top of page
RAW, YUV & JPEG Capture.png

RAW, YUV & JPEG Capture: Understanding Camera Data Formats

Camera systems can produce and capture image data in several different formats. The choice between RAW, YUV, and JPEG affects image quality, processing requirements, bandwidth, storage, and how the camera integrates with the rest of an embedded vision system.

Understanding these formats is important when designing and debugging camera pipelines.​

​

What Is RAW Capture?

RAW data is the image information captured directly from the image sensor before most image processing has been applied.

For a typical Bayer-pattern sensor, RAW data contains individual red, green, and blue pixel measurements rather than a finished color image.

A simplified pipeline is:

Image Sensor → RAW Data → ISP → YUV / RGB → Application

RAW formats are commonly represented using bit depths such as RAW8, RAW10, RAW12, or RAW14, depending on the sensor.

RAW capture is particularly useful for:

  • ISP tuning

  • Image quality analysis

  • Sensor evaluation

  • Computer vision

  • Custom image processing

  • Camera pipeline debugging

Because RAW data has not gone through the complete ISP pipeline, it provides greater control over how the image is processed.​

​

What Is YUV Capture?

YUV is a family of image formats that separates image information into luma (Y) and chroma (U/V) components.

Unlike RAW, YUV data typically represents an already processed image. The sensor data may have passed through ISP functions such as demosaicing, white balance, color correction, noise reduction, and tone mapping.

A simplified pipeline is:

Image Sensor → RAW → ISP → YUV → Application

Common YUV formats include YUV420, YUV422, NV12, and NV21.

YUV is widely used in embedded vision because it provides a good balance between image quality, processing requirements, and memory bandwidth.

​​​

What Is JPEG Capture?

JPEG is a compressed image format. A camera system can take processed image data and compress it into a JPEG image.

A typical pipeline is:

Image Sensor → RAW → ISP → YUV / RGB → JPEG Encoder → JPEG

JPEG compression significantly reduces the amount of data that needs to be stored or transmitted.

This makes JPEG useful for applications such as:

  • Still image capture

  • Image storage

  • Remote monitoring

  • Network transmission

  • Image logging

JPEG is typically lossy, meaning some image information is discarded during compression. The amount of compression can be adjusted depending on the desired balance between image quality and file size.

​​​

Comparing RAW, YUV, and JPEG

Each format serves a different purpose.

RAW provides direct access to sensor data and is best suited for ISP development, image quality analysis, and applications requiring control over the image-processing pipeline.

YUV is commonly used for processed video. It is well suited for real-time video, displays, computer vision, and AI applications.

JPEG is designed primarily for efficient image storage and transmission. Its compressed size makes it useful when bandwidth or storage capacity is limited.

The best format depends on the requirements of the camera system and the processing that happens after capture.

​​​

Capture Formats in Linux

On embedded Linux systems, camera capture is commonly handled through V4L2 (Video4Linux2).

A camera device may expose different capture formats depending on the hardware and driver. For example, a system could provide RAW10 for sensor-level capture, NV12 for processed video, and JPEG for compressed still images.

Applications can use V4L2 to select the desired format, resolution, frame rate, and other capture parameters.

The available formats depend on the sensor, ISP, camera driver, and processing platform.

​​​

Why Capture Format Matters

The selected format directly affects system performance.

RAW data can require significant memory bandwidth because it contains high-bit-depth sensor data and is generally uncompressed. However, it provides valuable information for advanced image processing.

YUV formats are often more efficient for real-time video and computer vision applications. Formats such as NV12 are commonly supported by hardware video and AI processing pipelines.

JPEG reduces storage and transmission requirements but requires JPEG encoding and decoding when images are processed.

Choosing the appropriate format can help reduce unnecessary processing and improve overall system performance.

​​​

RAW Capture for ISP Tuning

RAW capture is especially important during ISP tuning and camera development.

Engineers can capture RAW frames and evaluate how the ISP processes sensor data. This allows functions such as auto exposure, auto white balance, color correction, noise reduction, HDR, sharpening, and tone mapping to be evaluated and adjusted.

RAW capture can also help identify whether an image-quality problem originates from the sensor, optics, or image-processing pipeline.

​​​

YUV Capture for Computer Vision

YUV is commonly used when camera output needs to be processed in real time.

Computer vision and AI applications may receive YUV frames after the ISP has completed image processing. The frames can then be resized, converted, or passed directly to an AI accelerator depending on the platform.

Using a suitable YUV format can reduce unnecessary conversions and improve overall system efficiency.

​​​

JPEG Capture for Image Storage

JPEG is useful when the primary requirement is to save or transmit images rather than process every frame.

For example, an embedded device may continuously analyze a YUV stream for computer vision while capturing a JPEG image when an event occurs.

This allows the system to maintain a real-time processing pipeline while using compressed images for storage or communication.​

​

Choosing the Right Format

The appropriate capture format depends on the application and system architecture.

Important considerations include:

  • Required image quality

  • Processing requirements

  • Memory bandwidth

  • Storage capacity

  • Network bandwidth

  • Frame rate

  • ISP capabilities

  • Computer vision requirements

  • Hardware acceleration

For camera development, supporting multiple capture formats can also be useful. Engineers may use RAW during development and tuning, YUV for real-time processing, and JPEG for image storage or transmission.

​​​

Conclusion

RAW, YUV, and JPEG represent different stages and approaches to camera image capture. RAW provides direct sensor data for image processing and ISP development, YUV is widely used for processed video and real-time vision applications, and JPEG provides compressed images for storage and transmission.

Understanding how these formats move through the camera pipeline is essential for developing and debugging embedded vision systems.

​​​

Choosing the right capture format is an important part of designing an efficient camera pipeline, particularly when balancing image quality, ISP processing, memory bandwidth, and computer vision requirements. Explore our hardware design services to learn how we support camera integration, ISP development and tuning, embedded Linux pipelines, and complete embedded vision hardware development.

bottom of page