Hyperkernel-based intuitionistic fuzzy c-means for denoising color archival document images

In this article, we have addressed the problem of denoising and enhancement of color archival handwritten document images by separating noise from text and background. Indeed, archival document images that originated from scanning or photographing paper documents are mainly digitized in full color mode. Thus, it is necessary to preserve and exploit color information when applying an enhancement method or a denoising technique. Thus, the focus of our work has been to model a color image using a hyperspace. The defined hyperspace formed by the image pixels is obtained by using both topological and color spaces. The novelty of our work lies in exploiting the obtained hyperspace to cluster the extracted low-level features (topological and color) and, thereafter, to separate noise from text and background. Indeed, based on combining the obtained hyperspace with an adapted kernel-based intuitionistic fuzzy c-means (KIFCM) algorithm we have proposed a novel hyper-KIFCM (HKIFCM) method for denoising color historical document images. To illustrate the effectiveness of the HKIFCM method, a thorough experimental study has been firstly conducted with qualitative and quantitative observations obtained from color archival handwritten document images collected from both the Tunisian national archives and two datasets provided in the context of open competitions at ICDAR and ICFHR conferences. Then, we have compared the results achieved with those obtained using the state-of-the-art methods.