Function Of Image Resolution Research Articles

Uncovering the fractal nature of water vapor distribution above the surface of the earth

Fractals have been at the heart of geophysical and geospatial studies in the recent past. We explore the fractal character of water vapor distributions above the surface of the Earth as a function of both image resolution (number of pixels) and moisture content percentile and study the emergent scaling properties. The percolation phenomena associated with the vapor distribution clusters has been studied for both the cases. For our analysis, computational methods and algorithms have been used to derive the physically relevant quantities such as fractal dimension, number of clusters, and size of the largest cluster. The box counting method, used to calculate the fractal dimension, unravels a potential multi-fractal character of the data. We also test the applicability of Korcak’s law on our system and determine the quality of the fit using the Kolmogorov–Smirnov statistic. We show that the fractal character of the distribution is exact as a function of image resolution and approximate in some regimes as a function of the vapor percentile.

Impact of Image Resolution on Deep Learning Performance in Endoscopy Image Classification: An Experimental Study Using a Large Dataset of Endoscopic Images.

Recent trials have evaluated the efficacy of deep convolutional neural network (CNN)-based AI systems to improve lesion detection and characterization in endoscopy. Impressive results are achieved, but many medical studies use a very small image resolution to save computing resources at the cost of losing details. Today, no conventions between resolution and performance exist, and monitoring the performance of various CNN architectures as a function of image resolution provides insights into how subtleties of different lesions on endoscopy affect performance. This can help set standards for image or video characteristics for future CNN-based models in gastrointestinal (GI) endoscopy. This study examines the performance of CNNs on the HyperKvasir dataset, consisting of 10,662 images from 23 different findings. We evaluate two CNN models for endoscopic image classification under quality distortions with image resolutions ranging from 32 × 32 to 512 × 512 pixels. The performance is evaluated using two-fold cross-validation and F1-score, maximum Matthews correlation coefficient (MCC), precision, and sensitivity as metrics. Increased performance was observed with higher image resolution for all findings in the dataset. MCC was achieved at image resolutions between 512 × 512 pixels for classification for the entire dataset after including all subclasses. The highest performance was observed with an MCC value of 0.9002 when the models were trained on the highest resolution and tested on the same resolution. Different resolutions and their effect on CNNs are explored. We show that image resolution has a clear influence on the performance which calls for standards in the field in the future.

Rapid quantification of mitochondrial fractal dimension in individual cells.

An improved technique for fractal characterization called the modified blanket method is introduced that can quantify surrounding fractal structures on a pixel by pixel basis without artifacts associated with scale-dependent image features such as object size. The method interprets images as topographical maps, obtaining information regarding the local surface area as a function of image resolution. Local fractal dimension (FD) can be quantified from the power law exponent derived from the surface area and image resolution relationship. We apply this technique on simulated cell images of known FD and compared the obtained values to power spectral density (PSD) analysis. Our method is sensitive to a wider FD range (2.0-4.5), having a mean error of 1.4% compared to 6% for PSD analysis. This increased sensitivity and an ability to compute regional FD properties enabled the discrimination of the differences in radiation resistant cancer cell responses that could not be detected using PSD analysis.

Error analysis of helmholtz-based MR-electrical properties tomography.

MR electrical properties tomography (MR-EPT) aims to measure tissue electrical properties by computing spatial derivatives of measured B1+ data. This computation is very sensitive to spatial fluctuations caused, for example, by noise and Gibbs ringing. In this work, the error arising from the computation of spatial derivatives using finite difference kernels (FD error) has been investigated. In relation to this FD error, it has also been investigated whether mitigation strategies such as Gibbs ringing correction and Gaussian apodization can be beneficial for conductivity reconstructions. Conductivity reconstructions were performed on a phantom (by means of simulations and MR measurements at 3T) and on a human brain model. The accuracy was evaluated as a function of image resolution, FD kernel size, k-space windowing, and signal-to-noise ratio. The impact of mitigation strategies was also investigated. The adopted small FD kernel is highly sensitive to spatial fluctuations, whereas the large FD kernel is more noise-robust. However, large FD kernels lead to extended numerical boundary error propagation, which severely hampers the MR-EPT reconstruction accuracy for highly spatially convoluted tissue structures such as the human brain. Mitigation strategies slightly improve the accuracy of conductivity reconstructions. For the adopted derivative kernels and the investigated scenario, MR-EPT conductivity reconstructions show low accuracy: less than 37% of the voxels have a relative error lower than 30%. The numerical error introduced by the computation of spatial derivatives using FD kernels is one of the major causes of limited accuracy in Helmholtz-based MR-EPT reconstructions. Magn Reson Med 80:90-100, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Face Recognition as a Function of Image Resolution and Viewing Distance

