CMOS Image Sensors (CIS) are integral to both human and computer vision tasks, necessitating continuous improvements in key performance metrics such as latency, power, and noise. Despite experienced designers being able to make informed design decisions, novice designers and system architects face challenges due to the complex and expansive design space of CIS. This paper introduces a systematic methodology that elucidates the trade-offs among CIS performance metrics and enables efficient design space exploration. Specifically, we propose a first-principle-based CIS modeling method. By exposing low-level circuit parameters, our modeling method explicitly reveals the impacts of design changes on high-level metrics. Based on the modeling method, we propose a design space exploration process that swiftly evaluates and identifies the optimal CIS design, capable of exploring over 109 designs in under a minute without the need for time-consuming SPICE simulations. Our approach is validated through a case study and comparisons with real-world designs, demonstrating its practical utility in guiding early-stage CIS design.

Though considerable research has been carried out into capacitive touch sensing, the geometrical layout of sensors and its effects on accuracy and noise sensitivity have not been explored. This paper examines the sensor patterns and interpolation algorithms used in current touch screens. Its main aim is to evaluate sensor patterns using image processing techniques, thereby providing a methodology for objective comparison that lends itself to easily testing new designs. The paper presents results comparing the accuracy and reliability of three sensor patterns from the patent literature in combination with four well-known interpolation algorithms.

We show that bilateral symmetry plane estimation for three-dimensional (3-D) shapes may be carried out accurately, and efficiently, in the spherical harmonic domain. Our methods are valuable for applications where spherical harmonic expansion is already employed, such as 3-D shape registration, morphometry, and retrieval. We show that the presence of bilateral symmetry in the 3-D shape is equivalent to a linear phase structure in the corresponding spherical harmonic coefficients, and provide algorithms for estimating the orientation of the symmetry plane. The benefit of using spherical harmonic phase is that symmetry estimation reduces to matching a compact set of descriptors, without the need to solve a correspondence problem. Our methods work on point clouds as well as large-scale mesh models of 3-D shapes.

Propagating similarity information along the data manifold requires careful selection of local neighborhood. Selecting a "good" neighborhood in an unsupervised setting, given an affinity graph, has been a difficult task. The most common way to select a local neighborhood has been to use the k-nearest neighborhood (k-NN) selection criterion. However, it has the tendency to include noisy edges. In this paper, we propose a way to select a robust neighborhood using the consensus of multiple rounds of k-NNs. We explain how using consensus information can give better control over neighborhood selection. We also explain in detail the problems with another recently proposed neighborhood selection criteria, i.e., Dominant Neighbors, and show that our method is immune to those problems. Finally, we show the results from experiments in which we compare our method to other neighborhood selection approaches. The results corroborate our claims that consensus of k-NNs does indeed help in selecting more robust and stable localities.

Shape is a distinguishing feature of objects. But when objects have similar global shapes, discriminating them becomes difficult. This calls for the identification of important parts of the shapes. In this paper, we introduce the concept of discriminative parts and propose a method to identify the same. We show how levels of importance can be assigned to different regions of contours, based on their discriminative capability. Our experiments show that the method is promising and can identify semantically meaningful segments as being important segments.

QR codes are widely used as a means of conveying textual information, such as emails, hyperlinks, or phone numbers, through images that are interpreted using a smartphone camera. The codes take up valuable space in print media. The random appearance of QR codes not only detracts from the production values of the advertisement in which they appear, but the codes are also visually insignificant in the sense that a human cannot discern the vendor, brand, or purpose of the code just by looking at it, without the aid of scanning software. Though neither the aesthetics nor the visual significance of the code matter for scanning purposes, they do matter for advertising layout and, more importantly, can provide valuable brand distinction. In this paper, we show how the visually significant QR codes may be obtained by image blending. Unlike various ad-hoc methods that have been proposed by others, our method leaves completely intact the error correction budget of the code. Our method allows images as diverse as corporate logos and family photographs to be embedded in the code in full color. We provide a detailed statistical analysis of the method to show the effect of blending on error rates in noisy environments.

Directional chamfer matching (DCM) has shown good results in many areas such as object recognition and pose estimation. Currently DCM has been applied only for two-dimensional (2-D) matching. In this letter, we present a DCM scheme that utilizes depth in addition to 2-D input, which we refer to as 2.5D DCM. We show that in situations such as 3-D model-based pose estimation, depth information can be exploited to achieve robust performance. We apply the proposed method for human motion capture (HMC), using the Human Eva I dataset. We compare our approach with alternative methods used for HMC. Our results show that using depth information makes traditional DCM robust. Furthermore, the proposed method outperforms the alternatives used for HMC in state of the art systems.

