Image Processing for next-generation cameras

While computational cameras are set to disrupt the conventional camera environment, it is still not easy to integrate new optical and software designs into the existing ecosystem. With the many new optics, sensors and processors coming to market, the heterogeneous environment means ISPs need to be re-engineered with every new product; needless to say, this quickly becomes a resource-consuming process.

The solution is CRISP:
the Computationally Reconfigurable Image Signal Platform

Conventional camera pipeline

Separate processes aggregate errors

  • Each imaging process is an IP block, separately integrated to optics and hardware, which vary across devices.
  • The RAW imaging data is referred to only at the outset; each process then creates errors that carry over across the pipeline.
  • These errors are aggregated, creating suboptimal image & use of resources. For OEMs, this means a painstaking integration and testing, often delaying production.

CRISP

A unified framework optimizes for quality

  • CRISP is a software image reconstruction engine that incorporates all processes into one unified framework.
  • The Reconstruction Engine iterates based on the RAW imaging data, ensuring errors are minimized.
  • The framework’s objective is to optimize final image quality based on input variables. This leads to higher quality images, simple integration, and a better use of resources.

A software system that is optimized holistically for final image quality means that errors are minimized and changes can be implemented on the fly. Instead of working linearly and compounding the errors at each step, CRISP dynamically optimizes for final image quality and balances all the different functions.

This is especially useful for companies working on computational cameras. Whether they are focused on innovative optics or new applications, their ISP integration work can be minimized thereby allowing them to funnel energy into their core mission. The availability of a software ISP for next-generation cameras will accelerate innovation and push the limits of image quality.

CRISP means: Higher quality images & Simpler integration

CRISP is a cross-market, licensable software ISP that can provide higher quality images & easily adapt to:

New Optics

  • Stereo cameras
  • Array cameras
  • Lenslet arrays
  • Light-field cameras
  • Optical masks
  • Coded-aperture
  • Etc.

New Sensors

  • Bayer filter
  • Foveon X3
  • 3CCD
  • Thermography
  • Multispectral imaging
  • Medical scoping
  • Astronomy, etc.

New Applications

  • Improving images captured in low-light
  • Improved dynamic-range & HDR
  • Depth-mapping & 3D
  • Refocus capabilities
  • Complex image editing
  • Image stabilization & motion deblurring
  • Etc.

Sample Images

Example of CRISP’s burst denoising function

  • Before-Fiat
    After-Fiat
    Image capture Fiat With CRISP burst denoising

Example of CRISP’s Virtual Lens deblurring function

  • Before-Test
    After-Test
    Image Capture Test With CRISP Virtual Lens deblurring