The fundamental challenge in facial recognition lies in understanding the role that different facial features play in our judgments of identity. Notable in this regard are the relative contributions of the internal features (eyes, nose and mouth) as well as the external features (hair and jaw-line). Past studies that have investigated this issue have typically used high resolution images or high quality line drawings as facial stimuli. The results obtained have therefore been most relevant for understanding the how faces are identified at close range. Given that real-world viewing conditions are rarely optimal, it is important to know how image degradation (such as loss of resolution caused by large viewing distances) influences our ability to perceive those internal and external features. We report experiments designed to address this issue, and our data characterizes how the relative contributions of internal and external features change as a function of image resolution and viewing distance. While we have replicated results of previous studies that have shown internal features of familiar faces to be more useful for recognition than external features at high resolution, we found that the importance of the two feature sets is flipped as resolution decreases. Additionally, our results suggest that the visual system uses a highly non-linear cue-fusion strategy in combining internal and external features along the dimensions of image resolution and distance, and that the configural cues that relate the two feature sets play a profoundly important role in judgments of facial identity.

Physiological noise and signal-to-noise ratio in fMRI with multi-channel array coils

Sensitivity in BOLD fMRI is characterized by the signal to noise ratio (SNR) of the time-series (tSNR), which contains fluctuations from thermal and physiological noise sources. Alteration of an acquisition parameter can affect the tSNR differently depending on the relative magnitude of the physiological and thermal noise, therefore knowledge of this ratio is essential for optimizing fMRI acquisitions. In this study, we compare image and time-series SNR from array coils at 3T with and without parallel imaging (GRAPPA) as a function of image resolution and acceleration. We use the “absolute unit” SNR method of Kellman and McVeigh to calculate the image SNR (SNR0) in a way that renders it comparable to tSNR, allowing determination of the thermal to physiological noise ratio, and the pseudo-multiple replica method to quantify the image noise alterations due to the GRAPPA reconstruction. The Kruger and Glover noise model, in which the physiological noise standard deviation is proportional to signal strength, was found to hold for the accelerated and non-accelerated array coil data. Thermal noise dominated the EPI time-series for medium to large voxel sizes for single-channel and 12-channel head coil configurations, but physiological noise dominated the 32-channel array acquisition even at 1mm×1mm×3mm resolution. At higher acceleration factors, image SNR is reduced and the time-series becomes increasingly thermal noise dominant. However, the tSNR reduction is smaller than the reduction in image SNR due to the presence of physiological noise.

Reliability of the CFTM and GPA methods for strain analysis at ultra-thin layers

The reliability of the Computational Fourier Transform Moiré (CFTM) and Geometric Phase Analysis (GPA) techniques for strain analysis at ultra-thin layers has been investigated using computer-generated images. Our results revealed that the leakage effect creates error that is linearly dependent on the mask size used for Fourier filtering. Error due to the leakage effect has a significant impact on the analysis of strain for small-mismatched systems with low resolution in the original image. We demonstrate that the error due to the leakage effect can be minimized with improved resolution of the original image. In order to obtain a measurement of the reliability of the CFTM and GPA methods on ultra-thin layers, we systematically quantify the error due to the leakage effect as a function of image resolution and applied strain value for the original image. The presence of the leakage effect and the resulting limitations of the CFTM and GPA methods are demonstrated using a high-resolution transmission electron microscopy (HRTEM) image of an ultra-thin heterointerface from a strained layer superlattice.

