Digital colour cameras are dramatically falling in price, making them a ordable for ubiquitous appliances in many applications. Change in colour perception with changing light conditions induce errors that may escape a user's awareness. Colour constancy algorithms are based on inferring light properties (usually the white point) to correct colour. Other attempts using more data for colour correction such as (ICC based) colour management characterise a capturing device under given conditions through an input device pro le. This pro le can be applied to correct for deviating colour perception. But this pro le is only valid for the speci c conditions at the time of the characterisation, but fails with changes in light. This research presents a solution to the problem of long time observations with changes in the scene's illumination for common natural (overcast or clear, blue sky) and arti cial sources (incandescent or uorescent lamps). Colour measurements for colour based reasoning need to be represented in a robustly de ned way. One such suitable and well de ned description is given by the CIE LAB colour space, a device-independent, visually linearised colour description. Colour transformations using ICC pro le are also based on CIE colour descriptions. Therefore, also the corrective colour processing has been based on ICC based colour management. To verify the viability of CIE LAB based corrective colour processing colour constancy algorithms (White Patch Retinex and Grey World Assumption) have been modi ed to operate on L a b colour tuples. Results were compared visually and numerically (using colour indexing) against those using the same algorithms operating on RGB colour tuples. We can take advantage of the fact that we are dealing with image streams over time, adding another dimension usable for analysis. A solution to the problem of slowly changing light conditions in scenes with a static camera perspective is presented. It takes advantage of the small (frame-to-frame) changes in appearance of colour within the scene over time. Reoccurring objects or (background) areas of the scene are tracked to gather data points for an analysis. As a result, a suitable colour space distortion model has been devised through a rst order Taylor approximation (a ne transformation). By performing a multidimensional linear regression analysis on the tracked data points, parameterisations for the a ne transformations were derived. Finally, the device pro le is updated by amalgamating the corrections from the model into the ICC pro le for a single, comprehensive transformation. Following applications of the ICC based colour pro les are very fast and can be used in real-time with the camera's capturing frame rate (for current normal web cameras and low spec desktop computers). As light conditions usually change on a much slower time scale than the capturing rate of a camera, the computationally expensive pro le adaptation generally showed to be usable for many frames. The goal was to set out and nd a solution for consistent colour capturing using digital cameras, which is capable of coping with changing light conditions. Theoretical backgrounds and strategies for such a system have been devised and implemented successfully.

Digital colour cameras are dramatically falling in price, making them a ordable for ubiquitous appliances in many applications. Change in colour perception with changing light conditions induce errors that may escape a user's awareness. Colour constancy algorithms are based on inferring light properties (usually the white point) to correct colour. Other attempts using more data for colour correction such as (ICC based) colour management characterise a capturing device under given conditions through an input device pro le. This pro le can be applied to correct for deviating colour perception. But this pro le is only valid for the speci c conditions at the time of the characterisation, but fails with changes in light. This research presents a solution to the problem of long time observations with changes in the scene's illumination for common natural (overcast or clear, blue sky) and arti cial sources (incandescent or uorescent lamps). Colour measurements for colour based reasoning need to be represented in a robustly de ned way. One such suitable and well de ned description is given by the CIE LAB colour space, a device-independent, visually linearised colour description. Colour transformations using ICC pro le are also based on CIE colour descriptions. Therefore, also the corrective colour processing has been based on ICC based colour management. To verify the viability of CIE LAB based corrective colour processing colour constancy algorithms (White Patch Retinex and Grey World Assumption) have been modi ed to operate on L a b colour tuples. Results were compared visually and numerically (using colour indexing) against those using the same algorithms operating on RGB colour tuples. We can take advantage of the fact that we are dealing with image streams over time, adding another dimension usable for analysis. A solution to the problem of slowly changing light conditions in scenes with a static camera perspective is presented. It takes advantage of the small (frame-to-frame) changes in appearance of colour within the scene over time. Reoccurring objects or (background) areas of the scene are tracked to gather data points for an analysis. As a result, a suitable colour space distortion model has been devised through a rst order Taylor approximation (a ne transformation). By performing a multidimensional linear regression analysis on the tracked data points, parameterisations for the a ne transformations were derived. Finally, the device pro le is updated by amalgamating the corrections from the model into the ICC pro le for a single, comprehensive transformation. Following applications of the ICC based colour pro les are very fast and can be used in real-time with the camera's capturing frame rate (for current normal web cameras and low spec desktop computers). As light conditions usually change on a much slower time scale than the capturing rate of a camera, the computationally expensive pro le adaptation generally showed to be usable for many frames. The goal was to set out and nd a solution for consistent colour capturing using digital cameras, which is capable of coping with changing light conditions. Theoretical backgrounds and strategies for such a system have been devised and implemented successfully.

This thesis studies rapid object detection, focusing on feature-based methods. Firstly, modifications of training and detection of the Viola-Jones method are made to improve performance and overcome some of the current limitations such as rotation, occlusion and articulation. New classifiers produced by training and by converting existing classifiers are tested in face detection and hand detection. Secondly, the nature of invariant features in terms of the computational complexity, discrimination power and invariance to rotation and scaling are discussed. A new feature extraction method called Concentric Discs Moment Invariants (CDMI) is developed based on moment invariants and summed-area tables. The dimensionality of this set of features can be increased by using additional concentric discs, rather than using higher order moments. The CDMI set has useful properties, such as speed, rotation invariance, scaling invariance, and rapid contrast stretching can be easily implemented. The results of experiments with face detection shows a clear improvement in accuracy and performance of the CDMI method compared to the standard moment invariants method. Both the CDMI and its variant, using central moments from concentric squares, are used to assess the strength of the method applied to hand-written digits recognition. Finally, the parallelisation of the detection algorithm is discussed. A new model for the specific case of the Viola-Jones method is proposed and tested experimentally. This model takes advantage of the structure of classifiers and of the multi-resolution approach associated with the detection method. The model shows that high speedups can be achieved by broadcasting frames and carrying out the computation of one or more cascades in each node.

