Abstract :
As the existent filters haven´t good effect in denoising and preserving edge for intensity image, a wavelet threshold denoising algorithm based on homomorphic transform has been put forward. The wavelet threshold and the selection of threshold function are focused on, then the former is determined through the use of particle swarm optimization algorithm and the latter is obtained by the creation of a new threshold function having advantages of both hard threshold and soft threshold. In noise suppression, the maximum entropy of one-dimensional histogram of the indentation image is proposed to segment range image. Experimental results demonstrate that this method is not only applicable to the segmentation of images whose one-dimensional histogram is not the ideal bimodal shape, but also has strong anti-interference ability and small amount of computation. For outliers, there are two ways for the suppression of outliers. One is the suppression of outliers based on the intercepted morphological filter, and the other is the adaptive median filter algorithm based on pixel difference. The former reduces the error introduced by anomalous pixels through effectively removing abnormal pixels, making the smoothing effect ideal, and the latter can´t only suppress the outliers, but also meet the requirements of real-time.