Planar laser-induced fluorescence imaging in shock tube flows

The planar laser-induced fluorescence (PLIF) imaging method was used to perform flow visualization and quantitative planar thermometry in shock tube flow fields using toluene as a fluorescence tracer in nitrogen. Fluorescence quantum yield values needed to quantify PLIF images were measured in a static cell at low pressures (<1 bar) for various toluene partial pressures in nitrogen bath gas. Images behind incident and reflected shocks were taken in the core flow away from regions affected by boundary layers. Temperature measurements from these images were successfully compared with predicted values using ideal shock equations. Measured temperatures ranged between 296 and 800 K and pressures between 0.15 and 1.5 atm. The average temperature discrepancies between measurements and the predicted values behind the incident and reflected shocks were 1.6 and 3.6%, respectively. Statistical analyses were also conducted to calculate the temperature measurement uncertainty as a function of image resolution. The technique was also applied to the study of more complex supersonic flows, specifically the interaction of a moving shock with a wedge. Measured temperatures agreed well with the results of numerical simulations in all inviscid regions, and all pertinent features of the single Mach reflection were resolved.

Low-Complexity Binary Morphology Architectures With Flat Rectangular Structuring Elements

This article describes and evaluates algorithms and their hardware architectures for binary morphological erosion and dilation. In particular, a fast stall-free low-complexity architecture is proposed that takes advantage of the morphological duality principle and structuring element (SE) decomposition. The design is intended to be used as a hardware accelerator in real-time embedded processing applications. Hence, the aim is to minimize the number of operations, memory requirement, and memory accesses per pixel. The main advantage of the proposed architecture is that for the common class of flat and rectangular SEs, complexity and number of memory accesses per pixel is low and independent of both image and SE size. The proposed design is compared to the more common delay-line architecture in terms of complexity, memory requirements and execution time, both for an actual implementation and as a function of image resolution and SE size. The architecture is implemented for the UMC 0.13- mum CMOS process using a resolution of 640 t 480. A maximum SE of 63 t 63 is supported at an estimated clock frequency of 333 MHz.

Optimizing Image Resolution to Maximize the Accuracy of Hard Classification

There are three strategies by which the accuracy of classification can be improved after the imagery that will be used for the classification has been chosen. These are to improve the definition of the class decision surfaces, to maximize the between class distances, and to reduce the within class variances. This paper reports on work done to investigate the relationship between classification accuracy and within class variances, where generally accepted measures of accuracy derived from the Confusion Matrix are used as the indicators of classification accuracy. This paper shows that the within class variances are a function of image resolution, and it provides a mechanism based on the Average Local Variance (ALV) function to find the resolution that will yield the highest relative within field classification accuracy by minimizing the within class variances.

Comparison of physiological noise at 1.5 T, 3 T and 7 T and optimization of fMRI acquisition parameters

Previous studies have shown that under some conditions, noise fluctuations in an fMRI time-course are dominated by physiological modulations of the image intensity with secondary contributions from thermal image noise and that these two sources scale differently with signal intensity, susceptibility weighting (TE) and field strength. The SNR of the fMRI time-course was found to be near its asymptotic limit for moderate spatial resolution measurements at 3 T with only marginal gains expected from acquisition at higher field strengths. In this study, we investigate the amplitude of image intensity fluctuations in the fMRI time-course at magnetic field strengths of 1.5 T, 3 T, and 7 T as a function of image resolution, flip angle and TE. The time-course SNR was a similar function of the image SNR regardless of whether the image SNR was modulated by flip angle, image resolution, or field strength. For spatial resolutions typical of those currently used in fMRI (e.g., 3 × 3 × 3 mm 3), increases in image SNR obtained from 7 T acquisition produced only modest increases in time-course SNR. At this spatial resolution, the ratio of physiological noise to thermal image noise was 0.61, 0.89, and 2.23 for 1.5 T, 3 T, and 7 T. At a resolution of 1 × 1 × 3 mm 3, however, the physiological to thermal noise ratio was 0.34, 0.57, and 0.91 for 1.5 T, 3 T and 7 T for TE near T2*. Thus, by reducing the signal strength using higher image resolution, the ratio of physiologic to image noise could be reduced to a regime where increased sensitivity afforded by higher field strength still translated to improved SNR in the fMRI time-series.

