Leonardo Santurri

Sensitivity of Pansharpening Methods to Temporal and Instrumental Changes Between Multispectral and Panchromatic Data Sets

In this work, the authors investigate the behaviors of the two main classes of pansharpening methods: those based on component substitution (CS) or spectral methods and those based on multiresolution analysis (MRA) or spatial methods, in the presence of temporal and/or instrumental misalignments between the multispectral (MS) and panchromatic (Pan) data sets, that is, whenever MS and Pan are not jointly acquired at the same time and/or from the same platform. Starting from the mathematical formulation of CS and MRA pansharpening and from the spectral model between the Pan and MS channels, estimated through the multivariate linear regression between MS and spatially degraded Pan, it is proven that both CS and MRA methods may lose geometric sharpness in the case of a spectral mismatch, but spatial methods preserve the spectral diversity of the original MS data set regardless of the date or instrument of Pan image acquisition. Conversely, spectral methods also suffer from a loss of spectral fidelity to the original MS data set that is inversely related to the success of the spectral match between MS and Pan, measured by the coefficient of determination of the multivariate regression. An experimental setup exploiting GeoEye-1 and QuickBird data sets demonstrates the validity and intrinsic limitations of the proposed theoretical models. Depending of the target application, one class of methods may be preferred to another: Whenever spectral fidelity of pansharpened products to the original MS data sets is crucial, spectral methods should be avoided, and spatial methods are to be preferred.

Low-complexity lossless/near-lossless compression of hyperspectral imagery through classified linear spectral prediction

This paper presents a novel scheme for lossless/near-lossless hyperspectral image compression, that exploits a classified spectral prediction. MMSE spectral predictors are calculated for small spatial blocks of each band and are classified (clustered) to yield a user-defined number of prototype predictors for each wavelength, capable of matching the spatial features of different classes of pixel spectra. Unlike most of the literature, the proposed method employs a purely spectral prediction, that is suitable for compressing the data in band-interleaved-by-line (BIL) format, as they are available at the output of the on-board spectrometer. In that case, the training phase, i.e., clustering of predictors for each wavelength, may be moved off-line. Thus, prediction will be slightly less fitting, but the overhead of predictors calculated on-line is saved. Although prediction is purely spectral, hence ID, spatial correlation is removed by the training phase of predictors, aimed at finding statistically homogeneous spatial classes matching the set of prototype spectral predictors. Experimental results on AVIRIS data show improvements over the most advanced methods in the literature, with a computational complexity far lower than that of analogous methods by other authors.

Spatial resolution enhancement of ASTER thermal bands

Image fusion aims at the exploitation of the information conveyed by data acquired by different imaging sensors. A notable application is merging images acquired from space by panchromatic and multi- or hyper-spectral sensors that exhibit complementary spatial and spectral resolution. Multiresolution analysis has been recognized efficient for image fusion. The Generalized Laplacian Pyramid (GLP), in particular, has been proven as the most efficient scheme due to its capability of managing images whose scale ratios are fractional numbers (non-dyadic data) and to its simple and easy implementation. Data merge based on multiresolution analysis, however, requires the definition of a model establishing how the missing spatial details to be injected into the multi-spectral bands are extracted from the panchromatic image. The model can be global over the whole image or depend on the local space-spectral context. This paper reports results on the fusion of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data. Each of the five thermal infrared (TIR) images (90m) is merged with the most correlated visible-near infrared (VNIR) image (15m). Due to the 6:1 scale ratio, the GLP has been utilized. The injection of spatial details has been ruled by means of the Spectral Distortion Minimizing (SDM) model that minimizes the spectral distortion between the resampled and fused images. Notwithstanding the lack of a spectral overlap between the VNIR and the TIR bands, experimental results show that the fused images keep their spectral characteristics while the spatial resolution is enhanced.

