With the growing demand of embedded graphics applications, how to provide an efficient graphics hardware acceleration solution has drawn much attention. It is well known that computer graphics contains two major domains: two-dimensional (2D) and three-dimensional (3D) graphics. Each domain owns large amounts of applications, such that general embedded platforms will require both graphics acceleration supports. This thesis proposes an advanced texture unit architecture which can provide various 3D texture filtering functions including trilinear, anistrophics filtering etc , and 2D coloring, painting, and texturing functions. Our proposed design consists of a core computation unit, and a set of data registers. The equations for those supported functions are decomposed into a series of basic arithmetic operations such as multiply-add-accumulation, multiply, etc executed by the core computation unit. To evaluate those equations for each pixel may require some pre-computed parameters which will be computed outside our unit in advance by the system¡¦s micro-controller. The equations can be computed by our texture unit based on the selected finite-state machine sequences which is stored in the on-chip control table. By updating those sequences can change the functionality provided by our chip. The overall cost of the proposed unit is about 28.36k gates. In addition to various texturing functions, this thesis also proposes an implementation of texture function for high-dynamic range (HDR) textures. HDR textures can provide various color details according to the frame¡¦s global illumination environment. Therefore, the 3D rendering system has to incorporate a tone-mapping mechanism to map the HDR image into normal color range of output display system. To reduce the overall tone-mapping implementation cost, this thesis uses an extra accumulator between the standard per-fragment rendering pipeline stages to accumulate the global illumination intensity based on the depth comparison result of the incoming pixel. After all of the pixels have passed through the pipeline stages, every pixel of the stored rendering result will be fetched into a mapping unit which will generate its mapping color in the normal dynamic range. The overall cost of the additional hardware for the realization of HDR textures is about 6.98k gates.

A low voltage low power sigma¡Vdelta modulator for voice applications is presented. The implementation of proposed sigma-delta modulator is based on switched-capacitor circuit. Bootstrapped switches were used to replace CMOS transmission gates for increasing the insufficient driving of switched-capacitor circuit under the low voltage operation. To reduce the power dissipation, an improved current mirror OTA were designed with rail-to-rail output swing, which can also make the voltage gain enhance 10~20 dB and overcome the poor voltage gain shortage of traditional current mirror OTA. The post-simulation result shows that the modulator achieves a dynamic range of 77 dB, a peak signal-to-noise ratio of 82 dB, and the sigma-delta modulator dissipates 25£gW under 1.1-V voltage supply, using TSMC 0.18£gm 1P6M CMOS technology.

Today, infrared cameras are used especially for target tracking and surveillance operations. However, they have a high dynamic range output, and the standard display devices cannot handle them. In order to show them on common devices, the dynamic range is cropped. Thus, the contrast of the image is reduced. This is called as the High Dynamic Range (HDR) Compression. Although several algorithms have been proposed for preserving details during the HDR compression process, it cannot be used to enhance the local contrasts of image contents. In this thesis, we compare the performances of contrast enhancement techniques, which are suitable for real time applications. The methods experimented are generally histogram based methods. Some modifications are also proposed in order to reduce computational complexity of the process. Performances of these methods are compared with common objective quality metrics on different image sets.

Hard disk drive applications require a high Spurious Free Dynamic Range (SFDR), 6-bit Analog-to-Digital Converter (ADC) at conversion rates of 1GHz and beyond. This work proposes a robust, fault-tolerant scheme to achieve high SFDR in an av- eraging flash A/D converter using comparator chopping. Chopping of comparators in a flash A/D converter was never previously implemented due to lack of feasibility in implementing multiple, uncorrelated, high speed random number generators. This work proposes a novel array of uncorrelated truly binary random number generators working at 1GHz to chop all comparators. Chopping randomizes the residual offset left after averaging, further pushing the dynamic range of the converter. This enables higher accuracy and lower bit-error rate for high speed disk-drive read channels. Power consumption and area are reduced because of the relaxed design requirements for the same linearity. The technique has been verified in Matlab simulations for a 6-bit 1Gsamples/s flash ADC under case of process gradients with non-zero mean offsets as high as 60mV and potentially serious spot offset errors as high as 1V for a 2V peak to peak input signal. The proposed technique exhibits an improvement of over 15dB compared to pure averaging flash converters for all cases. The circuit-level simulation results, for a 1V peak to peak input signal, demon- strate superior performance. The reported ADC was fabricated in TSMC 0.18 ??mCMOS process. It occupies 8.79mm2 and consumes about 400mW from 1.8V power supply at 1GHz. The targeted SFDR performance for the fabricated chip is at least 45dB for a 256MHz input sine wave, sampled at 1GHz, about 10dB improvement on the 6-bit flash ADCs in the literature.

Typical consumer digital cameras capture the scene by generating a mosaic-like grayscale image, known as a color filter array (CFA) image. One obvious challenge in digital photography is the storage of image, which requires the development of an efficient compression solution. This issue has become more significant due to a growing demand for high dynamic range (HDR) imaging technology, which requires increased bandwidth to allow realistic presentation of visual scene. This thesis proposes two digital camera pipelines, efficiently encoding CFA image data represented in HDR format. Firstly, a lossless compression scheme exploiting a predictive coding followed by a JPEG XR encoding module is introduced. It achieves efficient data reduction without loss of quality. Secondly, a lossy compression scheme that consists of a series of processing operations and a JPEG XR encoding module is introduced. Performance evaluation indicates that the proposed method delivers high quality images at low computational costs.

