by Chung-fai Kwok. / Thesis (M.Sc.)--Chinese University of Hong Kong, 1994. / Includes bibliographical references (leaves 65). / Contents --- p.i / Table of Figures --- p.iii / Abstract --- p.v / Acknowledgment --- p.vii / Introduction --- p.viii / Chapter 1. --- Overview : Medical image network system --- p.1 / Chapter 1.1 --- MAGNET --- p.1 / Chapter 1.2 --- Medical image --- p.2 / Chapter 2. --- System configuration --- p.4 / Chapter 2.1 --- Hardware setting --- p.4 / Chapter 2.2 --- Software functions design --- p.5 / Chapter 3. --- Color handling --- p.7 / Chapter 3.1 --- Color --- p.7 / Chapter 3.2 --- Colormap and color display --- p.9 / Chapter 3.3 --- Static and dynamic color mapping --- p.10 / Chapter 4. --- Color image processing --- p.11 / Chapter 4.1 --- Color image quantization --- p.11 / Chapter 4.1.1 --- Pre-quantization --- p.13 / Chapter 4.1.2 --- Median cut Algorithm --- p.15 / Chapter 4.1.3 --- Remapping colors --- p.16 / Chapter 4.1.4 --- Hashing --- p.17 / Chapter 4.1.5 --- Distortion Measures --- p.21 / Chapter 4.1.6 --- Experiment results and Discussion --- p.22 / Chapter 4.2 --- Intensity mapping --- p.30 / Chapter 4.2.1 --- Graylevel image contrast enhancement and reduction --- p.30 / Chapter 4.2.2 --- Graylevel image brightness increment and reduction --- p.36 / Chapter 4.2.3 --- Contrast enhancement and reduction on color components --- p.40 / Chapter 4.2.4 --- Brightness increment and reduction on color components --- p.41 / Chapter 4.3 --- Pseudocoloring --- p.45 / Chapter 5. --- Color image analysis --- p.47 / Chapter 5.1 --- Region Measures --- p.47 / Chapter 5.1.1 --- Region measures function design --- p.47 / Chapter 5.1.2 --- Region growing mechanism --- p.48 / Chapter 5.1.3 --- Region smoothing --- p.49 / Chapter 5.2 --- Distance measures --- p.53 / Chapter 5.3 --- Statistical analysis --- p.53 / Chapter 6. --- Summary and future work --- p.57 / Appendix : User interfaces and functions --- p.58 / Bibliography --- p.65

Malignant melanoma is one of the most fatal forms of skin cancer. It has also become increasingly common, especially among white-skinned people exposed to the sun. Early detection of melanoma is essential to raise survival rates, since its detection at an early stage can be helpful and curable. Working out the dermoscopic clinical features (pigment network and lesion borders) of melanoma is a vital step for dermatologists, who require an accurate method of reaching the correct clinical diagnosis, and ensure the right area receives the correct treatment. These structures are considered one of the main keys that refer to melanoma or non-melanoma disease. However, determining these clinical features can be a time-consuming, subjective (even for trained clinicians) and challenging task for several reasons: lesions vary considerably in size and colour, low contrast between an affected area and the surrounding healthy skin, especially in early stages, and the presence of several elements such as hair, reflections, oils and air bubbles on almost all images. This thesis aims to provide an accurate, robust and reliable automated dermoscopy image analysis technique, to facilitate the early detection of malignant melanoma disease. In particular, four innovative methods are proposed for region segmentation and classification, including two for pigmented region segmentation, one for pigment network detection, and one for lesion classification. In terms of boundary delineation, four pre-processing operations, including Gabor filter, image sharpening, Sobel filter and image inpainting methods are integrated in the segmentation approach to delete unwanted objects (noise), and enhance the appearance of the lesion boundaries in the image. The lesion border segmentation is performed using two alternative approaches. The Fuzzy C-means and the Markov Random Field approaches detect the lesion boundary by repeating the labeling of pixels in all clusters, as a first method. Whereas, the Particle Swarm Optimization with the Markov Random Field method achieves greater accuracy for the same aim by combining them in the second method to perform a local search and reassign all image pixels to its cluster properly. With respect to the pigment network detection, the aforementioned pre-processing method is applied, in order to remove most of the hair while keeping the image information and increase the visibility of the pigment network structures. Therefore, a Gabor filter with connected component analysis are used to detect the pigment network lines, before several features are extracted and fed to the Artificial Neural Network as a classifier algorithm. In the lesion classification approach, the K-means is applied to the segmented lesion to separate it into homogeneous clusters, where important features are extracted; then, an Artificial Neural Network with Radial Basis Functions is trained by representative features to classify the given lesion as melanoma or not. The strong experimental results of the lesion border segmentation methods including Fuzzy C-means with Markov Random Field and the combination between the Particle Swarm Optimization and Markov Random Field, achieved an average accuracy of 94.00% , 94.74% respectively. Whereas, the lesion classification stage by using extracted features form pigment network structures and segmented lesions achieved an average accuracy of 90.1% , 95.97% respectively. The results for the entire experiment were obtained using a public database PH2 comprising 200 images. The results were then compared with existing methods in the literature, which have demonstrated that our proposed approach is accurate, robust, and efficient in the segmentation of the lesion boundary, in addition to its classification.