An improved H.263 video coder relying on weighted median filtering of motion vectors

The impact of a regularization of motion vectors (MVs) on the performance of a block-DCT based video coder (H.263) is addressed. Postprocessing is accomplished by exploiting both the spatial correlation of the vector field and the confidence of the estimated block vectors. A previously proposed adaptive scheme for MV smoothing, based on the theory of vector median filters, is adjusted and embedded into an H.263 coder. With a bit stream that is perfectly H.263-compatible, results are improved, especially for very low bit rates and complex motion of the scene.

Full-Scale Assessment of Pansharpening Through Polynomial Fitting of Multiscale Measurements

Pansharpening techniques aim at improving the spatial resolution of a multispectral data set (MS) by using a panchromatic image (Pan) acquired on the same scene with a greater spatial resolution and consequently a lower ground sample distance (GSD). Usually, a quantitative assessment of the fused products cannot be directly performed because of the lack of a reference MS data set with the same GSD of the Pan. A well-known solution is Walds protocol: The original MS and Pan are spatially degraded, the reducing factor being the ratio between their GSDs. Pansharpening is then performed between the reduced MS and Pan data sets, and the fused products are compared with the original MS, which can be used as reference. In this protocol, fusion performances are assumed to be independent of the scale so that the results at the reduced scale are an estimation of those at the original resolution. This hypothesis can be more or less reliable, depending on the sensor and/or the scene content. The objective of this paper is to propose a new methodology to infer the unknown performances of a pansharpening method at full scale. For this purpose, multiple sets of fused images are computed at degraded scales by downsampling Pan and MS data sets by means of the sensor modulation transfer function. Multiscale quality/distortion measurements are fitted by linear and quadratic polynomials in order to extrapolate their full-scale values. Once the proposed protocol has been assessed in the presence of reference originals, the obtained results are extended to the case where the reference image is not available.

Influence of spatial resolution on pan-sharpening results

Pan-sharpening techniques improve the spatial resolution of a Multispectral image (MS) by using a Panchromatic image (PAN) of the same scene contemporaneously acquired at higher resolution. Usually, a quantitative assessment of the resulting fused MS image cannot be directly performed because of the lack of a reference MS. Walds protocol offers a possible solution: original Pan and MS are spatially degraded, the reducing factor being the ratio between their resolutions. Pan-sharpening is then performed between the reduced MS and PAN images. The quality of the fused products is then evaluated by comparing them with the original MS used as reference, by assuming the hypothesis that the performances of the pan-sharpening methods are independent from scale. The objective of this work is to propose a methodology to verify this hypothesis. For this aim, pan-sharpening performances when varying spatial resolution are investigated and a viable strategy to devise pansharpening performances at full scale is suggested.

Spectral distortion evaluation in lossy compression of hyperspectral imagery

Goal of this work is to investigate lossy compression method- ologies from the viewpoint of spectral distortion introduced in hyperspec- tral pixel vectors, besides that of radiometric distortion. The main result of this analysis is that, for a given compression ratio, near-lossless methods, i.e., with constrained pixel error, either absolute or relative, are more suit- able for preserving the spectral discrimination capability among pixel vec- tors, which is perhaps the main source of spectral information. Therefore, whenever a lossless compression is not practicable, near-lossless compres- sion is recommended in such applications where spectral quality is crucial. Data compression is gaining an ever increasing relevance for the remote sensing community. Since technological progresses allow observations of the Earth to be available at increasing spa- tial, spectral, radiometric, and temporal resolutions, the asso- ciated data volume is growing much faster than the transmis- sion bandwidth does, either bandwidth of the downlink with the ground station for satellite platforms, or of the digital network supporting the data distribution. The introduction of data com- pression can alleviate bandwidth requirements at the price of a computational effort for encoding and decoding, as well as of a possible loss of quality. It is a widespread belief that the new generation of spaceborne imaging spectrometers (NASA/JPLs MODIS on Terra and Aqua, NASA/GSFCs HYPERION on EO-1, and ESAs MERIS on EnviSat) will create several prob- lems, on one side for on-board compression and transmission to ground stations, on the other side for an efficient dissemina- tion and utilization of the outcome hyperspectral imagery. In fact, the huge amount of data due to moderate ground resolu- tion, but extremely high spectral resolution (around 10nm), to- gether with the high radiometric resolution originates an amount of data of approximately 300 − 400 bytes/pixel. Therefore, the use of advanced compression techniques and of suitable analy- sis/processing procedures for dissemination to users of thematic information is mandatory. Compression algorithms are said to be fully reversible (loss- less) when the data that are reconstructed from the compressed bit stream are identical to the original, or lossy otherwise. The difference in performance expressed by the compression ratio (CR) between lossy and lossless algorithms can be of one or- der of magnitude without a significant visual degradation. For this reason lossy algorithms are extremely interesting and are used in all those applications in which a certain distortion may be tolerated. Actually these algorithms are more and more pop- ular and their use is becoming widespread also in such remote sensing applications as those in which it was rightly believed, Work supported in part by grants of the Italian Space Agency (ASI) and of the