Typical consumer digital cameras capture the scene by generating a mosaic-like grayscale image, known as a color filter array (CFA) image. One obvious challenge in digital photography is the storage of image, which requires the development of an efficient compression solution. This issue has become more significant due to a growing demand for high dynamic range (HDR) imaging technology, which requires increased bandwidth to allow realistic presentation of visual scene. This thesis proposes two digital camera pipelines, efficiently encoding CFA image data represented in HDR format. Firstly, a lossless compression scheme exploiting a predictive coding followed by a JPEG XR encoding module is introduced. It achieves efficient data reduction without loss of quality. Secondly, a lossy compression scheme that consists of a series of processing operations and a JPEG XR encoding module is introduced. Performance evaluation indicates that the proposed method delivers high quality images at low computational costs.

The performance, reliability, and ubiquity of automatic speech recognition systems has flourished in recent years due to steadily increasing computational power and technological innovations such as hidden Markov models, weighted finite-state transducers, and deep learning methods. One problem which plagues speech recognition systems, especially those that operate offline and have been trained on specific in-domain data, is the deleterious effect of noise on the accuracy of speech recognition. Historically, robust speech recognition research has focused on traditional noise types such as additive noise, linear filtering, and reverberation. This thesis describes the effects of nonlinear dynamic range compression on automatic speech recognition and develops a number of novel techniques for characterizing and counteracting it. Dynamic range compression is any function which reduces the dynamic range of an input signal. Dynamic range compression is a widely-used tool in audio engineering and is almost always a component of a practical telecommunications system. Despite its ubiquity, this thesis is the first work to comprehensively study and address the effect of dynamic range compression on speech recognition. More specifically, this thesis treats the problem of dynamic range compression in three ways: (1) blind amplitude normalization methods, which counteract dynamic range compression when its parameter values allow the function to be mathematically inverted, (2) blind amplitude reconstruction techniques, i.e., declipping, which attempt to reconstruct clipped segments of the speech signal that are lost through non-invertible dynamic range compression, and (3) matched-training techniques, which attempt to select the pre-trained acoustic model with the closest set of compression parameters. All three of these methods rely on robust estimation of the dynamic range compression distortion parameters. Novel algorithms for the blind prediction of these parameters are also introduced. The algorithms' quality is evaluated in terms of the degree to which they decrease speech recognition word error rate, as well as in terms of the degree to which they increase a given speech signal's signal-to-noise ratio. In all evaluations, the possibility of independent additive noise following the application of dynamic range compression is assumed.

Zugl.: Koblenz. Landau, Diss., 2007

Ο τομέας της ψηφιακής φωτογράφισης εξελίσσεται διαρκώς. Το γεγονός αυτό, καθιστά τη ψηφιακή φωτογραφία χρήσιμο εργαλείο για διάφορες επιστημονικές και όχι μόνο εφαρμογές. Για αυτό τον λόγο, γίνεται απαραίτητη η λεπτομερέστερη αποτύπωση της πληροφορίας. Η λεπτομερέστερη αποτύπωση της πληροφορίας αποτελεί αντικείμενο της εικονοποίησης υψηλής δυναμικής περιοχής(High Dynamic Range Imaging). Ο σκοπός της παρούσας διπλωματικής εργασίας είναι η μελέτη της εικονοποίησης υψηλής δυναμικής περιοχής και η υλοποίηση ενός αλγορίθμου, κάνοντας χρήση πολλαπλών λήψεων για την παραγωγή τελικής εικόνας με διευρυμένη δυναμική περιοχή. / The field of digital photography is constantly evolving. In fact, this makes digital photography a useful tool in various scientific applications. For this reason, a more detailed presentation of information is needed. The detailed presentation of information is the subject of high dynamic range (High Dynamic Range Imaging). The purpose of this thesis is to study the field of high dynamic range imaging and the implementation of an algorithm, making use of multiple shots to produce a final image with extended dynamic range.

abstract: This study examines the applicability of high dynamic range (HDR) imagery as a diagnostic tool for studying lighting quality in interior environments. It originates from the limitations in lighting quality assessments, particularly from the problematic nature of measuring luminance contrast--a significant lighting quality definer. In this research, HDR imaging method is studied systematically and in detail via extensive camera calibration tests considering the effect of lens and light source geometry (i.e. vignetting, point spread and modulation transfer functions), in-camera variables (i.e. spectral response, sensor sensitivity, metering mode,), and environmental variables (i.e. ambient light level, surface color and reflectance, light source spectral power distribution) on the accuracy of HDR-image-derived luminance data. The calibration test findings are used to create camera setup and calibration guidelines for future research, especially to help minimize errors in image extracted lighting data. The findings are also utilized to demonstrate the viability of the tool in a real world setting--an office environment combining vertical and horizontal tasks. Via the quasi-experimental setup, the relationship between line of sight and perceived luminance contrast ratios are studied using HDR images. Future research can benefit from the calibration guidelines to minimize HDR-based luminance estimation errors. The proposed tool can be used and tested in different contexts and tasks with varying user groups for revising the former luminance-contrast guidelines as well as surface reflectance recommendations. / Dissertation/Thesis / Ph.D. Architecture 2011