ELLFOU and ELLPRO work in very different ways. ELLPRO works by creating trial ellipses which it places on the galaxy. It then carefully adjusts the shape and position of the ellipses until the brightness variation around the ellipse is minimised. ELLFOU identifies the locations of all the pixels in a given isophotal range, takes a subset of those and then trys to fit an ellipse through the pixels that minimises the pixel/ellipse distance sum. These approaches are known as intensity and contour analysis respectively.
You might well ask then why two routines are provided. The reason is simply that galaxies vary enormously. If these routines are used with a normal galaxy of approximately elliptical shape they will normally both behave well. But this is not always the case. Limits to this sensible behaviour are outlined below.
The major difficulty with ELLFOU is its speed. With current machines (circa 1993 - SUN Classic) analysis of a large galaxy can take some time, clearly this will become less important in time as machine speed increases, but at present it is a serious problem. The problem is caused by the fact that many pixels can fall within a given isophotal range. If the image contains too many pixels within the defined isophotal range then the software has to select a subset and try to ensure that they are both representative and evenly distributed. This is time consuming. The minimisation routines then applied have been designed to be not only robust but (most importantly) to avoid false minima as far as possible. An algorithm inclined to allow false minima would have disasterous results, consequently, the software design implemented has verged on the side of caution and unashamedly sacrifices speed for accuracy.
A second problem with ELLFOU is that at the higher isophotes where the radius is small there will be very few pixels within a given isophote, this can make the shape of any resultant isophotal profile uncertain and limits the low radius limit.
ELLPRO is the faster of the two profiling application beings 3–4 times faster than ELLFOU in trials. This is because it does not have to sample all the image pixels and can instead sample the image as required using bi-linear (or bi-cubic spline if parameter FAST=FALSE) interpolation to provide it with the data it needs. As a result it can work well at both small and large radii. To help the application work at lower isophotes, where the signal to noise ratio is lowest, the number of points on the trial ellipses increases as the isophotal radius increases. This is subject to an upper limit defined in the ELLPRO INCLUDE file which may be adjusted by users with faster machines. Comparative trials showed that the results generated using ELLPRO, ELLFOU and RGASP on three different galaxies were consistent. Information on RGASP (VAX and UNIX versions) may be found in SUN/52.
Consequently, for most applications, particularly where the galaxy is reasonably small (in terms of pixels) ELLPRO is the better choice. However, it has been found that some types of barred spiral galaxy, can cause it difficulties. These occur mainly where the brightness of the profile and its orientation change drastically at the same time. The situation is exacerbated if the transition occurs at a low signal to noise ratio. You should be cautious in such cases and set FRZORI to FALSE and experiment with the LIM1, LIM2 and FINE parameter values. Alternatively, they can use ELLFOU. Such objects deviate greatly from elliptical form, particularly in their lower isophotes and SECTOR may also provide useful information.