This paper describes how phase-sensitive rotation invariants for three-dimensional data may be obtained. A “bispectrum” is formulated for rotations, and its properties are derived for spherical harmonic coefficients as well as for moments. The bispectral invariants offer improved discrimination over previously published magnitude-only invariants. They are able to distinguish rotations from reflections, as well as rotations of an entire shape from component-wise rotations of elements of the shape. As experiments show, they provide robust performance for both surface and voxel data.

The algebraic theory of finite groups appears in signal processing problems involving the statistical analysis of ranked data and the construction of invariants for pattern recognition. Standard signal processing techniques involving spectral analysis are, in theory, possible for data defined on finite groups by using the Fourier transform provided by group representations. However, one such technique, the bispectrum, which is useful for analysing non-Gaussian data as well as for constructing geometric invariants, has not been explored in detail for finite groups. This paper shows how to construct the bispectrum on an arbitrary finite group or homogeneous space and explores its properties. Examples are given using the symmetric group as well as wreath-product groups.

This paper proposes a multiresolution form of the singular value decomposition (SVD) and shows how it may be used for signal analysis and approximation. It is well-known that the SVD has optimal decorrelation and subrank approximation properties. The multiresolution form of SVD proposed here retains those properties, and moreover, has linear computational complexity. By using the multiresolution SVD, the following important characteristics of a signal may he measured, at each of several levels of resolution: isotropy, sphericity of principal components, self-similarity under scaling, and resolution of mean-squared error into meaningful components. Theoretical calculations are provided for simple statistical models to show what might be expected. Results are provided with real images to show the usefulness of the SVD decomposition.

This paper is concerned with the properties of estimators in O(n,p),the n/spl times/p orthogonal matrices. It is shown that it is natural to introduce the notion of a parallel estimator where the expected value of the estimator must lie in normal space (orthogonal complement of tangent space) of O(n,p) at the true value. An appropriate measure of variance, referred to as divergence, is introduced for a parallel estimator and a Cramer-Rao (CR) type bound is then established for the divergence. The well-known Fisher-von Mises matrix distribution is often used to model random behavior on O(n,p) and depends on parameters /spl Theta//spl isin/O(n,p) and H a p/spl times/p symmetric matrix. The bound for this distribution is calculated for the case n=p=3 and the divergence of the maximum-likelihood estimator (MLE) of /spl Theta/ is estimated by simulation. The bound is shown to be tight over a wide range of H.

A shape with a twice-differentiable boundary is convex if and only if the boundary has nonnegative curvature everywhere. We show how to formulate this condition equivalently in terms of the Fourier descriptors of the boundary: The shape is convex if and only if the boundary has a nonnegative definite "parametric" curvature spectrum (defined herein).

A procedure for estimating the parameters associated with a linear phase signal is developed. When the data being modelled is composed of a linear phase signal corrupted by additive Gaussian noise the approach taken results in maximum-likelihood estimates of the linear phase parameters. These estimates are useful for detecting and estimating the presence of symmetry in both one and two dimensions. The effectiveness of the estimates is tested on both synthetic and real images.

Edge localization occurs when an edge detector determines the location of an edge in an image. The authors use statistical parameter estimation techniques to derive bounds on achievable accuracy in edge localization. These bounds, known as the Cramer-Rao bounds, reveal the effect on localization of factors such as signal-to-noise ratio (SNR), extent of edge observed, scale of smoothing filter, and a priori uncertainty about edge intensity. By using continuous values for both image coordinates and intensity, the authors focus on the effect of these factors prior to sampling and quantization. They also analyze the Canny algorithm and show that for high SNR, its mean squared error is only a factor of two higher than the lower limit established by the Cramer-Rao bound. Although this is very good, the authors show that for high SNR, the maximum-likelihood estimator, which is also derived, virtually achieves the lower bound.>

Edge localization occurs when an edge detection algorithm is able to accurately determine the location of an edge in an image. Edge localization is formulated as a parameter estimation problem, and the authors derive a Cramer-Rao bound on achievable accuracy as measured by mean squared error. The bound reveals the effect on localization of factors such as signal to noise ratio, observation window size, scale of smoothing filter, and a priori uncertainty about edge intensity. The authors analyze the Canny (1986) algorithm and show that the variance of its localization error is only a factor of two higher than the lower limit established by the Cramer-Rao bound. Although this is very good, it is shown that the maximum-likelihood estimator, which is derived in this work, virtually achieves the lower bound.>