This PDF file contains the front matter associated with SPIE Proceedings volume 7876, including Title page, Copyright information, Table of Contents, Introduction (if any), and Conference Committee listing.

Digital photographers continuously demand more performance from their equipment. Digital camera performance is defined by a set of parameters including dynamic range, noise, frame rate, resolution, and color. Amongst these parameters dynamic range is becoming increasingly more important. This is true because the human eye typically has a wider dynamic range than a digital camera. In this paper we define dynamic range as the ratio of the maximum to the minimum signal that can be detected. At the heart of all digital cameras is either a CCD or a CMOS image sensor (CIS). The dynamic range of the sensor typically limits the dynamic range of the camera. In this paper we review five CIS architectures that are designed to improved dynamic range. We start by reviewing standard CCD and CIS architectures and then present a simple sensor model. Using this model we show how signal to noise ratio (SNR) can be used to evaluate different wide dynamic range (WDR) sensor architectures. Then we sequentially review five different wide dynamic range techniques. The first WDR technique is multiple gains, and the second technique is non-linear pixel response. The third technique is variable exposure, and the forth technique is well capacity recycling. The fifth and final technique is time to saturation. For each of these techniques we present the pixel level circuitry and its advantages and disadvantages. Furthermore, all of these techniques are compared based on SNR and implementation complex. We discuss how implementation complexity affects signal processing in a digital camera, and other parameters in the sensor such as quantum efficiency and read noise. We conclude with a few summary comments.

The utility of Compressed Sensing (CS) for demosaicing of digital images have been explored by few recent efforts. Most recently, a Compressive Demosaicing [3] framework, based on employing a random panchromatic Color Filter Array (CFA) at the sensing stage, has provided compelling CS-based demosaicing results by visually outperforming other leading techniques. Meanwhile, it is well known that the Bayer pattern is arguably the most popular CFA used in low-cost consumer digital cameras. In this paper, we explore and compare the Bayer and random panchromatic CFA structures using a generic approach for demosaicing of images based on recent advances in the field of CS. In particular, a key objective of this work is to provide a comparative analysis between these two CFA patterns (Bayer and random panchromatic) under the general umbrella of sparse recovery, which represents the cornerstone of CS-based decoding. We demonstrate the viability of the Bayer pattern under certain CS conditions. Meanwhile, we show that a random panchromatic CFA, which meets certain incoherence constraints, can visually outperform a Bayer based sparse recovery. As illustrated in our simulation results, a panchromatic CFA is more consistent in terms of providing better visual quality when tested on a wide range of color images.

In motion invariant photography, blur is introduced by a structured movement of the camera during capture. The structured movement results in a uniform blur which simplifies deblur through post-processing for objects moving at different speeds in a single motion plane. However motion invariance depends on the camera speed exceeding the object speed (in the image plane) by a significant amount. This can lead to noisy image results and may be a problem for implementation of the method in practice. We propose the introduction of a time varying shutter transmittance to this recently proposed computational imaging method and demonstrate through simulation how this can improve both the degree of motion invariance and the reconstructed image quality, despite a reduction in optical efficiency. Improvements in the order 6dB are demonstrated for the reconstructed, deblurred images in the presence of moderate noise. The work has the potential to bring motion invariant photography closer to use in real camera product.

This paper presents an optimal image acquisition methodology by replacing the traditional birefringent filter with slight out-of-focus blur generated by the camera lens. Since many cameras already have adjustable lenses and auto-focus systems, our method can exploit existing hardware by simply changing the focusing strategy. During the image acquisition, the optimal defocus setting is automatically adapted to the power spectrum of the scene which is evaluated by a generic autocorrelation model. A criterion to estimate reconstruction errors without the baseband knowledge of the scene is developed in the paper. This metric helps the camera to choose the optimal focus settings. An optimal Wiener filter then recovers the captured scene and yields sharper images with reduced aliasing. The numerical and visual results show that our method is superior to current acquisition methods used by most digital cameras.

