Iranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Brain Tumor Detection in 3D MRI images using Kapur's Entropy and flood fill algorithmBrain Tumor Detection in 3D MRI images using Kapur's Entropy and flood fill algorithm117139106FAMostafa NezamzadehPh.D Student of Electronics Engineering, Faculty of Engineering, Lorestan University, Khorramabad, IranVahid MehrdadDepartment of Electrical and Electronics, Engineering, Faculty of Engineering, Lorestan UniversityJournal Article20211022The Brain tumor is one of the most important factors in mortality, so timely and appropriate detection is necessary to treat the tumor. In this study, 3D images are used to detect tumor. 3D images have depth and therefore, blind spots that may be hidden in 2D images can be seen. This paper presents a threshold method using Kapur’s entropy to detect brain tumors in 3D MRI images. In the proposed method, in order to differentiate the tumor area, the images are normalized in three dimensions, which has the advantage that the brightness level of the tumor is brighter than the rest of brain. In the next step, the 3D image is sliced in 3D and converted into 2D images. By applying two steps of Kapur’s entropy to two-dimensional images of the tumor area with points that have a higher brightness level than the threshold value are separated. To remove Additional areas, a 3D image is first made by stacking 2D images on top of each other, and then the 3D area is extracted using a 3D morphology filter and flood-fill algorithm the advantages of the proposed method is the removal of excess areas while preserving the tumor area and covering all angles of the tumor in three dimensions. To show the efficiency of the proposed method, the BRATS database was used. The evaluation results for detecting tumor were evaluated with similarity, sensitivity and specificity coefficients of 0.9407, 0.9235 and 0.999, respectively, which have better performance than the proposed methods.The Brain tumor is one of the most important factors in mortality, so timely and appropriate detection is necessary to treat the tumor. In this study, 3D images are used to detect tumor. 3D images have depth and therefore, blind spots that may be hidden in 2D images can be seen. This paper presents a threshold method using Kapur’s entropy to detect brain tumors in 3D MRI images. In the proposed method, in order to differentiate the tumor area, the images are normalized in three dimensions, which has the advantage that the brightness level of the tumor is brighter than the rest of brain. In the next step, the 3D image is sliced in 3D and converted into 2D images. By applying two steps of Kapur’s entropy to two-dimensional images of the tumor area with points that have a higher brightness level than the threshold value are separated. To remove Additional areas, a 3D image is first made by stacking 2D images on top of each other, and then the 3D area is extracted using a 3D morphology filter and flood-fill algorithm the advantages of the proposed method is the removal of excess areas while preserving the tumor area and covering all angles of the tumor in three dimensions. To show the efficiency of the proposed method, the BRATS database was used. The evaluation results for detecting tumor were evaluated with similarity, sensitivity and specificity coefficients of 0.9407, 0.9235 and 0.999, respectively, which have better performance than the proposed methods.https://jmvip.sinaweb.net/article_139106_9406566bd913fc1de0d52901cbe5db43.pdfIranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Embedded Feature Representation in Dynamic Time Warping Space for 3D Action Recognition Using Kinect Depth SensorEmbedded Feature Representation in Dynamic Time Warping Space for 3D Action Recognition Using Kinect Depth Sensor1934139115FAMohsen TabejamaatMS.C Student of Electrical Engineering, Sharif University of TechnologyHoda MohammadzadeDepartment of Electrical Engineering, Sharif University of TechnologyJournal Article20211022This paper proposes a novel 3D action recognition technique which uses the skeletal information extracted from depth image sequences. First, each action is represented by a multidimensional time series where each dimension represents the position variation of one skeleton joint over time. The time series is then mapped into the kernel Hilbert space using a metric defined by Dynamic Time Warping distance. Afterwards, regularized Fisher strategy is used to remap the kernel space into a discriminative one. This incorporates the correlation-distinctiveness relationship of the sequences into the recognition process and also mitigates the curse of dimensionality effect in the kernel space. Unlike traditional kernel functions, the time warping used in the mapping strategy makes the kernel space robust to the temporal shift variations of the motion sequences. Moreover, our method eliminates the need for a complex design method for extracting the static and dynamic information of a motion sequence. A set of extensive experiments on three publically available databases; TST, UTKinect, and UCFKinect demonstrates the superiority of our method compared to a set of baseline algorithms.This paper proposes a novel 3D action recognition technique which uses the skeletal information extracted from depth image sequences. First, each