Integrating the Concept of Guided Image Filter and Coefficient Thresholding for Image Denoising

Image acquisition techniques introduce various types of artifacts and noise such as additive white gaussian noise, salt and pepper noise etc, so image denoising is an essential preprocessing step in digital image processing. It is clear from the background study of denoising, conventional methods are not much effective in reducing the noise in the image. In this work, a novel approach which integrates the concept of guided image filter and coefficient thresholding for the removal of different types of noise. Guided image filter is a spatial domain method and coefficient thresholding is a wavelet domain method. Guided image filter has no gradient distortion and has better performance near the edges. Inorder to enhance the output of the guided image filter, wavelet based edge detection is performed. The output obtained by wavelet based edge detection has better visual quality than the one obtained by conventional edge detectors. Also variance of the image has been calculated and smooth and texture regions are separated for coefficient thresholding. Threshold values are calculated and coefficient thresholding is performed. The tool used for this denoising algorithm is MATLAB R2012b. The denoised output obtained by this method has high peak signal to noise ratio (PSNR). By this proposed method, it is possible to get a denoised image which can be used for different applications such as haze removal, feathering etc.