This paper presents a mobile phone imaging module with extended depth of focus (EDoF) by using axial irradiance equalization (AIE) phase coding. From radiation energy transfer along optical axis with constant irradiance, the focal depth enhancement solution is acquired. We introduce the axial irradiance equalization phase coding to design a two-element 2-megapixel mobile phone lens for trade off focus-like aberrations such as field curvature, astigmatism and longitudinal chromatic defocus. The design results produce modulation transfer functions (MTF) and phase transfer functions (PTF) with substantially similar characteristics at different field and defocus positions within Nyquist pass band. Besides, the measurement results are shown. Simultaneously, the design results and measurement results are compared. Next, for the EDoF mobile phone camera imaging system, we present a digital decoding design method and calculate a minimum mean square error (MMSE) filter. Then, the filter is applied to correct the substantially similar blur image. Last, the blur and de-blur images are demonstrated.

Extended depth of field (EDOF) cameras have recently emerged as a low-cost alternative to autofocus lenses. Different methods, either based on longitudinal chromatic aberrations or wavefront coding have been proposed and have reached the market. The purpose of this article is to study the theoretical performance and limitation of wavefront coding approaches. The idea of these methods is to introduce a phase element making a trade-off between sharpness at the optimal focus position and the variation of the blur spot with respect to the object distance. We will show that there are theoretical bounds to this trade-off: knowing the aperture and the minimal MTF value for a suitable image quality, the pixel pitch imposes the maximal depth of field. We analyze the limitation of the extension of the depth of field for pixel pitch from 1.75μm to 1.1μm, particularly in regards to the increasing influence of diffraction.

In this work the use of the Transverse Field Detector (TFD) as a device for multispectral image acquisition is proposed. The TFD is a color imaging pixel capable of color reconstruction without color filters. Its basic working principle is based on the generation of a suitable electric field configuration inside a Silicon depleted region by means of biasing voltages applied to surface contacts. With respect to previously proposed methods for performing multispectral capture, the TFD has a unique characteristic of electrically tunable spectral responses. This feature allows capturing an image with different sets of spectral responses (RGB, R'G'B', and so on) simply by tuning the device biasing voltages in multiple captures. In this way no hardware complexity (no external filter wheels or varying sources) is added with respect to a colorimetric device. The estimation of the spectral reflectance of the area imaged by a TFD pixel is based in this work on a linear combination of six eigenfunctions. It is shown that a spectral reconstruction can be obtained either (1) using two subsequent image captures that generate six TFD spectral responses or (2) using a new asymmetric biasing scheme, which allows the implementation of five spectral responses for each TFD pixel site in a single configuration, definitely allowing one-shot multispectral imaging.

Multispectral color imaging is a promising technology, which can solve many of the problems of traditional RGB color imaging. However, it still lacks widespread and general use because of its limitations. State of the art multispectral imaging systems need multiple shots making it not only slower but also incapable of capturing scenes in motion. Moreover, the systems are mostly costly and complex to operate. The purpose of the work described in this paper is to propose a one-shot six-channel multispectral color image acquisition system using a stereo camera or a pair of cameras in a stereoscopic configuration, and a pair of optical filters. The best pair of filters is selected from among readily available filters such that they modify the sensitivities of the two cameras in such a way that they get spread reasonably well throughout the visible spectrum and gives optimal reconstruction of spectral reflectance and/or color. As the cameras are in a stereoscopic configuration, the system is capable of acquiring 3D images as well, and stereo matching algorithms provide a solution to the image alignment problem. Thus the system can be used as a "two-in-one" multispectral-stereo system. However, this paper mainly focuses on the multispectral part. Both simulations and experiments have shown that the proposed system performs well spectrally and colorimetrically.

We present here a method that makes use of multispectral image data and generates a novel "photometric-invariant multispectral image" for this type of data. For RGB, an invariant image has been constructed independent of the colour and intensity of the illuminant and of shading. To generate this image either a set of calibration images is required, or entropy information taken from a single image can be used to develop the parameters necessary to produce the invariant. Nonetheless, generating an invariant image remains a complex and error-prone task for RGB image data. For multispectral images, we show that photometric-invariant image formation is in essence greatly simplified. One of the requirements for forming an invariant is the necessity of narrowband-sensor sensors. Here this is the case, and we show that with the simple knowledge of peak sensor wavelengths we can generate a high-D multispectral invariant. The PSNR is shown to be high between the respective invariant multispectral features for multispectral images taken under different illumination conditions, showing lighting invariance for a per-pixel measure; and the s-CIELAB error measure shows that the colour error between the 3-D colour images used to visualize the output invariant high-D data is also small.