Sizing it up: Cellular MRI using micron‐sized iron oxide particles

There is rapidly increasing interest in the use of MRI to track cell migration in intact animals. Currently, cell labeling is usually accomplished by endocytosis of nanometer-sized, dextran-coated iron oxide particles. The limitations of using nanometer-sized particles, however, are that millions of particles are required to achieve sufficient contrast, the label can be diluted beyond observability by cell division, and the label is biodegradable. These problems make it difficult to label cells other than macrophages in vivo, and to conduct long-term engraftment studies. It was recently demonstrated that micron-sized iron oxide particles (MPIOs) can be taken up by a number of cell types. In this study we examined the MRI properties of single MPIOs with sizes of 0.96, 1.63, 2.79, 4.50, and 5.80 microm. Furthermore, the capacity of cells to endocytose these MPIOs was investigated, and the MRI properties of the labeled cells at 7.0 and 11.7 Tesla were measured as a function of image resolution and echo time (TE). Cells labeled with MPIOs generally contained iron levels of approximately 100 pg, which is approximately threefold higher than those obtained with the best strategies to label cells using nanometer-sized particles. On occasion, some cells had levels as high as approximately 400 pg. We demonstrate that these large particles and the cells labeled with them can be detected by spin echo (SE)-based imaging methods. These measurements indicate that MPIOs should be useful for improving cell tracking by MRI.

The Axis Ratio Distribution of Local and Distant Galaxies

Surface photometry from 16 HST/WFPC2 fields in the I(F814W) filter is used to derive the distribution of apparent axis ratios for galaxies in progressively fainter magnitude intervals for I<25. We assess the systematic and accidental errors in ellipticity measurements as a function of image resolution and signal-to-noise ratio, and statistically correct for the effect of cosmological surface brightness dimming on our isophotal measurements. The axis ratio distribution for the local galaxy population was computed using logR measurements for 1569 RC3 galaxies with Bt<13 mag. Nonparametric tests are used to show that our distant samples, in the redshift range 0.1<z<1.5, are not statistically different from the local sample. We present image montages of galaxies selected randomly from different axis ratio and apparent magnitude ranges and discuss the evolutionary consequences of the lack of a strong difference between the ellipticity distributions in near and far data sets.

The convergence of object dependent resolution in maximum likelihood based tomographic image reconstruction

Study of the maximum likelihood by EM algorithm (ML) with a reconstruction kernel equal to the intrinsic detector resolution and sieve regularization has demonstrated that any image improvements over filtered backprojection (FBP) are a function of image resolution. Comparing different reconstruction algorithms potentially requires measuring and matching the image resolution. Since there are no standard methods for describing the resolution of images from a nonlinear algorithm such as ML, the authors have defined measures of effective local Gaussian resolution (ELGR) and effective global Gaussian resolution (EGGR) and examined their behaviour in FBP images and in ML images using two different measurement techniques. For FBP these two resolution measures are equal and exhibit the standard convolution behaviour of linear systems.

Making Progressive Transmission Adaptive

Progressive transmission hierarchically reorganizes the information in an image with the aim of increasing transmission efficiency on a low bandwidth channel. We have developed methods for adaptively choosing initial and successive approximations to be sent when the end user is a human telebrowsing through a database to find particular desired images. Because of progressive transmission overhead, it is desirable to send in one step the approximation which is just good enough to allow a user to make a positive decision (e.g., want/don't want). During past and present sessions of telebrowsing, positive response times are recorded and used to estimate the actual positive response time probability mass function. This estimate is used with a previously calculated overhead function to choose the best initial and successive approximations. We discuss the application of two-channel coding to progressive transmission and conclude that little is gained by further coding of the initial lowresolution approximation. However, the interpretation of these lowresolution images is enhanced if the displayed size is a function of image resolution. We present a method for determination of the best displayed image size which is based on the modulation transfer function of the human visual system.