PhD Theses
Permanent URI for this collectionhttp://drsr.daiict.ac.in/handle/123456789/2
Browse
3 results
Search Results
Item Open Access New techniques for spatial resolution enhancement of hyperspectral images(Dhirubhai Ambani Institute of Information and Communication Technology, 2014) Patel, Rakeshkumar Chandulal; Joshi, Manjunath V.Vision plays a central role for human perception and interpretation of the world. With the beginning of the space age during late 1950s presented opportunities for remote sensing of earth resources [1]. Multispectral sensors image the earth in a few strategic areas of the electromagnetic spectrum using a small number of carefully chosen spectral bands (typically 3 to 10) spread across the visible and infrared regions of the electromagnetic spectrum. These bands are not contiguous and omit many wavelength ranges. The ability of a sensor to distinguish between wavelength intervals in the bands describes the spectral resolution. Higher the spectral resolution, narrower the wavelength range for a band. The spectral resolution determines the materials discrimination ability of the sensor. The high spectral resolution of multispectral imaging was found useful for ground-cover classification, mineral exploration, and agricultural assessment to name a few. In remote sensing, acquisition of image details helps accurate localization and correct identification of minerals and vegetation, and hence better classification of the landmass. The hardware in the remote sensing sensors limits the amount of detailed information captured (i.e., spatial resolution) whenever the spectral width of the acquired image is small. The size of the ground area expressed as meter x meter represented by a single pixel in an image defines the spatial resolution of the image. Smaller the ground area per pixel means higher spatial resolution. It depends on the sensor design and the height of the sensor above the ground. To increase the spatial resolution without affecting spectral resolution, the sensor should have a small instantaneous field of view (IFOV). But this reduces the signal power falling on the detector and hence signal to noise ratio is reduced. One can increase signal to noise ratio by widening the bandwidth of the acquired spectral band, but this reduces the spectral resolution of the image. Thus, there exists a trade-off between spatial and spectral resolutions of remotely sensed images. Multispectral images provide higher spectral resolution (of the order of 100 nanometers) but they suffer from low spatial resolution. Improved versions of these early multispectral imaging sensors known as hyperspectral imager provide spectral width of 10 nanometers for each band of hyperspectral image (HSI). Having very high spectral resolution they provide ample spectral information to identify and differentiate spectrally unique materials [1]. Hence, presently they are used in wide range of military and civilian applications that include target detection and tracking of objects, agriculture planning, forest inventory, mineral exploration, and urban planning to mention a few. Similar to the multispectral images hyperspectral images also suffer from low spatial resolution due to very small spectral width. Many times it is not feasible to capture the spatially high resolution (HR) images due to the limitation in implementation such as requirement of large memory, higher transmission bandwidth, high power requirement and higher camera cost. Since HR imaging leads to better analysis, classification and interpretation one may look for algorithmic approaches to obtain the HR images. Hence, we need to perform post processing of the hyperspectral images (HSIs) to increase their spatial resolution and hence the image details, without affecting their spectral resolution. Super-resolution enhancement refers to an algorithmic approach to increase the spatial resolution of a low spatial resolution image by using either multiple low-resolution (LR) observations or using a database of high and low-resolution images. Many satellites like WorldView-1, 2, 3, SPOT, Landsat, Quickbird, Ikonos, etc. capture two different types of images, namely, the high spectral but low spatial resolution multispectral (MS) images and high spatial but low spectral resolution registered panchromatic (PAN) image (auxiliary image). The reason behind configuring satellite sensors this way is to reduce weight, cost, bandwidth and complexity of the satellite. In this thesis, we develop different algorithms to enhance the spatial resolution of hyperspectral images. To start with, we first address the problem of enhancing the spatial resolution of MS images by merging information from PAN image, called multiresolution fusion, using two step approach. In the first step, the high-resolution edge details of the fused multispectral image are learned in the form of an initial estimate using discrete wavelet transform and compressive sensing (CS).We know that PAN and MS images are obtained from the same geographical region with the difference that PAN image is acquired with high spatial resolution. We assume panchromatic image and the multispectral bands are in registered form. This results in high spatial correlation between the MS observation and the coarser part of the PAN image. Our approach uses CS technique to obtain the detailed wavelet coefficients of the MS image by assuming the same sparseness of MS image with the coarser level, as well as detailed level of the PAN image. This way we obtain initial estimate of the fused image. To better preserve spatial homogenity, in the second step, we regularize it further to obtain the final solution. We restrict the solution space by using maximum a posteriori - Markov random field (MAP-MRF) approach that imposes smoothness constraint