Dimensionality Reduction of Hyperspectral Images Using Pooling

Abstract

Hyperspectral image having huge numbers of narrow and contiguous bands involves high computation complexity in processing and analysing the image. Hence dimensionality reduction is applied as an essential pre-processing step for hyperspectral data. Pooling is a technique of reducing spatial dimension and successfully applied in convolutional neural network. There are various types of pooling strategies present viz. max pool, mean pool and having their respective merits. In the present article, the concept of pooling is applied in the spectral dimension of the hyperspectral data to reduce the dimensionality and compared the result with standard reduction process like principal component analysis. Different pooling methods are applied and compared across and the mean pooling is found to be performing better. The results are compared in terms of overall accuracy and execution time.