What are some instances of real-world sensors, cameras, filetypes, or datasets whose image compression method is (at least mostly-based on) the truncated-SVD?
A first difficulty with your question is that "usual instances" may not be fully publicized nor documented. A second one is related to the fact that "compression standard" are indeed "decompression procedures". In other words, they offer several options to compress a file, and make sure anybody could decompress, given the knowledge of the options chosen.
Therefore, you might get a compression scheme potentially using SVD, but not using it in practice. One of the reason is that SVD or PCA compression rely on adaptive bases. The latter are often obtained from second-order empirical statistics. Hence, the projection vectors are learned, and ought to be transmitted along with the truncated eigenvalues or singular values. For only one image, the additional cost is often prohibitive, with respect to the thought compression ratio/error balance. Hence, the discovery of the discrete cosine transform (DCT) and its ability to emulate well SVD for correlated data or Markovian models won the deal.
For a set of images, the basis vectors could be shared, to reduce the storage cost. From the papers I have read, SVD/Karhunen-Loève transforms have be used in satellite picture coding (especially in multi-spectral images), and multi-channel physiological signals (EEG)
It has been said that the DCT reasonably closely matches the KLT for a representative set of images. KLT is essentially the same as PCA, I believe and SVD is only a different way to compute the same?
An image coder that computes the DCT, sorts the result, then truncate could perhaps be said to be an approximate «truncated SVD»? Usually, you dont sort adaptively but zigzag scan assuming that low frequencies are the most energetic. And you quantize terms, but this effectively leads to a «truncation» of (usually) low energy high frequency terms.