You can also use artificial intelligence to learn a set of image bases given a dataset of images. This paper (http://proceedings.mlr.press/v28/gupta13b.pdf) uses AI techniques to generate a set of image bases that help them classify MRI images of Alzheimer's disease, and they obtain great results! So if you thought image bases were simple and/or boring, then think again! Generating relevant bases can be a complex challenge, and can have significant results for image classification!
zhouwen
With AI, we can also do image compression using K-Means: https://ijcsmc.com/docs/papers/April2015/V4I4201588.pdf
kalebm
Is there a basis that is generally considered to be the best in practice? If not, what sorts of bases are good for certain tasks and why?
sc2019
In lecture it was mentioned that Walsh-Hadamard would be good for something that had a lot of edges. After doing some further research online it seems like the DCT is the standard for JPEG compression: https://www.ece.ucdavis.edu/cerl/reliablejpeg/compression/
jtburkle
Which of these are best for different applications? (e.g. line drawings, photographs, diagrams)
You can also use artificial intelligence to learn a set of image bases given a dataset of images. This paper (http://proceedings.mlr.press/v28/gupta13b.pdf) uses AI techniques to generate a set of image bases that help them classify MRI images of Alzheimer's disease, and they obtain great results! So if you thought image bases were simple and/or boring, then think again! Generating relevant bases can be a complex challenge, and can have significant results for image classification!
With AI, we can also do image compression using K-Means: https://ijcsmc.com/docs/papers/April2015/V4I4201588.pdf
Is there a basis that is generally considered to be the best in practice? If not, what sorts of bases are good for certain tasks and why?
In lecture it was mentioned that Walsh-Hadamard would be good for something that had a lot of edges. After doing some further research online it seems like the DCT is the standard for JPEG compression: https://www.ece.ucdavis.edu/cerl/reliablejpeg/compression/
Which of these are best for different applications? (e.g. line drawings, photographs, diagrams)