云昴(Mao Yun)

ML for imaging (2)

| 机器学习图像

摘要

Feature extraction

Intensity

  1. Pixel-level
  2. Patch-level
  3. Image-level

SIFT

尺度不变特征转换 (Scale-invariant feature transform), 它是一种检测局部特征的算法。

Scale-space Extrema Detection

构建尺度空间。

  • for all scales, compute convolution with Gaussian
  • Compute difference of Gaussians
  • Detect local maxima and minima of D

Keypoint localisation

云昴(Mao Yun)