<p>This thesis is about improving the image quality of image sequences scanned by the film scanner GoldenEye. Film grain is often seen as an artistic effect in film sequences but scanned images can be more grainy or noisy than the intention. To remove the grain and noise as well as sharpen the images a few known image enhancement methods have been implemented, tested and evaluated. An own idea of a thresholding method using the dyadic wavelet transform has also been tested. As benchmark has MATLAB been used but one method has also been implemented in C/C++. Some of the methods works satisfactory when it comes to the image result but none of the methods works satisfactory when it comes to time consumption. To solve that a few speed up ideas are suggested in the end of the thesis. A method to correct the color of the sequences has also been suggested.</p>

<p>The medical imaging system Sectra PACS from Sectra Imtec contains a 3D mode that can be used for visualising image stacks from e.g. computed tomography. Various structures of human anatomy can be visualised in the 3D mode, but visualisations of the bronchial tree of the lungs rarely become good enough to be useful. The goal of this work was to investigate ways of improving such visualisations.</p><p>Various approaches were studied, evaluated and tested. The fact that most effort was needed for small structures with sizes similar to the resolution of the images made things slightly more complicated. A method classifying neighbourhoods based on local structure emerged as most promising, and was used as foundation for a proposed algorithm. It creates a mask representing the presence of bronchi, allowing the hiding of uninteresting structures in its proximity. The algorithm was then implemented so that it could be tested together with the existing system.</p><p>The method was found to work well and was able to detect the smaller tubes of the bronchial tree and output the desired classification mask. Its usefulness was somewhat reduced by issues relating to speed, and the fact that many computed tomography image stacks lack the necessary resolution for visualising the finer details of the bronchial tree.</p>

The medical imaging system Sectra PACS from Sectra Imtec contains a 3D mode that can be used for visualising image stacks from e.g. computed tomography. Various structures of human anatomy can be visualised in the 3D mode, but visualisations of the bronchial tree of the lungs rarely become good enough to be useful. The goal of this work was to investigate ways of improving such visualisations. Various approaches were studied, evaluated and tested. The fact that most effort was needed for small structures with sizes similar to the resolution of the images made things slightly more complicated. A method classifying neighbourhoods based on local structure emerged as most promising, and was used as foundation for a proposed algorithm. It creates a mask representing the presence of bronchi, allowing the hiding of uninteresting structures in its proximity. The algorithm was then implemented so that it could be tested together with the existing system. The method was found to work well and was able to detect the smaller tubes of the bronchial tree and output the desired classification mask. Its usefulness was somewhat reduced by issues relating to speed, and the fact that many computed tomography image stacks lack the necessary resolution for visualising the finer details of the bronchial tree.

