A Raman optical fiber distributed temperature sensor, using a wavelet transform based signal processing technique for backscattered anti-Stokes and Stokes signals, is presented. This technique mainly performs two functions. First, it equalizes the wavelength dependent optical fiber attenuation of two Raman backscattered signals and second, denoises these Raman signals without generating any appreciable spatial inaccuracy in locating the hot zones. The proposed technique enables automatic measurement of distributed temperature profile that has better temperature accuracy and very small spatial error in detecting the location of hot zones. The accuracy achieved in temperature measurement with processed Raman signals is much better than the accuracy obtained with unprocessed signals. Results show a maximum temperature error of ±3.5°C in a temperature range of 25–295°C and a maximum spatial error of ±3 cm in locating the hot zones over a sensing length of 205 m with spatial resolution of 1 m. The proposed technique has been used for the development of a prototype Raman optical fiber distributed temperature sensor (ROFDTS) system which employs 200/220μm Polyimide coated Multimode optical fiber. The technique is much simpler compared to other complex techniques described in the literature and is suitable for general temperature sensing applications. The technique supports automatic and dynamic self-calibration of ROFDTS to take care of slow variations/drifts in the observed Raman signals which are due to fluctuations in laser power and laser –fiber coupling. A self-calibration setup has also been developed to track the changes in the temperature of calibration zone.