An improved local distribution fitting model

This paper presents an improved local distribution fitting model. Based on the recently proposed local binary fitting (LBF) model and local distribution fitting (LDF) model, this paper modifies the local background and object distribution fitting functions. Thus our model overcomes the limitations of the LDF model with original distribution fitting functions. In addition, this new model incorporates particle swarm optimization (PSO) into Gaussian mixture model (GMM) based intensity distribution estimation to optimize the parameter estimation. We apply this improved LDF model to image segmentation on both synthetic and real images. Experimental results show desirable performances of our method and present the improved performance over the original model.