Computing, IT & AI
Edge-preserving diffusion and normalized cuts in computer vision
1990-2000 CEJitendra Malik, with Pietro Perona and Jianbo ShiPerona & Malik (IEEE TPAMI, 1990); Shi & Malik, Normalized cuts (IEEE TPAMI, 2000)Diaspora (work abroad)
1990
Methods for smoothing images while keeping edges, and for dividing an image into meaningful regions.
- Dated
- 1990–2000
- As scholars put it
- 1990-2000 CE
- Originator
- Jitendra Malik, with Pietro Perona and Jianbo Shi
- Source text
- Perona & Malik (IEEE TPAMI, 1990); Shi & Malik, Normalized cuts (IEEE TPAMI, 2000)
- Region
- Diaspora (work abroad)
- Discipline
- Computing, IT & AI
- Evidence
- Documented
- Confidence
- High
- Verification
- Verified-web
Background
Jitendra Malik, an IIT Kanpur graduate at UC Berkeley, introduced the Perona-Malik equation, a non-linear version of the heat equation that smooths an image without blurring its edges. With Jianbo Shi he developed normalized cuts, which segment an image using the eigenvectors of a graph of pixel similarities.
Significance
Both are classics of computer vision. Normalized cuts became a general clustering method, and Malik's group also set up standard benchmarks and later the R-CNN object detector. He received the IEEE Computer Pioneer Award.
Path to modern innovation
Image denoising, segmentation and object detection.
Global impact
- Spectral clustering in machine learning.
- Medical and satellite image processing.
References
- 2Scale-space and edge detection using anisotropic diffusion
- 3Normalized cuts and image segmentation
Further reading
Corrections and additional sources: edit IND-396 on GitHub.