This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.
Crucial information barely visible to the human eye is often embedded in a series of low resolution images taken of same scene. Super-resolution enables the extraction of this information by reconstructing a single image, at a higher resolution than is preset in any of the individual images Jain (1989). This is particularly useful in forensic imaging where the extraction of minute details in an image can be solving a crime (Douglas, 2003), and remote sensing imaging where identification of object in high resolution may gives the most precise and reliable spatial information (Andrew, 2001).
Currently, Charge-Couple-Devices (CCDs) are used to capture such high resolution images digitally. Although it is adequate for most of today's applications, in near future this will not be acceptable. Technology of CCD and high precision optics cannot keep up with the demand for images of higher and higher resolution Muresan and Parks (2000).
Thus, the presence of shot noise, which is inevitable in any imaging system, prescribes an upper limit on the resolution of CCDs. Upper limit arises once reducing the area of each CCD which increases the resolution (more CCDs), the signal strength is correspondingly decreased, while the noise strength remains the same (Brian et al, 1995). This limit on the size of each CCD is roughly 50 μm2, and current CCD technology has almost reached this limit (Stark and Oskoui, 1989). In addition, cost is another concern in using high precision optics and CCDs. Launching a high resolution camera into space and on board satellite can be costly, and even risky. It is more cost-efficient to launch a cheaper camera with a lower resolution into orbit if higher resolution images can be obtained on the ground through image processing (Jain, 1989).
Most of super-resolution image reconstruction method consists of three basic components : (i) image registration or motion compensation (ii) interpolation (iii) blur and noise removal (Choudhuri, 2001). Image registration is implemented for motion compensation which used to map the motion from all available low resolution frames to a common reference frame. The motion field can be modelled in terms of motion vectors or as affine transformations. Interpolation refers to mapping the motion-compensated pixels onto a super-resolution grid. Lastly, the third component is needed to remove the sensor and optical blurring in restoration. Figure 1 shows schematic construction of super resolution imaging.
The low resolution frames Y1, Y2, Y3, …. Yn are input motion estimation or registration module, following which the registered image is interpolated onto a high resolution grid. Post processing of the interpolated image through blurs and noise removal algorithms results in the generation of a super-resolution image.
The basic premise of most super-resolution techniques is to combine several images from the same scene considered with low spatial resolution (LR) in order to produce one or several images with a higher resolution (HR). Of course, it can only be
assumed that a HR can be obtained from LR images if they are undersampled and suffer from aliasing. Every LR image samples the scene as a different projection of the same scene on different sampling lattices, so they have different profiles in the aliased frequency range. Thus, none of the LR images can be obtained from the other LR ones because each one contains a certain amount of differential information from the same scene, even though it may be in the aliased frequency range. Super resolution techniques combine the LR images and attempt to recover as much as possible of this differential information to construct the HR result.
There exist many different potential techniques for super-resolution mapping from remotely sensed imagery. A simple approach involves converting a hard-classified image into the vector data model by replacing class object boundaries with vectors. Generalizing these vectors will produce sub-pixel spatial information on land cover. However, not withstanding the problems associated with hard classification, such an approach is under-constrained. Foody (1998) evaluated an interpolation-based technique for predicting the boundary of a lake with sub-pixel geometric precision. However, this approach was similarly under-constrained. In both of the above cases, the algorithm may be subject to effects such as conditional bias and smoothing which may affect the final vector boundary. Aplin and Atkinson (2001) developed a technique for converting the output from a per-pixel soft-classification of land cover into a per-parcel hard classification of land cover objects. Land-line vector data from the Ordnance Survey were used to constrain the placement of the soft proportions within each pixel. This requirement for vector data makes the technique redundant for (i) less developed areas of the world and (ii) updating the vector database.
2.0 THE STATEMENT OF THE THESIS
In remote sensing, to obtain the highly accurate and precise spatial information may require high resolution images. Due to the highest number of price for single high resolution scene and limitation of observation coverage, low number of projects relies on high resolution images that lead to low precision and inaccurate results (Aizawa et al, 1992). This is results from interferences of noises (Kim et al 1990), blurs (Bascle et al, 1996), aliasing (Lucas, 1989) and low number of sampling rates (Eren et al, 1997) degraded the quality of spatial resolution acquired by low resolution images (Jain, 1989), (Calle, 1989). To overcome these problems, promising image processing method like super-resolution is employed to estimate an image at a higher resolution than its present in any of individual images (Irani, 1991, Foody et al, 2005 ).
2.1 Aims Of The Study
a. In order to extract highly accurate and precision spatial information, combination of multiple low resolution images certainly needs good image processing methods which may considering on image registration, aliasing correction, interpolation, restoration and object identifications. Study will lead to construct robust, precise and accurate spatial object identification using super-resolution.
b. To access and evaluate the techniques based on its speed of processing, limitation of its interferences to noise shot, aliasing, sampling rate, motion compensation, blur, noise removal, robustness and precise image registration, interpolation and object identification.
3.0 LITERATURE REVIEW
In super-resolution, all literatures can be broadly divided into methods employed for still images and those for video. All of them are depicted in diverse of applications such as mapping in remote sensing, tomography, video based object tracking, number plate detection and so on. In still images super-resolution, there are 2 main studies been implemented either in 2D or 3D. However in this study, only in 2D super resolution of still images with some discussion and reviews of previous studies.
