I have an image with its type is 16 bit signed. I see some paper, the author often convert it to unsigned 8 bit. But they did not explain why they do it. Could you explain what is benefit of this work? And How to implement it by matlab code?
Many image display devices only have an 8-bit DAC. So if one wants to analyze only what will be displayed on such a device (without added dithering or noise filtering, etc.), one can convert to that 8-bit format and potentially save storage memory. Many standard image file formats also only support 8 bits of luminance, so the same is true for data planned to be stored or transmitted in such a format.
However, this process is lossy, and will remove information that a display supporting greater luminance resolution (10-bit DAC, etc.) can display.
Other than the data being smaller and thus requiring less memory there is usually no benefit to converting to 8 bit unsigned.
There is never a benefit in terms of image quality, and there is usually not a benefit in terms of processing speed. Most processors handle 16 bit values as fast as they do 8 bit values (though there are exceptions). The one possible speed benefit is that more data will fit in the processor cache, which can make the processing faster. At that level of detail you really just have to run it and see which way is faster.