Generalizing Spectral Difference Method for Multispectral Image Demosaicking and Analyzing the Role of MSFA Patterns

Abstract

Multispectral images contain more content and details in comparison to color images and are observed to be beneficial for several applications. However, large size and cost restricts the wider usage and adoption of the multi-sensors multispectral imaging systems. Multispectral image demosaicking methods can help develop single-sensor equivalent systems at low cost. In this work, we present an efficient generic method for multispectral image demosaicking by extending the spectral difference method and we also present the analysis of multispectral filter array (MSFA) patterns on the performance, evaluated using two different metrics and by considering two different benchmark multispectral images datasets. Experimental results demonstrate the efficacy of the proposed method, except in cases when non compact MSFA patterns are used.