MATLAB

Adaptive Spectral-Spatial Compression of Hyperspectral Image With Sparse Representation

ABSTRACT:

Sparse representation(SR) can change otherworldly marks of hyperspectral pixels into scanty coefficients with not very many nonzero sections, which can productively be utilized for pressure. In this project, a ghostly spatial versatile SR (SSASR) technique is proposed for hyperspectral picture (HSI) pressure by exploiting the otherworldly and spatial data of HSIs. To start with, we build superpixels, i.e., homogeneous locales with versatile sizes and shapes, to depict HSIs. Since homogeneous areas as a rule comprise of comparative pixels, pixels inside each superpixel will be comparative and offer comparable phantom marks.

At that point, the phantom marks of each superpixel can be all the while coded in the SR model to misuse their joint sparsity. Since various superpixels for the most part have diverse exhibitions of SR, their rate-contortion exhibitions in the meager coding will be unique. To accomplish the most ideal by and large rate-mutilation execution, a versatile coding plan is acquainted with adaptively allocate contortions to superpixels. At last, the got meager coefficients are quantized and entropy coded and comprise the last bitstream with the coded superpixel delineate. The trial results more than a few HSIs demonstrate that the proposed SSASR technique outflanks some cutting edge HSI pressure strategies as far as the rate-twisting and unearthly devotion exhibitions.

Leave a Reply

Your email address will not be published. Required fields are marked *