Marine heat flow surveys typically utilize Lister-type heat probes, which allow multi-penetrations to measure both temperature and thermal conductivity in situ. The temperature rise resulting from penetration friction must first be removed in order to obtain the equilibrium temperature and the thermal conductivity of the sediments. However, the time needed to return to the original sediment temperature may be very long. Using highresolution heat probes, temperatures with resolution of 0.1 mK were obtained within the first few meters below the seafloor off southwestern Taiwan. This research introduces an improved data processing technique that can extrapolate friction-raised temperature to infinite time and obtain equilibrium ambient temperature accurately from a limited amount of highresolution temperature recordings. The computing technique is to regress the temperature decay function numerically on parameters of ambient temperature, initial temperature rise, thermal conductivity, and heat capacity of the sediment. In order to obtain a better fit between the data and the model as well as a better prediction of the friction raised temperature, we allowed erroneous solutions to parameters for regression. This technique was tested on both synthetic and high-resolution field data to demonstrate accuracy and efficiency. From the field experiment, we estimated that the error between the prediction and the true equilibrium ambient temperatures to be within a few 0.1 mK which is much smaller than a typical resolution, of a few mK in marine heat flow measurement.