This thesis studies rapid object detection, focusing on feature-based methods. Firstly, modifications of training and detection of the Viola-Jones method are made to improve performance and overcome some of the current limitations such as rotation, occlusion and articulation. New classifiers produced by training and by converting existing classifiers are tested in face detection and hand detection. Secondly, the nature of invariant features in terms of the computational complexity, discrimination power and invariance to rotation and scaling are discussed. A new feature extraction method called Concentric Discs Moment Invariants (CDMI) is developed based on moment invariants and summed-area tables. The dimensionality of this set of features can be increased by using additional concentric discs, rather than using higher order moments. The CDMI set has useful properties, such as speed, rotation invariance, scaling invariance, and rapid contrast stretching can be easily implemented. The results of experiments with face detection shows a clear improvement in accuracy and performance of the CDMI method compared to the standard moment invariants method. Both the CDMI and its variant, using central moments from concentric squares, are used to assess the strength of the method applied to hand-written digits recognition. Finally, the parallelisation of the detection algorithm is discussed. A new model for the specific case of the Viola-Jones method is proposed and tested experimentally. This model takes advantage of the structure of classifiers and of the multi-resolution approach associated with the detection method. The model shows that high speedups can be achieved by broadcasting frames and carrying out the computation of one or more cascades in each node.

As computers have become more and more advanced, with even the most basic computer capable of tasks almost unimaginable only a decade ago, researchers and developers are focusing on improving the way that computers interact with people in their everyday lives. A core goal, therefore, is to develop a computer system which can understand and react appropriately to natural human behavior. A key requirement for such a system is the ability to automatically, and in real time, recognises human facial expressions. In addition, this must be successfully achieved regardless of the inherent differences in human faces or variations in lighting and other external conditions. The focus of this research was to develop such a system by evaluating and then utilizing the most appropriate of the many image processing techniques currently available, and, where appropriate, developing new methodologies and algorithms. The first key step in the system is to recognise a human face with acceptable levels of misses and false positives. This research analysed and evaluated a number of different face detection techniques, before developing a novel algorithm which combined phase congruency and template matching techniques. This novel algorithm provides key advantages over existing techniques because it can detect faces rotated to any angle, and it works in real time. Existing techniques could only recognise faces which were rotated less than 10 degrees (in either direction) and most could not work in real time due to excessive computational power requirements. The next step for the system is to enhance and extract the facial features. To successfully achieve the stated goal, the enhancement and extraction of the facial features must reduce the number of facial dimensions to ensure the system can operate in real time, as well as providing sufficient clear and detailed features to allow the facial expressions to be accurately recognised. This part of the system was successfully completed by developing a novel algorithm based on the existing Contrast Limited Adaptive Histogram Equalization technique which quickly and accurately represents facial features, and developing another novel algorithm which reduces the number of feature dimensions by combining radon transformation and fast Fourier transformation techniques, ensuring real time operation is possible. The final step for the system is to use the information provided by the first two steps to accurately recognise facial expressions. This is achieved using an SVM trained using a database including both real and computer generated facial images with various facial expressions. The system developed during this research can be utilised in a number of ways, and, most significantly, has the potential to revolutionise future interactions between humans and computers by assisting these reactions to become natural and intuitive. In addition, individual components of the system also have significant potential, with, for example, the algorithms which allow the recognition of an object regardless of its rotation under consideration as part of a project aiming to achieve non-invasive detection of early stage cancer cells.

This thesis describes the theoretical, experimental, and practical aspects of a noise-limited method for scene-change detection in images. The research is divided into three sections: noise analysis and modelling, dual illumination scene-change modelling, and integration of noise into the scene-change model. The sources of noise within commercially available digital cameras are described, with a new model for image noise derived for charge-coupled device (CCD) cameras. The model is validated experimentally through the development of techniques that allow the individual noise components to be measured from the analysis of output images alone. A generic model for complementary metal-oxide-semiconductor (CMOS) cameras is also derived. Methods for the analysis of spatial (inter-pixel) and temporal (intra-pixel) noise are developed. These are used subsequently to investigate the effects of environmental temperature on camera noise. Based on the cameras tested, the results show that the CCD camera noise response to variation in environmental temperature is complex whereas the CMOS camera response simply increases monotonically. A new concept for scene-change detection is proposed based upon a dual illumination concept where both direct and ambient illumination sources are present in an environment, such as that which occurs in natural outdoor scenes with direct sunlight and ambient skylight. The transition of pixel colour from the combined direct and ambient illuminants to the ambient illuminant only is modelled. A method for shadow-free scene-change is then developed that predicts a pixel's colour when the area in the scene is subjected to ambient illumination only, allowing pixel change to be distinguished as either being due to a cast shadow or due to a genuine change in the scene. Experiments on images captured in controlled lighting demonstrate 91% of scene-change and 83% of cast shadows are correctly determined from analysis of pixel colour change alone. A statistical method for detecting shadow-free scene-change is developed. This is achieved by bounding the dual illumination model by the confidence interval associated with the pixel's noise. Three benefits arise from the integration of noise into the scene-change detection method: - The necessity for pre-filtering images for noise is removed; - All empirical thresholds are removed; and - Performance is improved. The noise-limited scene-change detection algorithm correctly classifies 93% of scene-change and 87% of cast shadows from pixel colour change alone. When simple post-analysis size-filtering is applied both these figures increase to 95%.