Filtering data involves restricting the analysis to a particular sub-group of the sample. For example, if you view the data of people in your sample that are male and aged under 50 you have filtered the data.
Filters are created using rules regarding which respondents should be included and which should be excluded from the analysis. While there are some nuances, generally filters are created by various AND and OR rules. For example, your rule may be to include people that are aged under 50 AND are males, or, the rule may be aged under 50 OR are males. There is no consistence between the different data analysis programs in terms of how filters are created. In SPSS, for example, filters have to be created by typing an expression. For example, a filter of males under 50 would be entered as q2 <= 7 & q3 == 1, where q2 and q3 are Variable Names and 7 and 1 are specific values that represent age and gender categories respectively.
By contrast, Q instead uses the same basic logic, but presented in a 'tree' type format (on the left), whereas DataCracker uses a less-flexible but easier-to-use grid of checkboxes.
Almost all programs treat the creation of a filter as being equivalent to creating a new variable, where the variable contains two categories, one representing the people in the filter and one representing the people not in the filter group. Typically, these are added to the data file allowing them to be re-used.
Once a filter has been created it can usually be re-used by selecting it from a list of saved filters. The only prominent exception to this is SPSS, in which you need to create a new filter but can do so by using the older filter (e.g., if the previously-created was called var001 then the expression for the new filter if re-using it would be var001. Another difference between SPSS and most programs is that in SPSS a filter is either on or off, whereas in other programs the filter is specifically applied to separate analyses.