Estimation of digital image noise based on multiscale segmentation

Noise is an important factor that affects the quality of digital image, which can inhibit the useful information thus affecting the accuracy of information extraction. And signal to noise ratio is an important basis for the estimation of image information amount, so how to have a reasonable estimate of image noise and determine the signal to noise ratio with it is an important part of research into information measurement of digital images. According to the need of information measurement of digital images, a method for noise estimation based on multiscale segmentation has been put forward in this paper, which is also an improved local mean and local standard deviation noise estimation method. Both theories and experiments show that this method is more reasonable and has higher accuracy than other existing methods for image noise estimation.