Make threshold dependent on local noise

[MatthewWhiting]

It would be good to have the threshold depend on what the local noise is. For instance, with 2D images, there might be regions of lower noise, where you want to search deeper, compared to other regions that might be affected by artifacts where the threshold will be higher.

There might be a need to output an image showing the detection threshold as a function of position.

[MatthewWhiting]

The first part of this was done in [582]. The searching functions now are independent of the image classes, and are simply given a vector of bools, indicating whether or not each pixel is detected.

The next step will be to have different ways of calculating whether each pixel is a detection or not, for instance calculating a local threshold for each pixel.

[MatthewWhiting]

There appears to be something wrong with the findSources2D function, that does not preserve the information about subsections (ie. the offsets are not added to the objects afterwards).

This needs to be fixed!

[MatthewWhiting]

Problem actually lay with the copying of Detections and the offsets not being preserved. Have added a + operator for Detection (it was previously just using the Object3D one), and this seems to have fixed it.

[MatthewWhiting]

[582] had introduced a bug that I've only just tracked down -- lutz_detect had a rogue ydim in place of xdim.

This is fixed and the changes from [582] now actually work in all cases tested (non-square subsections were crashing).

[MatthewWhiting]

Put this in the 1.2 release. This could be useful - eg. to scan through a cube an image at a time, and get a different threshold for each image. This would be a step towards what we do with Selavy, but easily doable in the context of the Duchamp code (ie. without involving casacore).