High fidelity color image acquisition requires an accurate characterization of the camera's spectral sensitivity curves to perform color calibration or spectral estimation. Several methods have been proposed to perform this task; these include characterizations via test charts, narrowband filters and methods utilizing a monochromator. In most publications, RGB cameras are characterized. In this paper, we describe the characterization of the spectral sensitivity curves of a multispectral camera featuring seven optical bandpass filters. We show two different methods for the calibration using a monochromator - either by measuring the grayscale sensor of the camera and the filters separately or by characterizing the multispectral camera as a complete system. A comparison of both methods validates the measurement results. We furthermore develop different reconstruction methods (maximum value method, principal eigenvector method, linear or Wiener estimation). We perform also simulations of the characterization process to evaluate the methods and show the impact of the bandwidth of the monochromator stimuli on the reconstruction.

We present a study on the the applicability of hyperspectral images to evaluate color filter array (CFA) design and the performance of demosaicking algorithms. The aim is to simulate a typical digital still camera processing pipe-line and to compare two different scenarios: evaluate the performance of demosaicking algorithms applied to raw camera RGB values before color rendering to sRGB, and evaluate the performance of demosaicking algorithms applied on the final sRGB color rendered image. The second scenario is the most frequently used one in literature because CFA design and algorithms are usually tested on a set of existing images that are already rendered, such as the Kodak Photo CD set containing the well-known lighthouse image. We simulate the camera processing pipe-line with measured spectral sensitivity functions of a real camera. Modeling a Bayer CFA, we select three linear demosaicking techniques in order to perform the tests. The evaluation is done using CMSE, CPSNR, s-CIELAB and MSSIM metrics to compare demosaicking results. We find that the performance, and especially the difference between demosaicking algorithms, is indeed significant depending if the mosaicking/demosaicking is applied to camera raw values as opposed to already rendered sRGB images. We argue that evaluating the former gives a better indication how a CFA/demosaicking combination will work in practice, and that it is in the interest of the community to create a hyperspectral image dataset dedicated to that effect.

We propose here a strategy for the automatic annotation of outdoor photographs. Images are segmented in homogeneous regions which may be then assigned to seven different classes: sky, vegetation, snow, water, ground, street, and sand. These categories allows for content-aware image processing strategies. Our annotation strategy uses a normalized cut segmentation to identify the regions to be classified by a multi-class Support Vector Machine. The strategy has been evaluated on a set of images taken from the LabelMe dataset.

A fast inpainting algorithm for restoring damaged digitized photographs based on orientation driven multi structure morphology is proposed. The inpainting is guided by the orientation of the edges at the surrounding known regions of the missing domain. The edges are grouped with respect to their orientation by compass operators. Each edge group is represented as separate planes. The edge-plane information is used twice manifold for reconstructing the regions within the missing part, as well as for guiding the integration that follows. The damaged regions are morphologically eroded using the structuring elements of corresponding orientations dictated by the edge-planes. Simultaneously the edge planes are interpolated. The resultant image is obtained by integrating the eroded images with reference to the interpolated edge planes. The novelty of our approach is to explicitly specify the direction of filling herby ensuring ease in convergence in different orientations and then streamlining the process to guarantee complete and natural look. By implementing region-filling through morphological erosion, a non-iterative and non-sampling process makes the method faster than many traditional texture synthesis inpainting algorithms and successfully recovers images with better Peak Signal to Noise ratios even for massive damages.

