Did you compare it also with JPEG? JPEG XL? Your method is also lossy so I think it would be a fair comparison. And JPEG XL is also new like yours. Good work
A possible correction on your word choice: from my understanding of it, 0.05 to 0.1 for accuracy is a *de*crease and not an increase. You said that 0.05 was an accuracy of 5cm while 0.1 was an accuracy of 10cm, which sounds like less accuracy rather than more.
The unit of 0.05 is in meter, this is why it become 5cm, basically smaller accuracy are more accurate and they will take more data! I am not sure what I said in the video, if I said something in contrary sorry about that!
@@mohsenzare2511 Rereading it, I suppose it actually wasn't clear exactly what I meant. You said in the video that 0.05 to 0.1 was an increase in accuracy, which definitely isn't the case. 0.1 to 0.05 would be though.
Nice results, good work! What I would have liked to see are load times, as compression times rarely make a difference, decompression / load times do always matter. Some compression methods even prioritise decompression performance over minimal file size. I guess you want to achieve the smallest file size, but it would still be very informative to see your method compared to some others including raw data at different resolutions. Just a minor word issue: You are not increasing the accuracy as you say. You are in fact increasing the maximum allowed margin of error, thus decreasing the accuracy. In short, you are increasing the inaccuracy. Something like "max. error" or "max. deviation" might be a better name for the parameter. I hope this makes more sense to you, as others have already pointed it out, but it looked like there might have been some confusion.
I did not test its load time yet! But I will do that later! compressing is more process costly anyway, but that does not matter as it happen only during development! Yeah I might be confuse about the accuracy stuff, And thanks for telling me that!
This is equivalent to lowering the png resolution, using mipmaps or subsamples. It might be reasonable to consider using many other formats designed for effective storage while sampling dimensionally. Consider the Carmack's megatexture format for example.
No, this is not like lowering the PNG resolution, this same as letting each pixel have a little error, and spending more data where height-map change more, for compression! we are not changing dimension here!