on the fused image by using first order neighborhood for MRF prior. We make use of the initial estimate to obtain the MRF parameter. Hyperspectral images are used for the same purpose as do MS images, but they have very high spectral dimensions that enables distinguishing the spectrally unique materials. The statistical classification (clustering) methods often used with multispectral images can also be applied to hyperspectral images by handling their high dimensionality [2]. Hyperspectral sensors like AVIRIS, Hyperion, HYMAP do not capture auxiliary HR image. In such circumstances we cannot use fusion to increase the spatial resolution of HSI. Our remaining three techniques discussed in this thesis deal with spatial resolution enhancement of HSIs using the concept of super-resolution without making use of auxiliary HR image (i.e. PAN image). The goal of super-resolution (SR) is to recover high-frequency details lost during image acquisition process which in turn increases the number of pixels in the input image. This is an inverse problem wherein the original high-resolution (HR) image has to be retrieved from the observed low-resolution data. There are large number of HR images which are consistent with the LR image. Hence, while solving such an ill-posed inverse problem, knowing the forward model alone is not sufficient to obtain a satisfactory solution. We need to add proper constrains by using priors to limit the solution space. This procedure to get a solution of the inversion problem in accordance with the prior information is called regularization. Selection of appropriate model as the prior information and use of regularization helps to obtain improved solution. In our work, we have considered different kinds of priors in regularization in order to obtain improved solution. We make use of compressive sensing theory and estimated wavelet filter coefficients to obtain SR results for HSIs. To reduce high computational load due to large number of spectral bands of HSIs, we use principal component analysis (PCA) to reduce the dimensions and work on reduced dimensional space to obtain SR results. In the first method, we use CS based approach to obtain initial SR of the most informative PCA image which represents highest spectral variance of the HSI. Here we use LR and HR raw dictionaries having large number of atoms in the CS based framework. Using the sparsity constraint, LR test patch is represented as a sparse linear combination of relevant LR dictionary elements adaptively. Assumption of same sparsity to LR and HR images with respect to their dictionaries gives SR image as an approximate. The final SR solution is obtained using a regularization in which AR prior model parameters are obtained from the initial SR estimate obtained using CS. SR results of the other significant PCA components are obtained using the same AR parameters and using the regularization framework. While regularization, decimation process is modeled as an averaging process. The decimation process modeled as the averaging process represents the aliased pixel in the low-resolution image by averaging the corresponding pixels in the high-resolution image. This means, the point spread function (PSF) of sensor considered is square and is same for all spatial and spectral region. However, in practice PSF depends on several factors like camera gain, zoom factor, and imaging hardware etc. This motivates us to estimate the PSF i.e., the aliasing and then perform SR. Here our CS based approach is further extended to obtain initial SR of all significant PCA components that represent most of the spectral variance (98 %) of the HSI where the aliasing is estimated for all the significant PCA components. Here we use jointly trained LR and HR dictionaries having very less number of atoms (i.e. 1000) using training algorithm called K-singular value decomposition (K-SVD). This is onetime and offline procedure. Regularization using our new prior i.e., Gabor prior preserves various bandpass features in the final SR image. Also the use of estimated entries of degradation matrix in the form of PSF represents imaging hardware more effectively in image observation model. This leads to better solution of final SR result. Finally, we address learning based super-resolution in wavelet domain using estimated wavelet filter coefficients. In this work, we estimate the PSF in the form of wavelet filter coefficients to take care of the degradation between LR and HR images. Here we do not consider spatially varying PSF, which is quite involved as this requires the estimation of PSF at every pixel. However, the space invariant PSF is estimated for individual spectral bands. The estimated filter coefficients are also used to learn high frequency details by using the HR training images in wavelet domain. This gives us an initial estimate of SR image for each HS band and they are used in deriving the sparse coefficients that are used as priors. The final SR image is obtained using the sparsity based regularization that also has the observation model constructed using the estimated filter coefficients. Since the cost function is differentiable, a simple gradient descent optimization is used to obtain final solution. We show the computational advantage of the proposed algorithm.Item Open Access Detection of tampering in digital images using feature based hash generation(Dhirubhai Ambani Institute of Information and Communication Technology, 2014) Mall, Vinod Kumar; Roy, Anil Kumar; Mitra, Suman KumarRecent years have witnessed an exponential growth in the use of digital images due to development of high quality digital cameras and multimedia technology. Easy availability of user-friendly image editing software has made modification of images a popular child’s play. In this environment, the