In the early 1980's the future of conventional silver-halide photographic systems was of great concern due to the potential introduction of electronic imaging systems then typified by the Sony Mavica analog electronic camera. The focus was on the quality of film-based systems as expressed in the number of equivalent number pixels and bits-per-pixel, and how many pixels would be required to create an equivalent quality image from a digital camera. It was found that 35-mm frames, for ISO 100 color negative film, contained equivalent pixels of 12 microns for a total of 18 million pixels per frame (6 million pixels per layer) with about 6 bits of information per pixel; the introduction of new emulsion technology, tabular AgX grains, increased the value to 8 bit per pixel. Higher ISO speed films had larger equivalent pixels, fewer pixels per frame, but retained the 8 bits per pixel. Further work found that a high quality 3.5" x 5.25" print could be obtained from a three layer system containing 1300 x 1950 pixels per layer or about 7.6 million pixels in all. In short, it became clear that when a digital camera contained about 6 million pixels (in a single layer using a color filter array and appropriate image processing) that digital systems would challenge and replace conventional film-based system for the consumer market. By 2005 this became the reality. Since 2005 there has been a "pixel war" raging amongst digital camera makers. The question arises about just how many pixels are required and are all pixels equal? This paper will provide a practical look at how many pixels are needed for a good print based on the form factor of the sensor (sensor size) and the effective optical modulation transfer function (optical spread function) of the camera lens. Is it better to have 16 million, 5.7-micron pixels or 6 million 7.8-micron pixels? How does intrinsic (no electronic boost) ISO speed and exposure latitude vary with pixel size? A systematic review of these issues will be provided within the context of image quality and ISO speed models developed over the last 15 years.

In this paper, a high-speed CMOS image sensor having a new architecture and a new operating principle has been developed. The image sensor achieves both the continuous capturing and the burst capturing by a single chip, and has low power consumption, low heat generation, high sensitivity and high S/N ratio. This image sensor consist of mainly four blocks, two dimensional pixel array of 4-transister CMOS active pixel, analog memory arrays connected with each pixel output line independently to the pixel array, scanning circuits and multiple number of output amplifiers. A prototype image sensor was fabricated using a 0.18μm 2-Poly 3-Metal CMOS technology with the die size of 5550 μm^H x 4575 μm^V, the pixel size of 48 μm^H x 48 μm^V, the number of pixels of 72^H x 32^V, the number of analog memories of 104 memories per pixel and the 6 parallel horizontal output circuits and output amplifiers. The aperture ratio is 35% and the conversion gain is 60 μV/e-(input referred). It has been confirmed that this image sensor achieves 10,000,000 fps during burst capturing mode and 10,000 fps during the continuous capturing mode through the image capture experiments of high speed phenomena such as rotating object and discharge phenomenon.

We present a novel method for accurately measuring the optical transfer function (OTF) of a camera lens by digitally imaging a tartan test pattern containing sinusoidal functions with multiple frequencies and orientations. The tartan pattern can be tuned to optimize the measurement accuracy for an adjustable set of sparse spatial frequencies. The measurement method is designed to be accurate, reliable, and fast in a wide range of measurement conditions, including uncontrolled lighting. We describe the design of the tartan pattern and the algorithm for estimating the OTF accurately from a captured digital image. Simulation results show that the tartan method has significantly better accuracy for measuring the modulus of the OTF (the modulation transfer function, or MTF) than the ISO 12233 standard slanted-edge method, especially at high spatial frequencies. With 1% simulated imaging noise, the root mean square (RMS) error of the tartan method is on average 5 times smaller than the RMS error of the slanted-edge method. Experiments with a printed tartan chart show good agreement (0.05 RMS) with MTFs measured using the slanted-edge method and that, like the slanted-edge method, our method is tolerant to wide variations in illumination conditions.

Image sensor arrays may have defect pixels, either originating from manufacturing or being developed over the lifetime of the image sensor array. Continuous defect pixel detection and correction performing during camera runtime is desirable. On-the-fly detection and correction is challenging since edges and high-frequency image content might get identified as defect pixel regions and intact pixels become corrupted during defect pixel replacement. We propose a table-based detection and correction method which by and by fills the non-volatile table during normal camera operation. In this work we model defect pixels and pixel clusters to be stuck to fixed values or at least fixed to a narrow value range whereas the local neighborhood of these pixels indicate a normal behavior. The idea is to temporally observe the value ranges of small group of pixels (e.g. 4x4 pixel blocks) and to decide about their defective condition depending on their variability with respect to their neighbor pixels. Our method is computationally efficient, requires no frame buffer, requires modest memory, and therefore is appropriate to operate in line-buffer based image signal processing (ISP) systems. Our results indicate high reliability in terms of detection rates and robustness against high-frequency image content. As part of the defect pixel replacement system we also propose a simple and efficient defect pixel correction method based on the mean of medians operating on the Bayer CFA image domain.

