Noise In Image Data Research Articles

Combating Label Noise in Image Data Using MultiNET Flexible Confident Learning

Deep neural networks (DNNs) have been used successfully for many image classification problems. One of the most important factors that determines the final efficiency of a DNN is the correct construction of the training set. Erroneously labeled training images can degrade the final accuracy and additionally lead to unpredictable model behavior, reducing reliability. In this paper, we propose MultiNET, a novel method for the automatic detection of noisy labels within image datasets. MultiNET is an adaptation of the current state-of-the-art confident learning method. In contrast to the original, our method aggregates the outputs of multiple DNNs and allows for the adjustment of detection sensitivity. We conduct an exhaustive evaluation, incorporating four widely used datasets (CIFAR10, CIFAR100, MNIST, and GTSRB), eight state-of-the-art DNN architectures, and a variety of noise scenarios. Our results demonstrate that MultiNET significantly outperforms the confident learning method.

Analysis of Camera Calibration with Respect to Measurement Accuracy

Machine vision is used for applications such as automated inspection, process control and robot guidance, and is directly associated with increasing of manufacturing process flexibility. The presence of noise in image data affects robustness and accuracy of machine vision, which can be an obstacle for industrial applications. Accuracy depends on both feature detection, resulting in pixel values of the measures of interest, and vision systems calibration, which allows transforming pixel measurements into real-world coordinates. This paper analyzes the camera calibration process, and proposes a new method for camera calibration, based on numerical analysis of probability distributions of the calibration parameters and removal of outliers. The method can be used to improve accuracy and robustness of the vision systems calibration process.

Regularization parameter selection methods for ill-posed Poisson maximum likelihood estimation

In image processing applications, image intensity is often measured via the counting of incident photons emitted by the object of interest. In such cases, image data noise is accurately modeled by a Poisson distribution. This motivates the use of Poisson maximum likelihood estimation for image reconstruction. However, when the underlying model equation is ill-posed, regularization is needed. Regularized Poisson likelihood estimation has been studied extensively by the authors, though a problem of high importance remains: the choice of the regularization parameter. We will present three statistically motivated methods for choosing the regularization parameter, and numerical examples will be presented to illustrate their effectiveness.

Recursive least squares smoothing of noise in images

In the recent past considerable attention has been devoted to the application of Kalman filtering to smoothing out observation noise in image data. A generalization of the one-dimensional Kalman filter to two dimensions was earlier suggested by Habibi, but it has since been shown that this generalization is invalid since it does not preserve the optimality of the Kalman filter. A new method is proposed here that enables well-established Kalman-filter theory to yield a simple two-dimensional filter for images that can be modeled by two-dimensional wide-sense Markov (WSM) random fields.