Epitome based transform domain Image Denoising

An image modeling scheme named “Epitome” was proposed by Brendan J. Frey and Nebjsa Jojic in 2002. Image epitome is a miniature, condensed version of the image. It is much smaller in size compared to the image but still it contains the most constituent elements of the corresponding image. Image epitome is applied to a wide variety of image processing tasks such as, Image Segmentation, Parts-based Image Retrieval, Image Inpainting etc. Image Denoising is also one of the popular application of image epitome. This article suggests a method to improve epitome based denoising. The state-of-the-art image denoising methods use transform domain processing for better noise removal. This article introduces the transform domain processing along with the epitome based denoising framework. Basis using Orthogonal Locality Preserving Projection (OLPP) are learnt from the epitome and denoising is performed in the OLPP domain. The experimental results with different noise levels suggest a significant amount of improvement over the original epitome based denoising.