This paper describes a novel image scaling algorithm based on a simple orientation adaptive anti-aliasing filter. The Projection-Slice theorem of Fourier theory is applied to characterize the structure of aliasing errors along edges in scaled images. This structure is used to explain the operation of our anti-aliasing filter which removes the dominant Fourier component of the aliasing error. It is shown that edge detail may be recovered from an aliased image by applying the anti-aliasing filter to the output of a separable wide-band interpolating filter. We also describe a simple one-dimensional inverse filter that may be combined with our anti-aliasing filter to restore further detail. Comparison of our scaling algorithm with other methods demonstrates that our scaling method is generally superior to other methods in removing visible aliasing artifacts along edges. Our method also competes favorably with state-of-the-art methods in recovering edge detail, without introducing sharpening artifacts exhibited by other methods.

Digital video stabilization allows to acquire video sequences without disturbing jerkiness by removing from the image sequence the effects caused by unwanted camera movements. One of the bottlenecks of these approaches is the local motion estimation step. In this paper we propose a Block Selector able to speed-up the block matching based video stabilization techniques without considerably degrading the stabilization performances. Both history and random criteria are taken into account in the selection process. Experiments on real cases confirm the effectiveness of the proposed approach even in critical conditions.

Image sensor crosstalk can be divided into spectral crosstalk and pixel crosstalk. This paper focuses on the pixel crosstalk and its effect on signal to noise ratio (SNR). Pixel crosstalk occurs in the spatial domain and is due to the signal leakage between adjacent pixels either by imperfect optical isolation or diffusion of electrons. This will have a negative impact on image quality mainly in two ways: spatial blurring and decreased SNR due to more aggressive color correction required. A method for modeling the spectral broadening due to the pixel crosstalk is used where a matrix is calculated from crosstalk kernels representing the spatial leakage between neighboring pixels. In order to quantify the amount of crosstalk we present a method in which ratios of integrals of the same color channel but within different wavelength intervals are calculated. This provides a metric that is more robust with respect to color channel scaling. To study the impact on SNR due to pixel crosstalk, a number of SNR metrics are compared to results from a limited psychophysical study. The studied SNR metrics are the metric used for calculating the SNR10 value in mobile imaging, the ISO 12232 noise metric and a metric where the signal is transformed into orthogonal color opponent channels, thereby enabling the analysis of the luminance noise separate from the chrominance noises. The results indicate that the ISO total noise and SNR10 metric yield very similar results and that the green channel has the largest individual impact on the crosstalk.

The image quality evaluation of CMOS sensors is a big challenge for camera module manufacturers. In this paper, we present an update of the Image Quality Evaluation Tool, a graphical user interface simulating image sensors to assess the performance of a pixel. The simulated images are computed from operating conditions and sensor's characteristics like Quantum Efficiency including off-axis effect. Simulation of QE off-axis impact has been based on characterization data. The method does not require optics, making it suitable for early design phases as for optimizations and investigations. Both measurement and implementation in the tool will be explained. The QE degradation with angle effect will be highlighted on simulated images. A uniform gray scene or coloured image simulation from QE off-axis measurement will help engineers to calculate post-processing digital correction like colour shading correction or colour correction matrix versus pixel position.

The research on image quality assessment (IQA) has been become a hot topic in most area concerning image processing. Seeking for the efficient IQA model with the neurophysiology support is naturally the goal people put the efforts to pursue. In this paper, we argue that comparing the edges position of reference and distorted image can well measure the image structural distortion and become an efficient IQA metric, while the edge is detected from the primitive structures of image convolving with LOG filters. The proposed metric is called NSER that has been designed following a simple logic based on the cosine distance of the primitive structures and two accessible improvements. Validation is taken by comparison of the well-known state-of-the-art IQA metrics: VIF, MS-SSIM, VSNR over the six IQA databases: LIVE, TID2008, MICT, IVC, A57, and CSIQ. Experiments show that NSER works stably across all the six databases and achieves the good performance.