NPR (Non-Photo-Realistic Rendering) is a new and quick-developed research topic in Image Processing. The main purpose of NPR is to generate pencil sketching¡Bwatercolor and oil painting, something different from photos, automatically by computer algorithms. On the other hand, there is another technique called PR (Photo-Realistic Rendering). The goal of PR is to generate real objects by computer algorithms, such as Matting or Inpainting. Furthermore, NPR includes two modes¡Gone is with physical model, researchers could write programs to simulate NPR by the properties of physical model. Without physical model, researchers could write programs to simulate NPR by their observation and deliberation. This thesis to the latter, there is no physical model in NPR. In the viewpoint of artists, drawing is performance of light and shadow. Then, in scientific, drawing depends on the degree of luminance. Luminance supports artists block and direction when drawing. Moreover, This thesis is mainly simulating oil painting with impressionist.

NPR (Non-Photo-Realistic Rendering) is a new and quick-developed research topic in Image Processing. The main purpose of NPR is to generate pencil sketching or watercolor, something different from photos, automatically by computer algorithms. On the other hand, there is another technique called PR (Photo-Realistic Rendering). The goal of PR is to generate real objects by computer algorithms, such as Matting. Furthermore, NPR includes two modes: one is with physical model and the other is not. With physical model, researchers could write programs to simulate NPR by the properties of physical model. Without physical model, researchers could write programs to simulate NPR by their observation and deliberation. This thesis belongs to the latter, NPR without physical model. In the viewpoint of artists, drawing is performance of light and shadow. Then, in scientific, drawing depends on the degree of luminance. Luminance supports artists block and direction when drawing. In this thesis, chapter 1 is introduction of art and previous researches. Chapter 2 describes theories what we can use and present the results. Chapter 3 describes the methods what this thesis use and necessary amendment, and present the results.