Full scale assessment of pansharpening methods and data products

Quality assessment of pansharpened images is traditionally carried out either at degraded spatial scale by checking the synthesis property ofWald’s protocol or at the full spatial scale by separately checking the spectral and spatial consistencies. The spatial distortion of the QNR protocol and the spectral distortion of Khan’s protocol may be combined into a unique quality index, referred to as hybrid QNR (HQNR), that is calculated at full scale. Alternatively, multiscale measurements of indices requiring a reference, like SAM, ERGAS and Q4, may be extrapolated to yield a quality measurement at the full scale of the fusion product, where a reference does not exist. Experiments on simulated Pl´eiades data, of which reference originals at full scale are available, highlight that quadratic polynomials having three-point support, i.e. fitting three measurements at as many progressively doubled scales, are adequate. Q4 is more suitable for extrapolation than ERGAS and SAM. The Q4 value predicted from multiscale measurements and the Q4 value measured at full scale thanks to the reference original, differ by very few percents for six different state-of-the-art methods that have been compared. HQNR is substantially comparable to the extrapolated Q4.

Detector shape in hexagonal sampling grids

Recent improvements in CCD technology make hexagonal sampling attractive for practical applications and bring a new interest on this topic. In the following the performances of hexagonal sampling are analyzed under general assumptions and compared with the performances of conventional rectangular sampling. This analysis will take into account both the lattice form (squared, rectangular, hexagonal, and regular hexagonal), and the pixel shape. The analyzed hexagonal grid will not based a-priori on a regular hexagon tessellation, i.e., no constraints will be made on the ratio between the sampling frequencies in the two spatial directions. By assuming an elliptic support for the spectrum of the signal being sampled, sampling conditions will be expressed for a generic hexagonal sampling grid, and a comaprison with the well-known sampling conditions for a comparable rectangular lattice will be performed. Further, by considering for sake of clarity a spectrum with a circular support, the comparison will be performed under the assumption of same number of pixels for unity of surface, and the particular case of regular hexagonal sampling grid will also be considered. Regular hexagonal lattice with regular hexagonal sensitivity shape of the detector elements will result as the best trade-off between the proposed sampling requirement. Concerning the detector shape, the hexagonal is more advantageous than the rectangular. To show that a figure of merit is defined which takes into account that the MTF (modulation transfer function) of a hexagonal detector is not separable, conversely from that of a rectangular detector. As a final result, octagonal shape detectors are compared to those with rectangular and hexagonal shape in the two hypotheses of equal and ideal fill factor, respectively.

Near-lossless compression of multi/hyperspectral image data through a fuzzy-matching-pursuit interband prediction

In this work, near-lossless compression yielding strictly bounded reconstruction error, is proposed for high-quality compression of remote sensing images. A space-varying linear-regression prediction is obtained through fuzzy-logic techniques as a problem of matching pursuit, in which a predictor different for every pixel is obtained as an expansion in series of a finite number of prototype nonorthogonal predictors, that are calculated in a fuzzy fashion as well. To enhance entropy coding, the spatial prediction is followed by context-based statistical modeling of prediction errors. Performance comparisons with JPEG 2000 and previous works by the authors, highlight the advantages of the proposed fuzzy approach to data compression.