BITS Faculty Publications
Permanent URI for this communityhttp://localhost:4000/handle/123456789/1867
Browse
5 results
Search Results
Item MIRACLE: multi-satellite Island image registration using anisotropic coherence locality enhanced nonlinear diffusion and Mahalanobis distance guided marginalization(Taylor & Francis, 2023-07) Rohil, Mukesh KumarFeature in an image plays a crucial role for geometric image registration. The geo-registration problem becomes difficult for scanty feature islands’ images captured by high and medium spatial resolution remote sensing satellites over deep ocean water. The article presents an automatic multi-satellite image registration methodology for scanty feature island scenes, termed as MIRACLE. In data pre-processing stage, the multi-spectral and reference image are enhanced using anisotropic coherence for better localized feature demarking of island regions. The input multi-spectral images are transformed using Principal Component Analysis (PCA) to maximize the variance information for improved feature matching with reference image. Enhanced features are detected and described using nonlinear diffusion filtering, and the matched control points are pruned using Mahalanobis distance guided marginalization optimization technique. The estimated affine parameters are applied to generate multi-satellite co-registered data products. MIRACLE is evaluated with multi-temporal Indian Resourcesat multi-spectral images and NASA-USGS Landsat−8 OLI panchromatic images that span from 5.0-metre spatial resolution to 15.0-metre spatial resolution and cover the Lakshadweep islands in deep ocean. The visual quality assessment indicates that different island regions are aligned at sub-pixel level registration accuracy. The matching accuracy of MIRACLE is quantified for multi-resolution images and is found to have 2.6% improvement in Correct Matching Ratio (CMR) as compared to the state-of-the-art feature based image registration techniques. The average Root Mean Square Error (RMSE) of island regions after precise geometric correction is found to be 0.45 pixel.Item Effect of Illumination and Blur Change on Markerless Image Registration Methods(IEEE, 2018) Rohil, Mukesh KumarImage registration forms a basis for a wide variety of applications in Computer Vision. The methods used for image registration are generally divided into two categories: 1) Extrinsic: based on some external object placed in an image. 2) Intrinsic: based on image information. Intrinsic methods work upon image features, pixel intensity levels etc. to determine measurements with respect to the requirements of a particular application. This paper presents a comparative analysis of three widely used image registration methods i.e. SIFT, ASIFT and SURF, for intrinsic image registration process. Quality of images describing medical, natural and structured scenes with different illumination and blur conditions is correlated with the performance analysis of the three image registration methods. Results show that the total count of extracted features in an image and correspondences found between two images (for determining the correct number of matches between an image pair, RANSAC algorithm is used for eliminating outliers) decreases with decreasing quality in terms of both blur and illumination conditions. Also, ASIFT outperforms SIFT and SURF in these changing imaging conditions.Item Feature based remote sensing image registration techniques: a comprehensive and comparative review(Taylor & Francis, 2022-08) Rohil, Mukesh KumarEarth observation using remote sensing data is a trending research field across the globe. In this perspective, image registration is a mandatory data processing step for any kind of time series data analytics. Feature-based image registration is one of the prominent categories for superimposing multi-temporal and multi-modal images over each other. The paper provides a comprehensive and comparative survey of feature detection/description techniques for remote sensing images. In addition, outlier removal algorithms are explored in detail to generate putative keypoint correspondences for accurate transformation parameter estimation. The experiments are conducted on multiple remote sensing image pairs comprising visible, Synthetic Aperture Radar (SAR) and infrared images that provide diverse characteristics of the feature target. The co-registration accuracy is quantified for all possible combinations of feature detection/description with outlier removal, and best amalgamation is visually verified to check sub-pixel spatial alignment by swiping multi-temporal or multi-modal images over each other. It has been found that SIFT performs better in optical image registration, whereas A-KAZE feature detection has an upper edge at SAR image registration task. Marginalizing Sample Consensus (MAGSAC) and Mode Guided (MG) based outlier removal techniques achieves better pruning performance for multi-temporal optical and SAR images respectively. The comparative evaluation indicates that Motion Smoothness Constraint (MSC) optimization shows optimal performance for multi-modal remote sensing images. The best possible Root Mean Square Error (RMSE) achieved for multi-temporal optical images is 0.44 pixel whereas for multi-temporal SAR images, it is 0.46 pixels. The RMSE achieved for multi-modal images is 0.48 pixel using combination of SIFT with MSC.Item EPOCH: enhanced procedure for operational change detection using historical invariant features and PCA guided multivariate statistical technique(Taylor & Francis, 2021) Rohil, Mukesh KumarIn this article, we have presented a methodology developed for automatic historical change detection using multi-decadal time-lapse remote sensing images, which we call as EPOCH. The unpaired bi-temporal images are spatially aligned using Mode Improved Scale Invariant Feature Transform (M-SIFT) to achieve sub-pixel co-registration accuracy. The surface changes are detected using Guided Image Filter Enhanced Multivariate Alteration Detection (GIF-MAD). The guidance image is extracted using Principal Component Analysis (PCA), and an operational processing framework is devised to generate change detection map. EPOCH is evaluated with Indian Remote Sensing (IRS) images and Landsat multi-temporal images that observe Earth for more than three decades. The procedure is generalized to detect changes using different satellite images over one of our neighboring planet Mars. EPOCH is compared with state-of-the-art techniques, and found to have closest consensus with ground truth data. The proposed approach achieved an overall accuracy of 90.9% with kappa value of 0.81Item FIRM: Framework for Image Registration Using Multistage Feature Detection and Mode-Guided Motion Smoothness Keypoint Optimization(IEEE, 2021) Rohil, Mukesh KumarRemote-sensing image registration is a pivotal preprocessing step for earth observation data analytics. In this regard, georeferencing corrects the systematic geometric degradation in the image. However, it is difficult to achieve subpixel geometric accuracy across multitemporal scenes. This article focuses on hindmost part of geometric correction that uses reference layer and feature detection in hierarchical stages to improve the geometric fidelity of images at subpixel level. The methodology developed is based on patch affine-oriented fast and brief with mode-guided tiled scale invariant feature transform (MT-SIFT) techniques in a coordinate manner at a multistage processing architecture, which we refer to as FIRM. Motion smoothness constraint (MSC) keypoint correspondence optimization is used in FIRM to remove the outliers at gross stage and estimate segmented affine transformation parameters at finer stage. The automatic coregistration pipeline is evaluated in Indian Resourcesat multispectral camera images covering diverse landscapes. The capability of the designed framework is demonstrated to handle relatively large geometrical error. With more than a decade difference in acquisitions, multitemporal images are superimposed over each other and compared with state-of-the-art feature-based methods. The potential of the proposed approach FIRM is assessed on multisatellite imagery acquired from Resourcesat-2 and Landsat −8. It is observed that the root mean square error (RMSE) between coregistered images is 0.12 pixel at a spatial resolution of 5 m.