Decreasing precision while saving data

6 次查看（过去 30 天）
Brandon
Brandon 2012-5-15
Greetings -
I am working on a project where I need high precision for the calculations but I do not need the high precision while saving the data for plotting. The data files are currently about 8mbs a piece and I need to save off several hundred of them. I would like to get them smaller. I tried to reduce it to single precision:
data = single(data)
and got the file down to about 2mb each. Is there any way I can further reduce the precision. I only need maybe 4 significant figures.
I am saving the files in .mat format if that is helpful.
Best Regards,

请先登录，再回答此问题。

回答（1 个）

Walter Roberson
Walter Roberson 2012-5-15
Do you have an estimate for min() of your data, and max() of your data?
4 significant figures after the decimal point requires 14 bits. Signed would require 1 more. If your value before the decimal point can exceed 1, you would need at least 2 bits for it. This plausibly gets you beyond 16 bits per value. 16 bits per value would allow you to pack 2 values into 32 bits (the space occupied by 1 float), but we could expect that the compressibility of the data would go down (.mat files are usually compressed.) It is not certain that the space would be reduced.

类别

在 Help Center 和 File Exchange 中查找有关 Image Arithmetic 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by