action is represented by a multidimensional time series where each dimension represents the position variation of one skeleton joint over time. The time series is then mapped into the kernel Hilbert space using a metric defined by Dynamic Time Warping distance. Afterwards, regularized Fisher strategy is used to remap the kernel space into a discriminative one. This incorporates the correlation-distinctiveness relationship of the sequences into the recognition process and also mitigates the curse of dimensionality effect in the kernel space. Unlike traditional kernel functions, the time warping used in the mapping strategy makes the kernel space robust to the temporal shift variations of the motion sequences. Moreover, our method eliminates the need for a complex design method for extracting the static and dynamic information of a motion sequence. A set of extensive experiments on three publically available databases; TST, UTKinect, and UCFKinect demonstrates the superiority of our method compared to a set of baseline algorithms.https://jmvip.sinaweb.net/article_139115_d47b9921ad20e1a8f4a11977c69947f7.pdfIranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Feature Selection for no reference multi-distortion image quality assessment Based on Particle Swarm Optimization AlgorithmFeature Selection for no reference multi-distortion image quality assessment Based on Particle Swarm Optimization Algorithm3548139316FAZahra DavoodiIT Engineer, Shahid Beheshti UniversityJournal Article20211027In this paper, a no-reference metric for evaluating the quality of multi-distortion images is introduced. This metric is based on a combination of structural features and image brightness. First, the structural features and brightness of the image, which change drastically due to distortion, were extracted. For different datasets, an optimal combination of properties was obtained by the particle swarm optimization algorithm. The optimal combination of features was supported by regression vector regression to the training model so that the trained model could measure the quality of other images. Due to the comprehensiveness of the selected features, this metric has the ability to measure image quality with a variety of degradations. According to the results of the implementation of the criterion, we had a significant improvement and also according to the research, the optimal combination of image properties has been obtained to investigate specific degradations, which can be useful for further research in the future.In this paper, a no-reference metric for evaluating the quality of multi-distortion images is introduced. This metric is based on a combination of structural features and image brightness. First, the structural features and brightness of the image, which change drastically due to distortion, were extracted. For different datasets, an optimal combination of properties was obtained by the particle swarm optimization algorithm. The optimal combination of features was supported by regression vector regression to the training model so that the trained model could measure the quality of other images. Due to the comprehensiveness of the selected features, this metric has the ability to measure image quality with a variety of degradations. According to the results of the implementation of the criterion, we had a significant improvement and also according to the research, the optimal combination of image properties has been obtained to investigate specific degradations, which can be useful for further research in the future.https://jmvip.sinaweb.net/article_139316_31aa0ea5f61c4db40a4b41226eb11aed.pdfIranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Automatic Tassel Detection to Estimate Flowering Date in the UAV Images using Deep Learning TechniquesAutomatic Tassel Detection to Estimate Flowering Date in the UAV Images using Deep Learning Techniques4963141101FASeyedeh Farveh MusaviImage Processing Lab, Dep. of Physics, Shahid Bahonar University, Kerman, IranAzam KaramiDep. of Physics, Shahid Bahonar University, Kerman, IranJournal Article20211202Estimating crop yields and examining growth trends in different species of a crop in precision agriculture is very important for researchers and agricultural experts. In this article, a new technique based on one-stage objection detection called GP-YOLOv5 for automatic tassel detection in the UAV images of a large maize field at different growing stages and flowering date estimation is presented. Because of the existing small number of tassels in the early stages of growth, GP-GAN is used to augment the training data. After that, the hyperparameters of the YOLOv5 are optimized to increase the tassel detection accuracy. Plant counting using CenterNet in the early stage of growth is calculated to determine the flowering date. Finally, well-known interpolation and prediction algorithms are used to estimate the flowering date. The proposed method is compared with two state-of-the-art methods based on detection “CenterNet” and regression “TasselNetv2+” technique for tassel counting. The average accuracy of GP-YOLOv5 for tassel detection is around 96.81 % and for the CenterNet method, it is around 81.78 %, which indicates that the accuracy of the proposed method is higher than the CenterNet technique.Estimating