A new computational approach for edge-preserving image decomposition

Abstract

Decomposition of an image into its cartoon part and texture part has been an interesting area of research. It is an important pre-processing step in many computer vision and image processing techniques such as image segmentation, pattern matching, object recognition, tone mapping as both cartoon and texture parts contain two different kinds of information. Significant work is already available in the literature. At the time of decomposition of an input image into its cartoon and texture part, we required texture-smoothing along with edge-preservation. Most of the approaches available in the literature establish a trade-off between the edge preservation and the texture-smoothing. This common drawback of the existing approaches motivates us to design a new image decomposition method by which we can achieve texture-smoothing without having any loss of edge information in cartoon part. To achieve this aim we introduce a new approach based on Joint bilateral filter and DCT (Discrete cosine transform). We show experimental results on several images to show the effectiveness of our method and comparison with some state-of-the-art methods.