Compression technique
category: code [glöplog]
I was wondering if any coders have tried this compression idea before, and know if there's a name for it. Its a lossless technique, and im wondering what the compression ratio could be for something like this.
Let's say one has the compressed data. The compressed data has hamming weight number = 5, so (it's just an RLE-stream of 1's and) the data looks like this:
11111
To decompress we have an index to where to split the data and move the stream to the right, where the data is splitted and no ones, there is zero. Since during the iteration we will always have to move one step, the step number for one-step will be a zero. We choose that the step number is 1 next:
110111
The new start number starts where the bitstream was divided. Next step-number we choose are 0:
1100111.
(we could do this until we got the original, but lets just stop here).
This is our decompressed data: 1100111 = 103.
From the above, it looks like some kind of delta's. If anyone can confirm if this has anything to do with delta-coding please elaborate. Ive made delta encoder/decoder before, but can't tell if it is yet. In the case above 103 was compressed to 5. But we need to check wether or not the compression can be done backwards. And we can see that 5 are not enough, because we need two-step sequences. In this case we could use two two-bit numbers. So, we have the bitstream: 1100111. We use a technique to read bitpairs. first is 11 => do nothing. second is 10 => ... the bits are stored in a way that we can decode. The three 1's are moved from a position to the next posision -1. same after the next iteration. and the result is 11111 = 5. meanwhile we stored the position of where the splitting occurs.
Let's say one has the compressed data. The compressed data has hamming weight number = 5, so (it's just an RLE-stream of 1's and) the data looks like this:
11111
To decompress we have an index to where to split the data and move the stream to the right, where the data is splitted and no ones, there is zero. Since during the iteration we will always have to move one step, the step number for one-step will be a zero. We choose that the step number is 1 next:
110111
The new start number starts where the bitstream was divided. Next step-number we choose are 0:
1100111.
(we could do this until we got the original, but lets just stop here).
This is our decompressed data: 1100111 = 103.
From the above, it looks like some kind of delta's. If anyone can confirm if this has anything to do with delta-coding please elaborate. Ive made delta encoder/decoder before, but can't tell if it is yet. In the case above 103 was compressed to 5. But we need to check wether or not the compression can be done backwards. And we can see that 5 are not enough, because we need two-step sequences. In this case we could use two two-bit numbers. So, we have the bitstream: 1100111. We use a technique to read bitpairs. first is 11 => do nothing. second is 10 => ... the bits are stored in a way that we can decode. The three 1's are moved from a position to the next posision -1. same after the next iteration. and the result is 11111 = 5. meanwhile we stored the position of where the splitting occurs.
http://encode.ru/
Would be a great place to ask if no one knows here. I've seen a few scener's pop their heads in from time to time like one of the FRequency guys.
Would be a great place to ask if no one knows here. I've seen a few scener's pop their heads in from time to time like one of the FRequency guys.
Thanks Intrinsic, ill do that
You are forgetting that your index on where to split isn't free. This doesn't work.
If you want to watch Silicon Valley S1E8 this is a spoiler. But it's definitely related.
Ah fuck now I definitely have to watch that show.
why to invent what invented already?
