An improved homomorphic spatial filtering algorithm is presented. With the reversed low frequency image this algorithm compensates imbalanced illuminating in the uneven brightness image in order to reduce the loss of low frequency components. Then two optimum schemes about spatial filtering are proposed for efficient processing: a method of 2D window template decomposition

which is constructed by tensor production and an approximate spatial filtering based on sampling and interpolation. Simulation indicates that the improved algorithm with the two optimum methods enhances illuminating uniformity

reduces the computational load and keeps the low frequency components effectively. Besides

it can be performed by more 2D windows.