High dynamic range (HDR) imaging seems to have developed to a level of soon being a standard feature in consumer cameras. This study was motivated by the need for evaluating tone mapping operators especially for consumer imaging applications. A no-reference method based on ISO 20462-2:2005 triplet comparison was created for evaluating tone mapping operators. Multiple HDR test images were photographed and the method was validated by evaluating 25 tone mapping operators with five test images. Tone mapping operators were evaluated based on image naturalness and pleasantness. The results indicate that the method successfully ranked the method in terms of naturalness and pleasantness. The test image set could be improved for example based on an imaging photo space for HDR photography. The test images of this study are available for non-commercial research purposes.

A well-known technique in high dynamic range (HDR) imaging is to take multiple photographs, each one with a different exposure time, and then combine them to produce an HDR image. Unless the scene is static and the camera position is fixed, this process creates the so-called "ghosting" artifacts. In order to handle non-static scenes or moving camera, images have to be spatially registered. This is a challenging problem because most optical flow estimation algorithm depends on the constant brightness assumption, which is obviously not the case in HDR imaging. In this paper, we present an algorithm to estimate the dense motion field in image sequences with photometric variations. In an alternating optimization scheme, the algorithm estimates both the dense motion field and the photometric mapping. As a latent information, the occluded regions are extracted and excluded from the photometric mapping estimation. We include experiments with both synthetic and real imagery to demonstrate the efficacy of the proposed algorithm. We show that the ghosting artifacts are reduced significantly in HDR imaging of non-static scenes.

The evaluation of sensor's performance in terms of signal-to-noise ratio (SNR) is a big challenge for both camera phone manufacturers and customers. The first ones want to predict and assess the performance of their pixel while the seconds require being able to benchmark raw sensors and processing pipes. The Reference SNR metric is very sensitive to crosstalk whereas for low-light issue, the weight of sensitivity should be increased. To evaluate noise on final image, the analytical calculation of SNR on luminance channel has been performed by taking into account noise correlation due to the processing pipe. However, this luminance noise does not match the perception of human eye which is also sensitive to chromatic noise. Alternative metrics have been investigated to find a visual noise metric closer to the human visual system. They have been computed on five pixel technologies nodes with different sensor resolutions and viewing conditions.

Computational imaging has been using for depth of field extension, distance estimation and depth map for stereo imaging and displaying with great successfully, which are realized by using special designed imaging lens and optimized image post-processing algorithm. Several special coding structures have been presented, like cubic, generalized cubic, logarithmic, exponential, polynomial, spherical and others. And different image post-processing algorithms like Wiener filter, SVD method, wavelet transform, minimum mean square error method and others are applied to achieve jointly-optimization. Although most of studies have shown excellent invariant of optical transfer function for imaging lens, but such invariance will be unsatisfied when manufacturing errors are considered. In this paper, we present a method to consider behavior of tolerance in computational imaging system from pure optical to optical - digital, which means lens and image post-processing are both included. An axial irradiance equalization phase coded imaging system is illustrated for tolerance sensitivity by using similarity of point spread function (PSF), Strehl ratio (SR), and root mean square error (RMSE) of restored images. Finally, we compare differences between presented method and Zemax.

We propose a new image deblurring method using differently exposed image pair. Regular-exposure image has more blur but less noise, while short-exposure image has more noise but less blur. Conventional approaches blend the two images using only good features of them based on the difference between the degradations. Although these approaches are effective under normal conditions, it is difficult to distinguish blur from noise under low light conditions. So we made two improvements to deal with large noise. One is using the gradient information of the regular-exposure image to refine the motion blur detection. The other is that noise on the edges is effectively suppressed considering the edge shapes and the noise levels on each pixel in the blended image. Finally, we implemented our method on the digital still camera and we successfully obtained the higher-quality images with less blur and noise through the simulations as well as the real camera examinations.

