Image representations are abstract descriptions of images. They refer to ideas describing or modelling images, in a particular way, for further processing and analysis. For a given image application and a given image, there are various image representations that can be employed. It is desired that the best image representation is used. The best image representation minimises storage size while preserving maximum image information. The thesis proposes a framework to evaluate the effectiveness of image representations. The effectiveness of image representations is measured by their storage size and preserved image information, with respect to a given image application. In consequence, the proposed framework allows us to seek the best image representation for a given image. In the proposed framework, different image representations for a given image are realised by their corresponding image transformations. This step generates different sets of coefficients for different image representations. Next, methods for ordering coefficients in each set are employed to sort coefficients from most-to-least significant. This step allows us to obtain the best subset of coefficients in each representation, by selecting the most significant coefficients until the required storage is met. Then, an appropriate quality measure is applied to select the best subset of coefficients. The best subset of coefficients corresponds to the best image representation, for a given image. This thesis also proposes a quality measure for image representations. It is based on the perceptual quality evaluation of image representations. Experimental results have shown that the proposed quality measure correlates well with evaluation by the human visual system. Finally, a validation of the proposed framework is carried out. Following the proposed framework, a method to obtain the best wavelet representation is introduced. Given a set of wavelets to represent an image, the proposed method enables us to attain the best wavelet representation, in terms of the highest perceptual image quality for a given storage size. The best wavelet image representation can be used to gain better performance for image coding, image enhancement and image comparison applications.
Facial feature extraction is one of the most important challenges in the area of facial image processing. This paper introduces a new method for locating eye features that is capable of processing images rapidly while achieving high detection rates. The proposed method is applicable to an n-dimensional space. Therefore, a new representation is considered for image, where an m×n image consists of m observation sets in an n-dimensional space. The main contribution to this paper is proposing a one-to-one linear transform based on this new representation called Linear Principal Transformation (LPT). LPT reduces the dimension of the image from n to two and allows extraction of all image features rapidly and efficiently. A set of experiments on the FERET and IFDB image data set is presented. The performance of eye feature extraction system is comparable to the best previous systems, where the success rate of the proposed method is 95.2%.
A number of image set compression algorithms have been proposed in the literature. A key component of these algorithms is a numeric measure used to quantify how similar two images are to each other from the point of view of a compression algorithm. Since most of these image set compression algorithms use wavelet-based image compression algorithms to compress the prediction error, we propose a number of related image prediction measures based on wavelet transforms. We also show some experimental results using the proposed measures. The proposed measure performs better than previous measures proposed in
We present the principle and the main steps of a new method for the visuo-spatial analysis of geometrical sketches recorded online. Visuo-spatial analysis is a necessary step for multi-level analysis. Multi-level analysis simultaneously allows classification, comparison or clustering of the constituent parts of a pattern according to their visuo-spatial properties, their procedural strategies, their structural or temporal parameters, or any combination of two or more of those parameters. The first results provided by this method concern the comparison of sketches to some perfect patterns of simple geometrical figures and the measure of dissimilarity between real sketches. The mean rates of good decision higher than 95% obtained are promising in both cases.
The accurate analysis of Coronary angiographic images plays an important role in diagnosing arterial diseases. All angiograms suffer from varying background, consisting of heart tissues, which harden accurate detection of vessels. Therefore removing background is one of the important steps which have essential effect on improving performance of subsequent steps of vessel detection algorithm, especially in the presence of noise. In this paper a new method for automatic precise detection of blood vessels in angiograms is proposed. Our method follows the steps firstly background removing, Secondly enhancement of vessel edges using fuzzy inference system and finally post processing the derived image for better visualization. Experiments on Angiograms show promising detection results.
Dermoscopy is one of the major imaging modalities used in the diagnosis of melanoma and other pigmented skin lesions. Due to the difficulty and subjectivity of human interpretation, automated analysis of dermoscopy images has become an important research area. Image segmentation is often the first step in this analysis. Numerous dermoscopy image segmentation methods have been proposed in the literature. Although, texture is one of the most significant features distinguishing a lesion from its surrounding skin, existing segmentation methods have solely relied on color information. By utilizing the G-means clustering algorithm that determines the number of clusters automatically, we propose a simple yet robust dermoscopy image segmentation method based on texture features. Preliminary experiments show that we can obtain good segmentation results without considering color information and texture information alone is sufficient to distinguish a lesion from its surrounding skin.
Faces are among the most important classes of objects, computers have to deal with. Automatic processing of facial images has attracted considerable attention in the last decades and many algorithms are proposed for that. Without doubt the facial image databases are the most important means which are used to test these algorithms. Despite of various databases provided, less attention is being paid to develop a benchmark for creating a face image database. A comparison among more than 40 available face image databases is carried out to extract their intrinsic and extrinsic features and to propose a benchmark. This paper introduces an efficient and innovative portable design for gathering face image data. It covers many photography conditions such as different poses, backgrounds, and illuminations. A set of 178 images from 20 individual is gathered and a comparative evaluation is carried out to prove the efficiency of developed benchmark and database design.
