In this paper, we present a novel approach for image selective smoothing by the evolution of two paired nonlinear partial differential equations. The distribution coefficient in de-noising equation controls the speed of distribution, and is determined by the edge-strength function. In the previous works, the edge-strength function depends on isotropic smoothing of the image, which results in failing to preserve corners and junctions, and may also result in failing to resolve small features that are closely grouped together. The proposed approach obtains the edge-strength function by solving a nonlinear distribution equation governed by the norm of the image gradient. This edge-strength function is then introduced into a well-studied anisotropic distribution model to yield a regularized distribution coefficient for image smoothing. An explicit numerical scheme is employed to efficiently solve these two paired equations.Compared with the existing methods, the proposed approach has the advantages of more detailed preservation and implementational simplicity. Experimental results on the synthesis and real images confirm the validity of the proposed approach.

Due to extensive SAR applications and the need to recognize SAR image details, the issue of improving the quality of these images after formation has been widely considered. Due to the nature of SAR image formation, the multiplicative speckle noise is considered as the most important factor in the quality degradation of these images. In this paper, a new method for removing speckle noise is presented. The main ideas of this article are using MAP estimator in accordance with the noise distribution function and presentation of a local convex optimization problem along with employment of adaptive smoothing, sparse representation regularizations and projection to the feature space. The local optimization model and adaptive smoothing provide proper noise removal and strong edges preservation and prevent image over smoothing. Also using sparse representation leads to texture preservation, and projection to the feature space enhances the algorithm against high noise levels. In order to solve the optimization problem, a method based on alternating minimization is introduced. The simulation results show good performance of the proposed method in noise reduction and preservation of image details which is better than many existing methods.

In this work, a robust rain removal algorithm is proposed for removing rain from still images. The algorithm uses a deep network architecture called DerainNet for effective rain removal. The proposed network directly learns the mapping relationship between rainy and clean image detail layers from the given set of data. In order to modify the objective function and also to improve the deraining process, other Deep CNN based architecture increases the width or depth of the neurons, which in turn increases the complexity of the network. But this work makes use of the Image Processing domain knowledge which reduces the complexity of the network. Instead of training the entire image, only the detail layer of the image is trained. The detailed layer of the image is obtained using two low-pass filters one after the other. They are guided filter and L0-Smoothing filter. The results obtained prove that the proposed network performs better deraining on images in comparison to paper [2] with light rain streaks. Python version 3.8 is used for this work.

Objective(s): The aim of this study is to examine the effect of different smoothing filters on the image quality and SUVmax to achieve the guideline recommended positron emission tomography (PET) image without harmonization.Methods: We used a Biograph mCT PET scanner. A National Electrical Manufacturers Association (NEMA) the International Electrotechnical Commission (IEC) body phantom was filled with 18F solution with a background activity of 2.65 kBq/mL and a sphere-to-background ratio of 4. PET images obtained with the Biograph mCT PET scanner were reconstructed using the ordered subsets-expectation maximization (OSEM) algorithm with time-of-flight (TOF) models (iteration, 2; subset, 21); smoothing filters including the Gaussian, Butterworth, Hamming, Hann, Parzen, and Shepp-Logan filters with various full width at half maximum (FWHM) values (1-15 mm) were applied. The image quality was physically assessed according to the percent contrast (QH, 10), background variability (N10), standardized uptake value (SUV), and recovery coefficient (RC). The results were compared with the guideline recommended range proposed by the Japanese Society of Nuclear Medicine and the Japanese Society of Nuclear Medicine Technology. The PET digital phantom was developed from the digital reference object (DRO) of the NEMA IEC body phantom smoothed using a Gaussian filter with a 10-mm FWHM and defined as the reference image. The difference in the SUV between the PET image and the reference image was evaluated according to the root mean squared error (RMSE).Results: The FWHMs of the Gaussian, Butterworth, Hamming, Hann, Parzen, and Shepp-Logan filters that satisfied the image quality of the FDG-PET/CT standardization guideline criteria were 8-12 mm, 9-11 mm, 9-13 mm, 10-13 mm, 9-11 mm, and 12- 15 mm, respectively. The FWHMs of the Gaussian, Butterworth, Hamming, Hann, Parzen, and Shepp-Logan filters that provided the smallest RMSE between the PET images and the 3D digital phantom were 7 mm, 8 mm, 8 mm, 8 mm, 7 mm, and 11 mm, respectively.Conclusion: The suitable FWHM for image quality or SUVmax depends on the type of smoothing filter that is applied.

In this paper, we focus on solving the system of absolute value equations (AVE), which is one of the most popular classes of nonlinear equations. First, a new smoothing technique with three different smoothing functions is introduced, and the AVE is transformed into a family of parametrized smooth equations with the help of these smoothing functions. Then, a smoothing Newton-type algorithm with hybridized inexact line search is developed based on the proposed smoothing technique. The numerical experiments have been carried out on some well-known and randomly generated test problems, and the results are analyzed in terms of line search techniques. The numerical results show that the proposed hybrid approach is more efficient than the other algorithms.

Integration of global positioning system (GPS) and inertial navigation system (INS) is used in airborne gravimetry to gravity field recovery. Since GPS computed position is noisy therefore the GPS acceleration which is the result of twice differentiation of GPS position will be too noisy as well. In this paper IIR low-pass filter and Kalman filter are used to smoothing the GPS acceleration and their result compared to B-spline smoother result. B-spline smoothing accuracy is reported about 1mGal in this paper data, therefore B-spline smoothing considered as a reference smoothing method. The correlation of IIR low-pass filter and Kalman filter results with B-spline smoothing result is about 97.55 and 99.83 percent, respectively. It shows that the Kalman filter result is closer to B-spline smoother. On the other hand, along with ease of design of IIR lowpass filter some other advantages such as fast computing algorithm in signal processing unlimited response hit and less memory requirement are worth mentioning. Therefore, in project with huge among of data the IIR low-pass filter could be efficient and causes the time and cost saving. Mentioned smoothing methods can also be used in INS instrumental noise reduction. Therefore, less accurate INS can be used in integration with GPS, which causes the INS cost saving and project productivity promotion.

GPS system, a satellite positioning system with an accuracy about centimeter under special methods and conditions is used by smoothing algorithm for determining non real time location in Kinematic case. This algorithm works based on previous epoches phase observation and after the desired epoch. Since in this method more observations are used relative to real time location determining with Kalman filtering algorithm, it is expected to have good accuracy. To study this matter, both methods were tested with phase observations in Kinematic case. Comparing the variance covariance matrix of obtained unknowns showed the accuracy of the smoothing algorithm. The obtained results and the comparison are given by all epoch standard deviation graphs in this paper.

Purpose: The study purpose was to determine and compare the Recall and Precision of Bing and Google Image search engines for content based image retrieval. Methodology: The research used webometrics and comparative methods. Population includes images stored in the databases of Bing and Google search engines, and research sample includes 15 selective images searched in any of search engines. All the retrieved sources through the images by image content based image search were gathered, results’ Recall and Precision measures were calculated by relevance formula and their average percentage were obtained. Research hypotheses were tested by U Mann-Whitney test as well. Findings: Findings showed that the Google search engine functionality was higher with recall measure of % 88. 73 than recall rate (%20. 86) for Bing. But Bing search engine had higher precision (% 99. 86) than Google (%94. 80). Results: Hypotheses tests on recall and precision in two search engines’ image retrieval showed a significant difference for recall in favor of Google, indicating its better functionality than Bing but there was no significant difference between them concerning precision since both showed fair precision however Bing was relatively useful.