Cave systems in phong nha ke bang area

VNƯ Jo u rn al o f Science, E a rth Sciences 28 (2012) 264-275 Assessing the feasibility o f increasing spatial resolution of remotely sensed image using HNN super-resolution mapping combined with a forward model Nguyen Quang Minh* Faculty o f Surveying and Mapping, Hanoi University o f M ining and Geology Received 03 September 2012 Revised 24 September 2012; accepted 15 October 2012 A bstract Spatial resolution o f land covers from remotely sensed images can be increased using super-resolution mapping techniques for soft-classified land cover proportions A further development of super-resolution mapping techniques is downscaling the original remotely sensed image usmg super-resolution mapping techniques with a forward model In this paper, the model for increasing spatial resolution o f remote sensing multispectral image is tested with real SPOT imagery for an area in Bac Giang Province, Vietnam in order to evaluate the feasibility o f application o f this model to the real imagery The soft-classified land cover proportions obtained using a fuzzy c-means classification are then used as input data for a Hopfield neural network (HNN) to predict the multispecừal images at sub-pixel spatial resolution Visually, the resulted image is compared with a SPOT panchromatic image acquired at the same time with the multispecừal data The predicted image is apparently sharper than the original coarse spatial resolution image Keywords: Hopfield neural network optimisation, soft classification, image downscaling, forward model m apping such as Elad a n d F eu er [1], Tipping Introduction and Bishop [2] A lthough w idely applied in image processing, these approaches are hardly Spatial resolution o f im age and photos can be increased by the super-resolution algorithm s applicable In the image processing context, im age super- sensed m ultispectral (M S) im agery because o f resolution com m only refers to the process o f the lack o f a sequence o f im ages o f the scene at using a set o f cross-correlated coarse spatial the same or sim ilar tim es The only feasible resolution im ages o f the sam e scene to obtain a application o f the super-resolution approaches single higher spatial resolution im age There are using image sequences is for hyperspectral numerous im agery [3] For other com m on m ultispectral studies on such super-resolution for super-resolution o f rem otely rem otely sensed im agery, only few m ethods for * Tel-84-982721243 E-mail: nguyenquangminh@humg.edu.vn in creasin g the spatial reso lu tio n to sub-pixel 264 265 N.Q Mirth / V N U Journal of Science, Earth Sciences 28 (2012) 264-275 level have been proposed such as a Point spatial resolution o f the original MS image Spread Function-derived convolution filter [4], inừoduced by N guyen Q uang M inh et al [17] segmentation technique [5], and geostatistical The new m odel is based on the HNN super­ method [ ], resolution Sub-pixel spatial resolution land cover maps can be predicted using super-resolution m apping techniques The input data for super­ resolution m apping are com m only the land cover proportions estim ated by soft- classification [7], There is a list o f super­ resolution m apping inừoduced m axim isation techniques including [ ], spatial linear have been dependence optim isation techniques [9], H opfield neural netw ork (HNN) m apping technique IS from unsupervised soft-classification com bined with a forward m odel using local end-m em ber spectra [15,16] The m ethod is exam ined with a degraded rem ote sensing im age and both visual and statistical evaluations show n a good result However, there still exist some concerns about the feasibility o f the model because it is not tested in a m ore com plicated landscape with different kinds o f land cover features w hich are varying in sizes and shapes as well as specfral histogram characteristics This paper, therefore, is to optim isation [ 11 ], genetic algorithm s [ 12 ] and im plem ent the test o f the algorithm in a feed-forw ard com plicated landscape optim isation [ 10 ], neural tw o-point netw orks [13], The supplem entary data are also supplied to H N N to produce more accurate sub-pixel land cover m aps such as m ultiple sub-pixel shifted image [14], fused and panchrom atic (PAN ) im agery [15,16] These latter approaches produce a synthetic M S or PA N im age as an interm ediate step for super-resolution m apping based on a forward model and then these im ages are com pared w ith the predicted and observed MS or PAN im ages to produce an accurate sub­ pixel image classification G eneral m odel The proposed m odel is an extension o f the super-resolution m apping approach based on H NN optim isation T he prediction o f a MS im age at the sub-pixel spatial resolution is based on a forw ard m odel w ith local specfra as w as used in N guyen et al., 2006 [15], In addition to the goal functions and the proportion consừ aint o f the H N N for super- The creation o f the predicted M S and then resolution m apping, a reflectance constraint is PAN im age by a forw ard m odel suggested a used to retain the brighừiess values o f the possibility to im plem ent a super-resolution for original M S image the MS image A m ethod for increasing the 266 N.Q M in h / V N U Journal of Science, Earth Sciences 28 (2012) 264-275 Spatial convolution Synthetic MS image (4 m) SR MS image (2 m) Figure General model for super-resolution MS imagery prediction Figure presents the H N N sub-pixel MS spatial resolution of the original image im age prediction algorithm The procedure is as Follow ing a com parison o f the observed and follows: From the MS im ages at the original synthetic MS area produced to retain the brightness value o f the proportion images are predicted using a softclassifier A set o f local end-m em ber specừa pixels o f the original M S image The process is repeated until the energy function o f the HNN v a lu e s is calculated b a sed on the estim ated land is m inim ised and the synthetic M S image is cover proportions and the original M S image generated L and spatial resolution, land cover proportions error value is A dem onsfration o f the algorithm for an image o f 2x2 pixels can be descnbed in Figure m apping w ith a zoom factor z to produce the Firstly, the soft-classification predicts land land cover resolution HNN map at From the for the used an super-resolution the then im ages, to constrain are cover MS sub-pixel super-resolution spatial land cover m ap at the first iteration, an estim ated MS im age (at the sub-pixel spatial resolution) is then produced using a forw ard m odel and the estim ated local end-m em ber spectra The estim ated MS image is then convolved spatially to create a synthetic M S im age at the coarse cover proportion as in Figure lb from the MS specừal image as in Figure la There are two land covers in this im age called Class A and Class B From the land cover proportions in Figure lb , the land cover classes at sub-pixel level are predicted as in Figure Ic w here a pixel is divided into x sub-pixels and the 2x2 pixels im age is super-resolved to 16x16 pixels 267 N.Q M inh Ị V N U Journal of Science, Earth Sciences 28 (20Ĩ2) 264-275 land cover image o f Class A and Class B The and it is possible to produce a new spectra! brightness o f the new 16x16 pixels image is image by assigning all the sub-pixels belonging predicted using end-m em ber spectra (standard to Class A the brightness value o f 35 and the brightness for the Class A and B in this area o f sub-pixels o f Class B the brighừiess value o f 50 the image) For exam ple, the brightness o f the as in Figure Id pure pixel o f Class A is 35 and Class B is 50 100% land coverA 62.5% land cover A 37.5% land cover B 50% land coverA 100% land coverB 50% land cover B (b) Figure Creation o f !6>