Fractional-order Differintegral Based Multiscale Retinex Inspired Texture Dependent Quality Enhancement for Remotely Sensed Images

Non-integer orders for fractional differintegral based adaptive filtering are very useful for addressing the various levels of textural discrimination in any image. This paper addresses a novel contribution for image enhancement by harvesting the collective benefits of fractional-order differentiation (FOD) as well as fractional-order integration (FOI). A 2-D fractional-order (FO) “Differintegration” (FODI) based dual-operator is proposed for imparting “on-demand adaptive-filtering” through spatial masking for adaptive textural boosting along with contrast enhancement. In addition, FODI based adaptive filtering is also employed for improving the conventional Multiscale Retinex approach. Consequently, the second novel contribution in this paper is the proposed framework for Differintegration based Fractional-order Multi-scale Retinex (DFMSR) for effective reflectance channel computation. The proposed DFMSR module is readily compatible with other image pre-processing methods for effective suppression of airborne and/or illumination based visual imperfections. An optimal fractional-order two-dimensional adaptive filtering mechanism is proposed in this paper. The proposed model is highly modular. So, it can also be pipelined in parallel manner, along with any well-established state-of-the-art contrast enhancement approach. Total quality improvement for remotely sensed textural data is achieved in this work. Being highly modular, four different possible variants of proposed approach are also discussed in this draft, so that a generalized solution can be identified for overall quality improvement for diverse variety/domains of the visual data or images.