited for AI inference. For Windows host‑PC development, YUY2 is preferred for broader compatibility and seamless integration with mainstream AI frameworks without extra format conversion. For Linux‑based embedded on‑board projects using Rockchip or HiSilicon chips, select N12. It reduces data size by one‑third compared with YUY2, lowers CPU load and benefits from native hardware‑ISP optimisation, enabling smoother execution on resource‑limited on‑board hardware.
Q: For night‑time starlight‑full‑colour recording, how substantial is the practical performance gap between H.265 and H.264?
A: At comparable subjective quality, H.265 saves roughly half the bandwidth and storage capacity — a tangible advantage for bandwidth‑ and storage‑constrained FPV UAV deployments. Especially for noisy night‑time starlight footage, H.265 handles fine noise far more efficiently than H.264 and avoids excessive bitrate overhead for noise preservation. Use H.265 whenever supported by on‑board decoding hardware. Fall back to H.264 only on legacy hardware lacking H.265 hardware decoding.
References
[1] H.264. Baidu Baike. https://baike.baidu.com/item/H.264/1022230
[2] Introduction to MJPEG Video Coding. CSDN Blog. https://blog.csdn.net/gqd0757/article/details/140108003