Spatial variation in average lifetime in one or more cells
Here, the most robust thing to do is mono-exp pixel by pixel fitting.
The photon count in any one single exponential must be at least M* (say 1,000). If you waited long enough when acquiring, or have a low resolution image with big pixels then you may have >M* in each pixel. If you do not then you must bin pixels such that the count in each bin is >M*, reducing your resolution. If you can manage the larger file sizes then there is a visual benefit from binning a high resolution image as opposed to acquiring at a lower resolution.
If your data is clearly bi-exp or more, rather than a mono-exp, some people prefer to do a bi-exp and then calculate an average tau. Although the fits are visibly better and have better a chi squared value, I am not convinced that this has any benefit or mono-exp fitting.