4 ms·
Ah, this is a great idea, and I can see how this could be extended beyond simple deblocking. In essence, it computes the uncertainty from the JPEG quantization
by nneonneo 9y ago
Ah, this is a great idea, and I can see how this could be extended beyond simple deblocking.
In essence, it computes the uncertainty from the JPEG quantization step (the source of most of JPEG’s compression) and uses that information to select a decoding that minimizes a particular error metric (here, discontinuities at block boundaries).
But it’s not limited to just that, in principle. It could optimize towards any decoded block within the quantization range given a suitable prior - and I imagine that this could be paired with e.g. a DNN to faithfully reconstruct an image from a JPEG with much better accuracy.
- lokopodium 9y agoThe same should be applicable to MP3s. Also, the continuity measure could include derivatives as well.
- blt 9y agoyeah, the deblocked image looks a little softer. Makes sense, the optimization objective encourages it. Sometimes the block boundary aligns with an image boundary. Agree that incorporating a DNN could be a cool solution.
- sitkack 9y agoCould a DNN be thought of a continuous decompression dictionary?
- psandersen 9y agoAdditionally, I'd love to see a decoder add noise proportional to the uncertainty and surrounding structure as over smoothed images look unnatural as well. See activity masking demo https://people.xiph.org/~jm/daala/pvq_demo/ https://people.xiph.org/~jm/daala/pvq_demo/
- Ruud-v-A 9y agoPik can synthesize noise. https://github.com/google/pik https://github.com/google/pik
- disposablename 9y agoDNNs are not magic, and not everything needs them, especially a fucking JPEG decoder, come on.