Visibility is an Image Quality Index maintained for compatibility with scoring performed by human beings. However, there cannot be a perfect overlap between the "visibility" of humans and the "visibility" of a program. There are many reasons for this, for example:
- Humans apply cognitive processes that are very difficult to apply by computer when analysing images: for example automatic shape completion (when half a circle is visible, the completion is applied by human brain)
- Humans search details in the right place because they expect it: this helps the brain to "form" see the expected shape even if it is not complete
- Humans hardly apply geometric consistency check: a "spot" which is near the right place and with almost the right dimension is classified as visible detail even if it could not be
- Computer is much more sensitive than human beings: a gray level difference of 1 in a 65536 (16-bit) range is easily recognised by the PC but not by humans
- Humans work through monitors: non-specialized monitors can show 255 levels of gray, on 256th of the range available on a 16-bit image
Therefore, visibility as reported by AutoPIA shall be interpreted as "the algorithm recognises the detail as an area with right dimensions and in the right place having a certain signal".
The signal is