ISP is at the center of the imaging system.

The article directories

  • 1 the ISP function

  • 1.1 Device Control

  • 1.2 Format Conversion

  • 1.3 Picture quality optimization

  • 2 ISP algorithm flow

The ISP function

Device control

  • Control the Shutter and Gain of the Sensor

  • Control lens zoom and focus

  • Controls the lens aperture

  • Controls the switching of filters

  • Supplementary light control

Format conversion

  • RAG to RGB

  • RGB to YUV

  • YUV turns HSI

  • YUV444 YUV420, etc

Image quality optimization

  • Original image correction (uneven luminous flux and distortion)

  • Color management

  • The noise reduction

  • Dynamic range control

  • Clarity and sharpness are improved

  • post-processing

  • Digital to shake

The picture on the left below is not optimized, while the picture on the right is optimized by ISP.

ISP Algorithm Flow

  • Horizontal Flip: Horizontal Flip

  • Test Pattern: the Test Pattern is generated within ISP and used for debugging within ISP

  • Black Level Correction: Black Level Correction, Black Level refers to the corresponding signal Level when the image data is 0, the purpose of Black Level Correction; First, there is dark current in the sensor itself, which leads to the pixel’s voltage output even when there is no light coming in. However, this part is usually processed at the sensor end. Another reason is that the accuracy of the sensor’s ANALOG-to-digital conversion is not enough. Sensor may not be able to convert information close to 0. Due to the characteristics of human eyes (sensitive to details in dark places), sensor manufacturers generally add a fixed offset during conversion to make pixel output between 5 (non-fixed value) and 255, and then make a subtraction at ISP end to change 5 (non-fixed value) to 0

  • Bad Pixel Correction: Bad Pixel Correction

  • Lens Shadow Correction: Lens Shadow Correction (Lens defect, uneven light flux)

  • AF Statistics in Raw: indicates THE AF Statistics

  • CFA Interpolation: Go to Mosaic and convert RAW to RGB

  • D65 White Balance: make calibration so that under D65, White objects appear White

  • AWB Statistics: You can obtain AWB Statistics after D65 White Balance is completed

  • AWB: The AWB algorithm performs white balance correction based on the color temperature

  • CCM: 3×3 color matrix, the purpose of the imaging system to match the color of the human eye. AWB has calibrated white. CCM is used to calibrate white for all colors except white, using a 3X3 CCM matrix where each column of coefficients r1+ G1 + B1 equals a constant value 1. The final results of Ccm correction can be analyzed by taking 24 color card images and using IMATEST analysis

  • Gamma: Gamma correction, the reason is the nonlinear characteristics of human eyes, when the brightness changes, the human eyes do not feel the relationship of multiplication

  • CSC: RGB to YUV color space conversion

  • AE Statistics: Indicates AE Statistics

  • 2DNR: noise reduction in space domain

  • 3DNR: Noise reduction in the time domain

  • AF Statistics in YUV: indicates the AF Statistics

  • Local Tone Mapping: Each Local area of the image is clear enough

  • Sharpness: Sharpening makes images sharper

  • WDR: Wide dynamic algorithm makes the dynamic range of the image larger

  • Post-process: post-processing

  • LDC: Lens distortion correction

  • De – shaking: image stabilization

Transfer:Deepinout.com/isp/image-p…

* I have been engaged in Android Camera related development for 5 years.

* Currently working in Shenzhen,

* Welcome to follow my wechat public number: Chi Notes, and hope to communicate with more partners ~