Insulator Infrared Image Denoising Method Based on Wavelet Generic Gaussian Distribution and Map Estimation

The infrared techniques on failure detection in power grid have attracted widely attention in recent years. Since the infrared image of the insulator string has high noise and low-contrast, it will affect the judgment accuracy of the zero value insulators. This paper proposes a method based on wavelet generic Gaussian and maximum posterior probability estimation for the noise removing of insulator infrared images. Due to the sharp peak and long tails features of the wavelet coefficients of the infrared images, generalized Gaussian distribution (GGD) is used as the probability distribution function. Maximum posterior probability (MAP) estimation is used to obtain denoised signal from the posterior probability distribution function. Because the resolution of the maximum posterior probability estimation based on generalized Gaussian distribution cannot be achieved directly, Newton-Raphson law is used to obtain the resolution of the real signal wavelet coefficients. Compared by Signal noise ratio (SNR) and Mean square error (MSE), the results indicate that the proposed method can effectively remove the infrared image noise and the performance is much better than the wavelet soft threshold method and wavelet hard threshold method.