CFU/mL of sample can easily be determined using solid medium. However, difficulty is always encountered using broth culture in spectrophotometer which may be considered more appropriate by some researchers.
If you consider the horizontal axis to be a colony count and the vertical axis is an optical measurement, then after drawing the standard chart, you can reach the result by attaching a point on the graph on both axes.
The above answers all assume that the fraction of reproductively viable cells in the culture is constant. The OD vs. number of cells/ml does not distinguish between viable and dead cells. So one needs to get your curve for OD vs. cells/ml and then measure the PFU/ml at each cell concentration to get a corrected curve for CFU/ml vs. OD.
Alireza Mordadi, Sorry, you did mention them, but the original questioner may not have gotten from that the necessary details.
The cell concentration vs. OD curve is linear (or log-linear) for a restricted range of concentrations, and similarly for the PFU/ml vs. cells/ml curve, so one really needs to track the OD vs. cells/ml vs. PFU/ml to ensure that you are working in the "linear" parts of the curves. Each of these curves will have precision estimates, which must be combined to get PFU/ml from OD. Equipartition of variance works best for such problems.
As I noted above, a calibration curve is necessary to ensure that the formula you wish to use applies to all values on the curve. The problem is that when the data fall outside the linear or loglinear range of the calibration curve, the formula often no longer applies. Hence you need calibration curves for CFU/ml at each cells/ml value and cells/ml at each OD value. As long as your data points fall within the linear (or loglinear) parts of the calibration curves, the calculations are straight forward. When they fall outside of the, things become very complicated and precision estimates become very large (imprecise). The precision estimates for your measured and calculated values are just as (or perhaps even more important) than those for your mean values of interest. Regarding the latter point, I once had the uncomfortable situation where a colleague at another institution calculated the extrapolated intercept of a cell survival curve indirectly and obtained a value of 40, which was much higher than that for cells of other tissues (ranging from 1 to 10), indicating high resistance. I warned her several times that if she continued to cite the value she obtained, I would be forced to publish how she was in error. When I did the error analysis on her data, the 95% confidence limits were between 10-5 and 106, so any conclusion about this value being evidence of a high resistance was unsubstantiated by her data. My paper was titled: On the use and abuse of statistics in radiobiology: with examples from the data of your friends and colleagues. We had an overflow audience.
Alloysius Chibuike Ogodo when you talk about standard curve it means any tests according to your curve must be same so you can not use dilution specially in case of bacterial growth curve because that curve is non liner and you can not use a formula of liner curve for non liner curve.
As dear Christopher said it is complicated because we are talking about organisms.
I do not know that I can clearly explain or not, about this I am sorry.