# public documents.sextractor_doc

## [/] [detect.tex] - Blame information for rev 25

Line No. Rev Author Line
1 25 gam
\chapter{Detection and \index{segmentation} segmentation}
2 19 gam
In {\sc SExtractor}, the detection of sources is part of a process called
3 25 gam
{\em \index{segmentation} segmentation} in the \index{image} image-processing vocabulary.  Segmentation normally
4
consists of identifying and separating \index{image} image regions that have different
5
properties (brightness, \index{colour} colour, texture...) or are delineated by edges. In
6
the astronomical context, the \index{segmentation} segmentation process consists of separating
7 19 gam
objects from the sky background. This is however a somewhat imprecise
8 25 gam
definition, as astronomical sources have, on the \index{image} images --- and even often
9
physically ---, no clear \index{boundaries} boundaries, and may overlap. We shall therefore use
10 19 gam
the following working definition of an object in {\sc SExtractor}: a group of
11
pixels selected through some detection process and for which the flux
12
contribution of an astronomical source is believed to be dominant over that
13 25 gam
of other objects. Note that this \index{mean} means that a simple $x,y$ position vector
14 19 gam
alone cannot be handled by {\sc SExtractor} as a detection: most measurement
15
routines require some rough shape information about the objects.
16
 
17
Segmentation in {\sc SExtractor} is achieved through a very simple
18 25 gam
\index{threshold} thresholding process: a group of connected pixels that exceed some \index{threshold} threshold
19 19 gam
above the background is identified as a detection.  But things are a little
20 25 gam
bit more complicated in practice. First, on most astronomical \index{image} images, the
21 19 gam
background is not constant over the frame, and its determination can be
22
ambiguous in crowded regions. Second, the software has to operate on noisy
23 25 gam
data, and some filtering adapted to the characteristics of the \index{image} image has to
24 19 gam
be applied prior to detection, to reduce the contamination by noise
25 25 gam
peaks. Third, many sources that overlap on the \index{image} image are unlikely to be
26
detected separately with a single detection \index{threshold} threshold, and require a
27
de-blending procedure, which is actually \index{multi-thresholding} multi-thresholding in {\sc
28 19 gam
  SExtractor}. Each of these points will now be described in greater detail
29
below. It is worth mentioning here that these 3 difficulties could, to a
30 25 gam
large extent, be bypassed using a \index{wavelet} wavelet decomposition (e.g. Bijaoui \etal
31 19 gam
1998). Although such an algorithm might be implemented in a future version of
32 25 gam
{\sc SExtractor}, current constraints in processing speed, available \index{memory} memory
33
(processing of gigantic \index{image} images) often make the pedestrian approach'' still
34 19 gam
more interesting in the case of large scale surveys. \gam{Remove or update
35 25 gam
  last two sentences on \index{wavelet} \index{wavelets} wavelets?}