My wechat account is MultiAgent1024. I mainly study deep learning, machine game, reinforcement learning and other related topics. Look forward to your attention, welcome to learn and exchange progress together!

AI(artificial intelligence) is divided into three parts: perception, understanding and decision making. This understanding, in image processing and computer vision, is called image analysis (or image understanding). In the world to do this direction is more famous Stanford University artificial intelligence Laboratory (SAIL) director Professor Li Feifei. And the so-called understanding is to understand the deep meaning behind the image. The ultimate goal is to look at an old photo like a human, which may make you cry with deep emotion. (It contains a lot of information. . What Li’s team has done now is to understand the relationships between objects.

Image processing is also divided into three levels: low-level processing, intermediate processing, advanced processing.

Low-level processing: mainly for some simple operations on the image, such as noise reduction, contrast enhancement and image sharpening. Noise can be reduced by filtering. The principle of image enhancement is to process a given image, make the result more source images easier for subsequent operation and application, mainly to solve due to small image grayscale range of low contrast, the purpose is to make the output image grayscale amplified to the specified degree, making increased details in the image looks clear. Sharpen image features, such as edges, boundaries, contrast, etc., to make the image appear better or easier to analyze.

Intermediate processing: involves many tasks, such as dividing an image into different areas or targets, which is what the field of image segmentation does to make it better recognized and classified, which can also be called target detection.

Advanced processing: Why is it so hard to understand images? Because it’s hard for human beings to do this, just as there are 10,000 Hamlets in 10,000 readers’ eyes. Although the field of digital image processing is based on mathematical and probabilistic formula representations, human intuition and analysis play a central role in choosing one technique over another. This is true throughout science.

Computer vision series

The articles in this series are from professor Wei Ying’s lecture notes on image processing at Northeastern University:

  • Digital image processing (1) Introduction
  • Digital image processing (2) Basic digital image processing
  • Digital image processing (3) image transformation
  • Digital image processing (4) digital enhancement
  • Digital image processing (5) image restoration
  • Digital image processing (6) image compression
  • Digital image processing (7) image segmentation
  • Northeastern University graduate course – Digital image processing – final materials summary
  • Introduction to R-CNN and Faster R-CNN for target detection
  • Image Semantic Segmentation Based on FNC and Pascal-VOC Data Set
  • Image segmentation of coins, pins, rice code details
  • Principle of maximum entropy image restoration Method
  • Principle and Code Practice of bilateral Filtering Method
  • Traffic light identification code analysis based on OpenCV

Computer vision Practice

All the articles below are from the study notes of Netease Cloud class “Opencv Computer Vision Actual Combat”.

  • Computer vision combat (a) open a visual combat column
  • Computer vision actual combat (2) image basic operation
  • Computer vision combat (3) threshold and smoothing processing
  • Computer vision actual combat (4) image morphology operation
  • Computer vision combat (5) image gradient calculation
  • Computer vision combat (6) edge detection
  • Computer vision practice (7) image pyramid and contour detection
  • Histogram and Fourier transform
  • Credit card digital recognition project (complete code attached)
  • Harris Corner Detection of image features (complete code attached)
  • Scale Invariant Feature Transform (SIFT)
  • Computer Vision practice (12) Panoramic image Mosaic (complete code attached)
  • Computer vision combat (13) Parking lot identification (with complete code)
  • Answer Sheet Identification (With complete code)
  • Computer vision Practice (15) Background modeling (with complete code)
  • Computer Vision Field (16) Optical flow estimation
  • OpenCV DNN model (with complete code)

The purpose of this column is to convey and share more information, does not mean that the platform agrees with its views and is responsible for its authenticity, if quoted, copyright belongs to the original author, if there is infringement, please contact us to delete.