integrity of an image cannot be taken for granted. Malicious tampering has serious ramification in legal documents, copyright issues, photojournalism, celebrities’ lifestyles, fashion statements, beauty and fitness products, entertainment sector, medical science, biometric images and forensic cases. The proposed work is based on features based hash generation of a digital image and thereby detection and localization of image tampering that is done with malicious intention, no matter howsoever small the tampering is. The hashing algorithm generates a short binary string by extracting the feature vectors of the image and mapping them into robust hash values. This hash mapping meets the two requirements of being sensitive to tampering in the image and being robust against content preserving manipulations. The first part of the work uses correlation coefficient to generate a hash representation of the image which is then utilized for tampering detection. This method is extended later to use properties of Singular Value Decomposition (SVD) to derive the hash values. SVD enables us to generate key based hash values. We show that this hash can be used in secured transmission of image information on the public network. We emphasize that the malicious tampering is generally done on the structural part of the image. Therefore, in the third part of our study, we have used Canny Edge Detector to extract the features of the image by detecting the edges present in it. Having the edges identified, tampering detection and localization are carried out. This method of tampering detection and localization has been found to be promising. The last part of the thesis discusses Comprehensive Image Index which is able to detect multiple types of tampering, viz., brightness, contrast, and structural, simultaneously. We observe that a structural tampering is always followed by brightness and contrast changes in the image. We also establish that even brightness or contrast or both change can be seen as a malicious tampering if exceeds a threshold value. The sensitivity of the technique and its robustness have been discussed quantitatively for each method.Item Open Access Some new methods for multi-resolution image fusion(Dhirubhai Ambani Institute of Information and Communication Technology, 2016) Upla, Kishorkumar Parsottambhai; Joshi, Manjunath V.It is always an interest of a mankind to explore the various resources of the earth such as minerals, agriculture, forestry, geology, ocean etc., Before the invention of the remote sensing, in order to analyze the various resources it was required to visit the field to take the different forms of data samples and later on those were processed further. The revolution in terms of photography using satellite made it possible to view the earth's surface without being in touch with the area of interest. With the help of satellite technology it is also possible to view the locations on the earth which are not accessible by the mankind. Remote sensing has effectively enabled the mapping, studying, monitoring and management of various resources present on the earth. It has also enabled monitoring of environment and thereby helping in conservation. In the last four decades, the advancement of the remote sensing technologies have improved the methods of collection, processing and analysis of the data. Remote sensing involves the acquiring of the pictorial data of the earth's surface without any type of physical contact. It provides the information which not only helps in managing and protecting the natural resources but also helpful in the development of land usage in terms of urban planning. One of the major advantages of the remote sensing satellites is the ability to provide the repetitive observations of the same area. This capability is very useful to monitor dynamic phenomena such as cloud evolution, vegetation cover, snow cover, forest _res, etc. A farmer may use thematic maps to monitor the health of his crops without going out to the field. A geologist may use the images to study the types of minerals or rock structure found in a certain area. A biologist may want to study the variety of plants in a certain location. The image acquisition process of remote sensing system consists of sensing the reflected electromagnetic energy from the surface of the earth. The amount of energy reflected from the earth's surface depends on the composition of the material. The variations in the reflected energy are captured by the remote sensing sensors placed in the satellite or aircraft which are then quantized and digitized into the pictorial form i.e., images. The smallest element of an image i.e., pixel corresponds to an area of a few squared meters in the actual scene which is referred to the spatial resolution of the given sensor. The spatial resolution is limited by the instantaneous field of view (IFOV) of the remote sensing system. Smaller the IFOV, lesser is the area covered by sensor and hence the amount of collected light energy is reduced. By keeping the small IFOV, one can increase the amount of light falling on the sensor i.e. photo detector element by increasing the spectral width of the sensor. This results in wider spectral width with high spatial resolution. Alternatively, one can use the sensor with wide IFOV that covers large surface area. This makes the sensor to collect more light energy but the image formed has lower spatial resolution. However, in this case the spectral width of the sensor can be made narrower in order to sense the data in that spectral width which results in an image with high spectral resolution having fine spectral details. The data with narrower spectral width always helps in better classification since the materials present in the scene reflect the light energy of different