Can Multispectral Imaging been Inspired by Digital Camera?

Data:02-07-2020  |  【 A  A  A 】  |  【Print】 【Close

In recent years, multispectral filter array (MSFA) imaging is an emerging technique which inspired by digital cameras featuring color filter array (CFA). However, such method suffers from multispectral demosaicking from the captured raw data, and the sparseness is proportional to the number of bands, which lead to rich information but blurry images.


Dr. Bangyong Sun from Xi'an Institute of Optics and Precision Mechanics (XIOPM) of the Chinese Academy of Sciences (CAS) proposed 9-band MSFA imaging system, as well as a demosaicking algorithm to reconstruct a clear multispectral image. This work was published in the journal Mechanical Systems and Signal Processing.

As MSFA imaging needs to recovering information from other bands, which is known as multispectral demosaicking, the distribution of MSFA and the filter spectral sensitivity functions (SSFs) should be carefully selected.


During the imaging, The nine spectral bands can be classified into two categories, densely sampled band and sparsely sampled band.


The idea of the reconstructing algorithm is to recover the densely sampled band by the gradients of neighboring sampled pixels, then the recovered densely sampled band with the guided filter and residual interpolation plays a role of guided image to demosaicking other sparsely sampled band. Finally, the spectral reflectance values can be estimated from the multispectral image and the characterization matrix.

 Visual comparison of the demoaicing results of a cropped region of Image statue. (Image by XIOPM)


The proposed method is validated using CAVE dataset form Columbia university with 32 images, where supreme performance were shown. Compared to digital cameras, MSFA technique is capable of using more bands as most types of CFAs, and have advantages on imaging speed, size and cost. On the other hand, multispectral imaging techniques significantly reduce the hardware complexity with reasonable spectra precision compared to classical hyperspectral imaging methods. In the coming years, such method can be employed in object detection, material analysis and mechanical system diagnosis.