r.neighbours should offer to ignore centre value

[marisn]

Currently there is not possible to ignore the centre cell when performing calculation in r.neighbours. There should be a flag to ignore centre cell value in calculation. It should be implemented as a flag as it should be available to all calculation methods.

[wenzeslaus]

For now, in case you don't know, you can use r.mapcalc with [m, n] operator to achieve the same results in less convenient way.

[glynn]

Replying to marisn:

Currently there is not possible to ignore the centre cell when performing calculation in r.neighbours.

You can use the weight= option to supply a kernel in which the centre cell has zero weight.

A flag may still be worth having if this is a sufficiently common case.

[marisn]

Replying to glynn:

Replying to marisn:

Currently there is not possible to ignore the centre cell when performing calculation in r.neighbours.

You can use the weight= option to supply a kernel in which the centre cell has zero weight.

A flag may still be worth having if this is a sufficiently common case.

First - it is a documentation problem as module description is "Makes each cell category value a function of the category values assigned to the cells around it" - the problem with word "around" - I read it as "nearby cells excluding the centre cell". I took liberty to add a notice in r64860 that "around" means - "all surrounding cells + centre cell".

Second - it is not possible to use weights and circular neighbourhood at the same time (documentation states: "The -c flag and the weights parameter are mutually exclusive.")

wenzeslaus - thanks for hint. As I needed a count of neighbour cells, I ended with mapcalc expression "if(isnull(original), count, count-1)"

[glynn]

Replying to marisn:

Second - it is not possible to use weights and circular neighbourhood at the same time (documentation states: "The -c flag and the weights parameter are mutually exclusive.")

More accurately, it is not possible to use weight= and an automatically-generated mask (either the circular mask generated by -c or the Gaussian mask generated by gauss= or the square mask generated in the absence of any other applicable option). Nothing prevents weight= from being used to supply a circular mask.

However, any automatically-generated mask (whether square, circular or Gaussian) is simply a short-cut whose behaviour could also be obtained by passing the appropriate mask via weight=. So the question isn't about what's possible, but what's possible without providing an explicit weights file.

There's no end to the set of masks which could be generated internally; the only question is which ones are common enough to warrant a short-cut.

[marisn]

Replying to glynn:

Replying to marisn:

Second - it is not possible to use weights and circular neighbourhood at the same time (documentation states: "The -c flag and the weights parameter are mutually exclusive.")

There's no end to the set of masks which could be generated internally; the only question is which ones are common enough to warrant a short-cut.

Should there be a vote for and against providing a short-cut for "ignore centre cell"?

As r64860 now has a documentation enhancement to clarify current behaviour, this bug also can stay open for longer time and anyone can just then add "me too" comment.

[pvanbosgeo]

Me too (would like such a short-cut)

[neteler]

Ticket retargeted after 7.0.1 milestone closed

[neteler]

Ticket retargeted after milestone closed

[neteler]

Ticket retargeted after milestone closed

[neteler]

Ticket retargeted after 7.0.3 milestone closed

[martinl]

Milestone renamed

[neteler]

Ticket retargeted after milestone closed

[martinl]

All enhancement tickets should be assigned to 7.6 milestone.

[martinl]

Ticket retargeted after milestone closed

[martinl]

Ticket retargeted after milestone closed