Στη παρούσα διπλωματική εργασία έχοντας σα βάση το γεγονός ότι η μαστογραφία αποτελεί την καλύτερη τεχνική έγκαιρης ανίχνευσης του καρκίνου του μαστού και ότι το μέγιστο όφελος της μαστογραφίας βρίσκεται στην πρώιμη ανίχνευση του καρκινώματος στα αρχικά στάδιά του, ασχοληθήκαμε με την ανάλυση και εφαρμογή μεθόδων ψηφιακής επεξεργασίας εικόνας, που εφαρμόζονται στη μαστογραφία. Ορίσαμε και αναλύσαμε τους διάφορους αλγόριθμους ενίσχυσης εικόνας, που στηρίζονται στην ενίσχυση αντίθεσης μέσω: (α) αύξησης των διαβαθμίσεων του γκρι σε συγκεκριμένες υπό περιοχές (π.χ. μέθοδος παραθύρου, εξίσωση ιστογράμματος) (μέθοδοι ενίσχυσης αντίθεσης) (β) αύξησης των διαφορών διαβάθμισης του γκρι της εικόνας, που αντιστοιχούν στην ενίσχυση της παρυφής της εικόνας, που συνήθως παρέχεται από φίλτρα που υλοποιούν τελεστές παραγώγου εικόνας (μέθοδοι ενίσχυσης αιχμών). Οι τεχνικές ενίσχυσης εικόνας που χρησιμοποιήσαμε στα πλαίσια της παρούσας Διπλωματικής Εργασίας είναι η Μέθοδος παραθύρου (Intesity Windowing), η Εξισορρόπηση Ιστογράμματος με ψαλίδισμα (Clipped Adaptive Histogram Equalization) και η Καταστολή θορύβου-Ενίσχυσης παρυφής με μέθοδο μετασχηματισμού κυματίου (Wavelet). Πριν την εφαρμογή των μεθόδων ενίσχυσης εικόνας στις μαστογραφικές εικόνες ήταν αναγκαία η δημιουργία ψηφιακών συνθετικών εικόνων οι οποίες μιμούνται τα παθολογικά ευρήματα και μέσω αυτών εξάγουμε τις βέλτιστες παραμέτρους που θα χρησιμοποιηθούν στις μεθόδους ενίσχυσης εικόνας. Αυτό το πετύχαμε κάνοντας χρήση των ποσοτικών δεικτών αξιολόγησης ποιότητας εικόνας. Οι δείκτες αυτοί στηρίχτηκαν σε μετρήσεις θορύβου και αντίθεσης για συγκεκριμένη περιοχή ενδιαφέροντος μαστογραφικής εικόνας και είναι ο Δείκτης Βελτίωσης Αντίθεσης (Contrast improvement index-CII), ο Δείκτης Ενίσχυσης Θορύβου (Noise amplification index- NAI) και ο λόγος CII/NAI (Contrast to noise ratio index -CNRI). Στη συνέχεια κάνοντας χρήση των βέλτιστων παραμέτρων ενίσχυσης εικόνας που βρήκαμε μέσω των ποσοτικών δεικτών, ενισχύσαμε το δείγμα των 105 μαστογραφικών εικόνων (περιοχής ενδιαφέροντος) και με τη βοήθεια ακτινολόγου προχωρήσαμε στο πειραματικό στάδιο της Διπλωματικής Εργασίας αξιολογώντας τις μαστογραφικές εικόνες με βάση το σύστημα BIRADS του ACR. Τελευταίο βήμα της μελέτης μας αποτέλεσε η διαγνωστική ακρίβεια του ακτινολόγου βασιζόμενη στη ROC καμπύλη. Υπολογίσαμε το εμβαδό κάτω από την καμπύλη ROC για κάθε μέθοδο ενίσχυσης ξεχωριστά αλλά και για τις αρχικές μαστογραφικές εικόνες στις οποίες είχαμε τις εκτιμήσεις από δύο ακτινολόγους. / In the present postgraduate thesis, we studied the application of image enhancement methods in X-ray mammography, the best technique for detection of breast cancer in early stages. We defined and analyzed the various algorithms of mammographic image enhancement, which are based in contrast enhancement: (a) Increasing the differences in grey level values in specific regions (Intensity Windowing, Histogram Equalization), called contrast enhancement methods. (b) Increasing the gradation of the differences of grey level values of regions that correspond in edge enhancement, provided by filters such applying first or second order derivatives, called edge enhancement methods). Three image enhancement methods were used in the present study, Manual Intensity Windowing (MIW), the Contrast-Limited Adaptive Histogram Equalization (CLAHE) and a dual-stage (denoising and enhancement) Adaptive Wavelet-based Enhancement method (WAVELET). Before applying the three image enhancement methods in mammographic images (regions of interests, ROIs), it was necessary to determine the optimal values of parameters that will be used in image enhancement methods. Digital synthetic images simulating pathological findings, such as microcalcification clusters, were developed for the determination of optimal values of parameters for CLAHE and WAVELET of the image enhancement methods. This was achieved by using quantitative metrics for image quality evaluation. These metrics were based on measurements of noise and contrast for a specific ROI in mammographic images containing the simulated microcalcification clusters. These metrics are: Contrast Improvement Index (CII), Noise Amplification Index (NAI) and the Contrast-to-Noise Ratio Index (CNRI). For the two image enhancement methods (CLAHE and WAVELET) the optimal values of parameters were used, while for MIW a radiologist manually determined the values of parameters (centre and level of windowing) for enhancing the sample of 105 mammographic images (ROIs) containing microcalcification clusters. Evaluation of the three image enhancement methods was performed based on BI-RADS of ACR for patient management determined by one radiologist. Last step of our study was to estimate the diagnostic accuracy of radiologist based on ROC analysis. We calculate the area under ROC curve for each enhancement method, as well as for the initial mammographic images (original).