This article demonstrates that and how sensitometric and opto-electronic characteristic function (OECF) standards should be changed: The sensitivity S of all photosensitive arrays is--and, in standards, should be--determined by the midtone photosensitive exposure of the array Hmid, the binary logarithm of which is [log₂(H_max) + log₂(Hmin)]/2. The reference exposure H_o = S/H_sp of a photosensitive array is--and, in standards, should be--determined by the width of the photosensitive exposure range Δ, which is determined by the measured minimum and maximum usable photosensitive exposures, H_max and H_min. Knowledge of H_max and H_min (or H_mid and Δ) makes the reference exposure H_o, the speed point exposure H_sp and the safety factor σ unnecessary, and these extraneous quantities should be eliminated from the sensitometric standards. In addition, the sensitometric standard for solid-state arrays, ISO 12232-2006, and the OECF standard, ISO 14524-1999, should be changed because they use a photosensitive exposure range which is inappropriate for solid-state arrays. Finally, this article shows that all current standards establish midtone reflectances R_mid for standard photosensitive arrays that are much less than oft-touted 18% making most references to 18% inappropriate.

The paper provides a method of selectively controlling the strength of image noise reduction (NR) and sharpening in regions associated with specific colors that are visually impact on the human visual system. Since NR and sharpening operation can cause undesirable edge blurring or noise boosting in flat regions, the strength is adjusted according to the edge information. The proposed algorithm takes the following three stages: (i) color-based region classification, (ii) adaptive texture-based NR and sharpening, and (iii) H/W optimization for mobile image sensor compatibility. Experiments indicate that the proposed integrated algorithm platform provides significantly better results in terms of subjective visual quality.

We present CFA designs that faithfully capture images with specified luminance and chrominance bandwidths. Previous academic research has mostly been concerned with maximizing PSNR of reconstructed images without regard to chrominance bandwidth and cross-talk. Commercial systems, on the other hand, pay close attention to both these parameters as well as to the visual quality of reconstructed images. They commonly sacrifice resolution by using a sufficiently aggressive OLPF to achieve low cross-talk and artifact free images. In this paper, we present the so called Chrominance Bandwidth Ratio, r, model in an attempt to capture both the chrominance bandwidth and the cross-talk between the various signals. Next, we examine the effect of tuning photosite aspect ratio, a hitherto neglected design parameter, and show the benefit of setting it at a different value than the pixel aspect ratio of the display. We derive panchromatic CFA patterns that provably minimize the photo-site count for all values of r. An interesting outcome is a CFA design that captures full chrominance bandwidth, yet uses fewer photosites than the venerable color-stripe design. Another interesting outcome is a low cost practical CFA design that captures chrominance at half the resolution of luminance using only 4 unique filter colors, that lends itself to efficient linear demosaicking, and yet vastly outperforms the Bayer CFA with identical number of photosites demosaicked with state of the art compute-intensive nonlinear algorithms.

We have developed a robust color reproduction methodology by a simple calculation with a new color matrix using the formerly developed wide dynamic range WRGB lateral overflow integration capacitor (LOFIC) CMOS image sensor. The image sensor was fabricated through a 0.18 μm CMOS technology and has a 45 degrees oblique pixel array, the 4.2 μm effective pixel pitch and the W pixels. A W pixel was formed by replacing one of the two G pixels in the Bayer RGB color filter. The W pixel has a high sensitivity through the visible light waveband. An emerald green and yellow (EGY) signal is generated from the difference between the W signal and the sum of RGB signals. This EGY signal mainly includes emerald green and yellow lights. These colors are difficult to be reproduced accurately by the conventional simple linear matrix because their wave lengths are in the valleys of the spectral sensitivity characteristics of the RGB pixels. A new linear matrix based on the EGY-RGB signal was developed. Using this simple matrix, a highly accurate color processing with a large margin to the sensitivity fluctuation and noise has been achieved.

In the present article we focus on enhancing the contrast of images with low illumination that present large underexposed regions. For these particular images, when applying the standard contrast enhancement techniques, we also introduce noise over-enhancement within the darker regions. Even if both the contrast enhancement and denoising problems have been widely addressed within the literature, these two processing steps are, in general, independently considered in the processing pipeline. The goal of this work is to integrate contrast enhancement and denoise algorithms to proper enhance the above described type of images. The method has been applied to a proper database of underexposed images. Our results have been qualitatively compared before and after applying the proposed algorithm.