In 2D super-resolution, Tsai and Huang (1984) were firstly addressed the problem of reconstructing a high resolution image from a sequence of undersampled low resolution still images. They presumed that translational motion and solved registration and restoration problem imply estimating samples on a uniform grid with high sampler. They assumed those images are free from degradation and noise interferences. Latter, Kim et al. (1990) continued this approach with noisy and blurry low resolution observations and developed and algorithm based on weighted recursive least square theory. Then, this method furthers by Kim and Su (1993) who considered the case of different blurs in each of the low resolution observations and use Tikhonov regularization to determine the solution of an inconsistent set of linear equations (Kim and Su, 1993).
Interpolation in super-resolution been investigated by Ur and Gross (1992) who utilized the generalized multichannel sampling theorem to perform a non-uniform interpolation of an ensemble of spatially shifted low resolution pictures. They continued with deblurring process and the relative shifts of the input pictures are assumed to be known precisely. While Irani and Peleg (1991) describe a method based on the principle of reconstruction of a 2D object from its 1D projections in computer aided tomography. Image registration is carried out using method in Keren et al (1998) with an iterative super-resolution algorithm with minimum error between observed image and constructed image. Then, interdependence registration, interpolation and restoration have been investigated by Tom and Katsaggelos (1995). They posed expectation-maximation (EM) algorithm to solve the problem arose as a maximum likelihood (ML) estimation. The ML estimation problem solved the sub-pixel shifts, the noise variances of each image, and the high resolution image.
In remote sensing, super-resolution has been accessed on its capabilities to accommodate and support accuracy of classification (Andrew et al, 2001). They also identified land cover targets at the subpixel scale using super resolution technique to classified images with Hopfield neural networks. They found that super resolution cooperated with Hopfield neural network showed good performance and simplicity technique but it is dependent on nature constraints and needed prior information of imagery. Flack et al. (1994) also concentrated on super-resolution target identification at the borders of agricultural fields, where pixels of mixed class composition occur.
Then, Leavers (1993) shown edge detection and segmentation techniques were used to identify field boundaries and the Hough transform was applied to identify the straight, subpixel boundaries which enhanced the method of Flack. However, no validation or further work was carried out and this method remains unclear. Aplin et al. (1999) also made use subpixel scale vector boundary information, along with the fine resolution satellite sensor imagery to identify land cover targets by utilizing Ordnance Survey land line vector data and undertaking per-field rather than the traditional per-pixel land cover classification. In this study, target identification at a subpixel scale was demonstrated but accurate vector data sets with which to apply the approach will rarely be available.
4.0 HYPOTHESIS OF STUDY
The hypothesis of this study outlined in a topic base upon on workflow to construct super-resolution image and shoreline identification.
4.1. Image Registration
Essential to the successful super resolution algorithm is the need to find highly accurate point-to-point correspondence or registration between image input sequence. This problem had been addressed by Capel and Zisserman (2003), given two different views of the same scene, for each image point in one view find the image point in the second view which has the same pre-image, i.e. corresponds to the same actual point in the scene. Planar projective transformation or planar homography enable to estimate transformation of correspondence points (interest points) using geometric transform in 8 degrees of freedom. On the other hand, to locate these correspondence points feature-based registration needs to apply. In order to eliminate global illumination changes across the scene and intensity variations due to camera disturbances, photometric registration going to be applied.
4.2. Super-resolution image
Generally, the observed low resolution images are regarded as degraded observations of a real high resolution image. These are been implemented as shown in figure 2, where the stages in the super-resolution process described.
In this study, the objective is to generate high resolution image where to solve the inverse problem from the forward model. High resolution image from consecutive low resolution image then been applied with object identification methods which may construct with respect to image registration and super-resolution construction. All parameters are used iteratively and make object identification secured from error response and been processed in robustness, accurate and precision mode.
5.0 RESEARCH METHODOLOGY:
Methodologies will carry out with respect to hypothesis which may address some purpose methods of super-resolution construction for object identification.
5.1. Image Registration
Image registration consists of several methods to evaluate the resistances of geometric distortion, incorrect putative points and photometric differences.
i. Geometric Registration
Purpose of Figure 3 concentrates on the case of images which are related by a planar projective transformation or so-called planar homography. There are two different situations where (a) images of a plane viewed under arbitrary camera motion and (b) image of an arbitrary 3D scene viewed by a camera rotating about its optic centre and/or zooming .
Under a planar homography, points are map as: x' = Hx , where x' correspondence point of reference points x in other image and H is a 9 transformations projection. Three of these circumstances will investigate the registration of images based on transformation matrix approach below:
or equivalently; (1)
x' = Hx
The equivalent non-homogeneous relationship is
The last scenario depicts in which homography will occurs when a freely moving camera views a very distant scene, such case in high-aerial or satellite photography (Forte and Jones (1999).
6.0 A PRELIMINARY CHAPTER OF RESULTS
Shoreline information is important to navigation charting, marine boundary determination, and many coastal zone management activities, such as monitoring shoreline changes and delineating the inter-tidal zone, wetlands, and other coastal habitats. Satellite sensor imagery has proved its utility in all fields of earth science studies, including the study of coastal processes, because of the rapid, repetitive, synoptic and multispectral coverages of the satellites. However, mapping shoreline using coarse spatial resolution is difficult due to fact that actual shoreline could be located within the pixels. This chapter is a pilot study with the aims to determine shoreline positional errors using super resolution technique.
The methodology used in this chapter can be divided into several main sections. The study required a relatively coarse spatial resolution image data set of a coastal region for which the shore le location was known. NOAA AVHRR images with 1.1 km (Figure 4) spatial resolution was used as input image.