• 简体中文

This is a easy ISP (aka:ez_ISP) for RAW to RGB conversion. It is based on the package of numpy, and it is easy to use and understand. The ez_ISP project is implemented by python, and it is easy to transplant to other platforms such as C/C++ for speed up.

• Bad Pixel Correction， 坏点校正
• Black Level Correction， 黑电平校正
• Anti Aliasing Filter， 抗混叠
• Bayer Noise Reduction， RAW域去噪
• Auto White Balance， 自动白平衡
• Color Filter Array Interpolation， 去马赛克
• Color Correction Matrix， 颜色矫正
• Global Tone Mapping， 全局色调映射
• Gamma Correction， Gamma映射
• Edge Enhancement， 边缘增强
• Brightness Contrast Control，亮度控制
• Chorma Noise Reduction，Chorma域去噪
• Lens Shading Correction， 阴影矫正
• Luma Noise Reduction， Luma域去噪
• Local Tone Mapping， 局部色调映射

The ez_ISP project tree structure is listed as follows.

ez_ISP
│  .gitignore
│  run.py
|  isp_pipeline.py
│  LICENSE
│  README.md
│
├─config
│      isp_config.yaml
│
├─assets
│      raw.png
│
├─algorithm
|     __init__.py
|     aaf.py
|     awb.py
│     bcc.py
│     blc.py
│     bnr.py
│     bpc.py
│     ccm.py
│     cfa.py
│     cnr.py
│     ee.py
|     fir.py
|     gmc.py
|     gtm.py
|     ltm.py
|     r2y.py
│     utils.py
│     y2r.py
│
├─test_images
│      test.RAW
│

Device: AMD Ryzen 5 5600 6-Core [email protected] GHz, Image Resolution: 1920x1080, Running time cost here:

Time cost: 27.12 s for a 1920x1080 image, though it is not fast enough, it is easy to use and easy to understand.

You can install ez_ISP by pip install the packages below.

• The main package is numpy, and opencv-python is used for image I/O.
• Other packages are used for the demo such as rawpy and yaml, time, os.

pip install yaml, numpy, opencv-python, time, rawpy, os

Clone the ez_ISP project from github, and you can run the project.

git clone https://github.com/HuiiJi/ez_ISP.git
cd ez_ISP

Make sure that you have installed the packages above, or you will get an error when you run the project.

The ez_ISP project is run by the run.py file.

python run.py

But before you run the py, please config the config/isp_config.yaml file, The config file is listed as follows.

# -------------------- ISP Module Enable/Disable --------------------
enable:                
  BPC: True
  LCS: False                 # not implemented yet
  BLC: True
  AAF: True
  AWB: True
  BNR: False                 # not implemented yet
  CFA: True
  CCM: True
  GTM: True
  GMC: True
  R2Y: True
  CNR: True
  EE:  True
  BCC: True
  HSC: False                 # not implemented yet
  Y2R: True

# -------------------- Algorithm Params --------------------
RAW_img_path: '/mnt/cvisp/isp/ez_ISP/test_images/2DNR_Case_1_1.raw'
RAW_Height: 1080
RAW_Width: 1920
white_level: 1023
bayer_pattern: RGGB

BPC:
  bad_pixel_threshold: 30

LCS: ~

BLC:
  black_level_r: 256.0
  black_level_gr: 256.0
  black_level_gb: 256.0
  black_level_b: 256.0
  alpha: 1.        
  beta: 1.                  

AAF: ~

AWB:
  r_gain:  1.6             
  b_gain: 2.0            

BNR:
  BNR_method: 'bilateral'

CFA:
  CFA_method: 'bilinear'

CCM:
  ccm_matrix:
    - [1.631906, -0.381807, -0.250099]
    - [-0.298296, 1.614734, -0.316438]
    - [0.023770, -0.538501, 1.514732 ]

GTM:
  GTM_method: 'smoothstep'

GMC:
  gamma: 2.0

R2T: ~

CNR:
  CNR_method: 'gaussian'
  CNR_threshold: 0.3

EE:
  edge_enhancement_strength: 0.3

BCC:
  BCC_contrast: 0.1
  BCC_brightness: 10

HSC: ~

Y2R: ~

The params are listed as follows.

• enable: enable or disable the ISP module.
• RAW_img_path: the path of the RAW image.
• RAW_Height: the height of the RAW image.
• RAW_Width: the width of the RAW image.
• white_level: the white level of the RAW image.
• bayer_pattern: the bayer pattern of the RAW image.

If you don't want to use the ISP module, just set the enable to False. What you must to config is the RAW_img_path, RAW_Height, RAW_Width. The other params are the params of the ISP module, you can set them according to your needs.The result will be saved in demo_outputs folder.

Here are some courses about ISP, you can learn more about ISP from these courses.

• ISP Pipe1（了解）
• ISP Pipe2（了解）
• ISP Pipe3（了解）

You can learn more about ISP Pipe from these courses., but i think these courses are not useful for me, because i am not a camera engineer, i just want to learn the ISP algorithm, so i think the courses below are more useful for me.

• Camera History（了解）
• Recommend for ISP（推荐）
• Digital Photomontage（了解）
• Marc大佬开课（强烈推荐）

Here are some open source ISP projects, which are very helpful for me to complete this project.

• openISP
• fast-openISP

• HDR-ISP
• ISPLab

MIT Thanks for your attention! If you have any questions, please contact me @HuiiJi.