As with any sampling process, image acquisition is subject to spatial domain aliasing. Many designs of optical filters have been proposed that reduce the amount of aliasing in image acquisition. Still, the most common systems are based on birefringent optical materials. Unfortunately, these filters are expensive and the actual filter response curve can only partially be influenced. We propose a new refractive optical low-pass filter (ROLPF) based on refraction at a structured glass plate. The transparent plate features a single surface with a defined structure. This leads to refraction and will lead to a modified light distribution on the sensor. The plate is placed in front of the sensor and is moved in the image sensor plane during an exposure. This causes an averaging process and results in a spatial domain filtering. The response curve of this setup can be controlled by designing the gradient of the surface. This filter will be easy to manufacture, it offers a true low-pass characteristic which can be specifically designed and the bandwidth can be adjusted at runtime by variation of the distance to the sensor.

This paper presents a JPEG-like coder for image compression of single-sensor camera images using a Bayer Color Filter Array (CFA). The originality of the method is a joint scheme of compression.demosaicking in the DCT domain. In this method, the captured CFA raw data is first separated in four distinct components and then converted to YCbCr. A JPEG compression scheme is then applied. At the decoding level, the bitstream is decompressed until reaching the DCT coefficients. These latter are used for the interpolation stage. The obtained results are better than those obtained by the conventional JPEG in terms of CPSNR, ΔE2000 and SSIM. The obtained JPEG-like scheme is also less complex.

Objective Image Quality Assessment (IQA) model investigation is a hot topic in recent times. This paper proposed a novel and efficient universal Reduced Reference (RR) image quality assessment method based upon the statistics of edge discrimination. Firstly, binary edge maps created from the multi-scale wavelet transform modulus maxima were used as the low level feature to discriminate the difference between the reference and distorted image for IQA purpose. Then the gradient operator was applied on the binary map to produce the so called edge pattern map. The histogram of edge pattern map was used to verify the pattern of the edges of reference and distorted image, respectively. The RR features extracted from the histogram was used to discriminate the difference of edge pattern maps, and then form a new RR IQA model. Comparing to the typical RR model (Zhou Wang's method, 2005), only 12 features (96 bits) are needed instead of 18 features (162 bits) in Zhou Wang et al.'s method with better overall performance.

In this work, LED based flash solutions are evaluated for use in a camera phone application. The performance of a given flash solution is measured in terms of color accuracy and signal to noise ratio (SNR), both of which are standard test methods used in industry. Early in a camera phone design before completed camera modules are available, color accuracy and SNR are evaluated through a model which is based on knowledge of a given image sensor's color response as well as the power spectral distribution of the flash. Later in the design, when working camera modules become available, the evaluation is performed through direct measurement. These direct measurements are also used to verify the results of the aforementioned model. Color accuracy, SNR, how they are related, and the compromise between the two are discussed as well as the efficiency at which electrical power is converted to light that is detectable by the image sensor. Though many issues remain to be investigated, measuring color accuracy and SNR provides an evaluation method that builds on developed techniques and provides a practical foundation for flash evaluation as it applies to the camera phone industry.

Development of small pixels for high resolution image sensors implies a lot of challenges. A high level of performance should be guaranteed whereas the overall size must be reduced and so the degree of freedom in design and process. One key parameter of this constant improvement is the knowledge and the control of the crosstalk between pixels. In this paper, we present an advance in crosstalk characterization method based on the design of specific color patterns and the measurement of quantum efficiency. In a first part, we describe the color patterns designed to isolate one pixel or to simulate un-patterned colored pixels. These patterns have been implemented on test-chip and characterized. The second part deals with the characterization setup for quantum efficiency. Indeed, the use of spectral measurements allows us to discriminate pixels based on the color filter placed on top of them and to probe the crosstalk as a function of the depth in silicon, thanks to the photon absorption length variation with the wavelength. In the last part, results are presented showing the impact of color filters patterning, i.e. pixels in a Bayer pattern versus un-patterned pixels. The crosstalk directions and amplitudes are also analyzed in relation to pixel layout.