Made available in DSpace on 2014-06-11T19:29:12Z (GMT). No. of bitstreams: 0 Previous issue date: 2004-07-01Bitstream added on 2014-06-13T18:07:38Z : No. of bitstreams: 1 bissoli_cf_me_sjc.pdf: 1658392 bytes, checksum: 87521fc116b48dbe7d8b1a4b112dee4e (MD5) / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / O propósito deste trabalho é analisar e verificar as medidas verticais de dentes humanos in vitro por meio de radiografias panorâmicas. O estudo foi baseado na montagem de uma boca com 28 dentes em um manequim de borracha, sendo utilizado para a realização das radiografias dois aparelhos ortopantomográficos (Rotograph plus e Panoura-10). O manequim, com o auxílio de suportes de madeira, foi radiografado com o plano oclusal paralelo ao horizontal, e também com seu plano oclusal à +5 graus e à -5 graus em relação ao plano horizontal. Os resultados obtidos indicaram que para a região de incisivos para o aparelho Rotograph plus a ampliação foi de 18,97% na maxila e 17,91% na mandíbula; para a região de pré-molares foi de 16,38% na maxila e 16,30% na mandíbula; e para a região de molares foi de 14,90% na maxila e 14,24% na mandíbula. Já para o aparelho Panoura-10 foi na região de incisivos 19,70% na maxila e 19,31% na mandíbula, na região de pré-molares 18,49% na maxila e 17,88% na mandíbula e para molares 16,66% na maxila e 16,17% na mandíbula. Conclui-se que houve ampliações diferentes nas regiões anatômicas estudadas. Posteriormente foi feito o teste estatístico ANOVA, e não houve diferenças estatisticamente significantes entre as inclinações de +5 graus e -5 graus em relação ao plano oclusal paralelo ao horizontal nos dois aparelhos (p>0,05). Também foi concluído que o aparelho Panoura-10 obteve 18,03% de ampliação vertical geral média e que o Rotograph plus obteve 16,45% de ampliação vertical geral média. / The aim of this work is to analyze and to verify the vertical measurements of the human teeth in vitro by panoramic radiographs. The study was based in set up of 28 teeth in rubber manikin. The radiographs were made with two orthopantomographs (Rotograph plus and Panoura - 10). The manikin with wood supports was radiographic with oclusal plane parallel to horizontal plane and oclusal plane 5 degrees positive and 5 degrees negative to horizontal plane. The results has showed that in incisors region, in Rotograph plus, has enlarged 18,97 % for the maxilla and 17,91% for the mandible; in bicuspid region, 16,38% for the maxilla and 16,30%for the mandible; and molar region, 14,90% for the maxilla and 14,24% for the mandible. The Panoura - 10 has enlarged in incisor region, 19,70% for the maxilla and 19,31% for the mandible, mandible, in bicuspid region, 18,49 % for the maxilla and 17,88 % for the mandible and molar region 16,66 % for the maxilla and 16,17 % for the mandible. The conclusions were that different enlargements in anatomic regions studied occurred. The ANOVA test has known that significance statistical differences between the inclinations of +5 degrees and -5 degrees and oclusal plane parallel the horizontal plane in both (p>0,05) didn t happen. The Panoura - 10 has enlarged 18,03% in total vertical average and the Rotograph plus has enlarged, 16,45 % in total vertical average.

The quality of an underwater image is degraded due to the effects of light scattering in water, which are resolution loss and contrast loss. Contrast loss is the main degradation problem in underwater images which is caused by the effect of optical back-scatter. A method is proposed to improve the contrast of an underwater image by mitigating the effect of optical back-scatter after image acquisition. The proposed method is based on the inverse model of an underwater image model, which is validated experimentally in this work. It suggests that the recovered image can be obtained by subtracting the intensity value due to the effect of optical back-scatter from the degraded image pixel and then scaling the remaining by a factor due to the effect of optical extinction. Three filters are proposed to estimate for optical back-scatter in a degraded image. Among these three filters, the performance of BS-CostFunc filter is the best. The physical model of the optical extinction indicates that the optical extinction can be calculated by knowing the level of optical back-scatter. Results from simulations with synthetic images and experiments with real constrained images in monochrome indicate that the maximum optical back-scatter estimation error is less than 5%. The proposed algorithm can significantly improve the contrast of a monochrome underwater image. Results of colour simulations with synthetic colour images and experiments with real constrained colour images indicate that the proposed method is applicable to colour images with colour fidelity. However, for colour images in wide spectral bands, such as RGB, the colour of the improved images is similar to the colour of that of the reference images. Yet, the improved images are darker than the reference images in terms of intensity. The darkness of the improved images is because of the effect of noise on the level of estimation errors.