wavelengths based on their composition. If one can capture the reflected energy at different bands of wavelengths then it provides separate information about the same scene content. However, this set of images obtained at different spectral bands is possible with the compromise of poor spatial resolution. This trade off in high spatial and spectral resolutions imposes the limitations on the hardware construction in the remote sensing sensors. It is always of interest to visualize the content of the scene with high spatial and spectral resolutions. However, constraints such as the tradeoff between high spatial and spectral resolutions of the sensor, channel bandwidth, on board storage capability of a satellite system place the limitations on capturing the images with high spectral and spatial resolutions. Due to this, many commercial remote sensing satellites such as Quickbird, Ikonos and Worldview-2 capture the earth's information with two types of images: a single panchromatic (Pan) and a number of multispectral (MS) images. Pan has high spatial resolution with lower spectral resolution while MS image has higher spectral resolving capability with low spatial resolution. An image with high spatial and spectral resolutions i.e., fused image of MS and Pan data can lead to better land classification, map updating, soil analysis, feature extraction etc. Also, since fused image increases the spatial resolution of the MS image it results in sharpening the image content which makes it easy to obtain greater details of the classified maps. The pan-sharpening or multi-resolution image fusion is an algorithmic approach to increase the spatial resolution of the MS image with the preservation of spectral contents by making use of the high spatial resolution Pan image. In this thesis we address some new multi-resolution image fusion techniques. In multi-resolution image fusion problem, the given MS and Pan images have high spectral and high spatial resolutions, respectively. One can think of obtaining the fused image using these two by injecting the missing high frequency details from the Pan image into the MS image. The quality of the final fused image will then depend on the method used for high frequency details extraction and also on the technique for injecting these details into the MS image. In the literature various approaches have been proposed based on this idea. Motivated from this, we first address the fusion problem by using different edge preserving filters in order to extract the high frequency details from the Pan image. Specifically, we have chosen the guided filter and difference of Gaussians (DoGs) for detail extraction since these are more versatile in applications involving feature extraction, denoising, etc. Using these edge preserving filters we extract the high frequency details from the Pan image and inject them into the upsampled MS image. One of the drawbacks of the fusion methods using edge preserving filters is the upsampling operation required to perform on the MS image before the injection of high frequency details into the same. Since this operation do not consider the effect of aliasing it results in distortions in the final fused image. Solving the problem of fusion by model based approach is accurate since aliasing present due to undersampled MS observation can be taken care of while modeling. Many researchers have used the model based approaches for fusion with the emphasis on improving the fused image quality and reducing the color distortion. In a model based method, the low resolution (LR) MS image is modeled as the blurred and noisy version of its ideal high resolution (HR) fused image. Since this problem is ill-posed, it requires regularization to obtain the final solution. In the proposed model based approach a learning based method that uses Pan data is used to obtain the required degradation matrix that accounts for aliasing. We use sub-sampled as well as non sub-sampled contourlet transform based learning to obtain close approximation to fused image (initial estimate). Then using the proposed model, the final solution is obtained by solving the inverse problem where a Markov random field (MRF) smoothness prior is used for regularizing the solution. We next address the fusion problem based on the concept of self similarity and compressive sensing (CS) theory. In the earlier proposed approach, the degradation matrix entries were estimated by modeling the relationship between the Pan derived initial estimate of the fused MS image and LR MS image which may be inaccurate as the estimate depends on the low spectral resolution Pan data. If the initial fused estimate is derived using the available LR MS image only, then the transformation between the estimated high resolution initial estimate and the observed MS image would be more accurate. This makes the estimated degradation matrix to better represent the aliasing. In this case we obtain the initial estimate using the available LR MS image only. Here, we use the property of natural images that the probability of availability of same or similar information in the current resolution and its coarser resolution is high. We exploit this self similarity concept and combine it with CS theory in order to obtain the initial estimate of fused image which is then used in obtaining the degradation. Finally, in order to better preserve the spatial details and to improve the estimate of fused image, we solve the multi-resolution fusion problem in a regularization framework by making use of a new prior called Gabor prior. Use of Gabor prior ensures features at different spatial frequencies of fused image image to match those of the available HR Pan image. Along with Gabor prior we also include a MRF prior which maintains the spatial correlatedness among the HR pixels