crop yields and examining growth trends in different species of a crop in precision agriculture is very important for researchers and agricultural experts. In this article, a new technique based on one-stage objection detection called GP-YOLOv5 for automatic tassel detection in the UAV images of a large maize field at different growing stages and flowering date estimation is presented. Because of the existing small number of tassels in the early stages of growth, GP-GAN is used to augment the training data. After that, the hyperparameters of the YOLOv5 are optimized to increase the tassel detection accuracy. Plant counting using CenterNet in the early stage of growth is calculated to determine the flowering date. Finally, well-known interpolation and prediction algorithms are used to estimate the flowering date. The proposed method is compared with two state-of-the-art methods based on detection “CenterNet” and regression “TasselNetv2+” technique for tassel counting. The average accuracy of GP-YOLOv5 for tassel detection is around 96.81 % and for the CenterNet method, it is around 81.78 %, which indicates that the accuracy of the proposed method is higher than the CenterNet technique.https://jmvip.sinaweb.net/article_141101_718a915d8e40811fa014dff19fc9be14.pdfIranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Classification of medical images of skin lesions using capsular neural networkClassification of medical images of skin lesions using capsular neural network6578144807FANarges HasanpourMSc. of Artificial Intelligence, ShahidBahonar university of KermanOmid EslamBSc of Software Engineering, ShahidBahonar university of KermanHadis MohseniAssistant professor of Artifitial Intelligence, ShahidBahonar university of KermanJournal Article20220214Deep networks are a type of learning method that can model high-level relationships in data. One of the most widely used types of deep models are convolutional networks that are able to model spatial dependencies in images using convolutional layers, but do not consider the hierarchical spatial structures within the image. Capsule networks are one of the new ideas proposed for modeling the hierarchical structure of features in the image, which use grouped capsules or neurons with a dynamic routing algorithm. Despite the effectiveness of the idea of capsule networks on simple data sets, the performance of these networks on complex data is still unclear. In this paper, the performance of this network is examined on a complex skin cancer dataset, which has been selected due to the importance of skin lesions diagnosis in medicine, the complexity and huge number of images and the imbalance of categories. In order to better extract the diversity of skin lesions, changes were made in the initial layers of the network. Also, due to the imbalance in the mentioned data set, changes were made in the cost function of the network. The effect of using different activation functions in the network was also investigated. The results show that the idea of a capsule network can be used optimally on complex data sets by making appropriate adjustmentsDeep networks are a type of learning method that can model high-level relationships in data. One of the most widely used types of deep models are convolutional networks that are able to model spatial dependencies in images using convolutional layers, but do not consider the hierarchical spatial structures within the image. Capsule networks are one of the new ideas proposed for modeling the hierarchical structure of features in the image, which use grouped capsules or neurons with a dynamic routing algorithm. Despite the effectiveness of the idea of capsule networks on simple data sets, the performance of these networks on complex data is still unclear. In this paper, the performance of this network is examined on a complex skin cancer dataset, which has been selected due to the importance of skin lesions diagnosis in medicine, the complexity and huge number of images and the imbalance of categories. In order to better extract the diversity of skin lesions, changes were made in the initial layers of the network. Also, due to the imbalance in the mentioned data set, changes were made in the cost function of the network. The effect of using different activation functions in the network was also investigated. The results show that the idea of a capsule network can be used optimally on complex data sets by making appropriate adjustmentshttps://jmvip.sinaweb.net/article_144807_19622f393361f7893a84dcc2d7f718da.pdfIranian Society of Machine Vision and Image ProcessingJournal of Machine Vision and Image Processing2383-11979320220923Reconstruction of illegible QR codes using deep neural networkReconstruction of illegible QR codes using deep neural network7989144810FAMilad MonfaredMSc. student in Artificial Intelligence, Department of Mechanic, Electrical and Computer Science, Islamic Azad University Science and Research, Tehran, Iran0000-0001-5985-1828Abbas KoochariDepartment of Mechanic, Electrical and Computer Science, Islamic Azad University Science and Research, Tehran, IranJournal Article20220214Todays, barcodes play a significant role in various industries, and among the two-dimensional barcodes, the most famous one is QR code (Quick Response code) that has grown widely.<br />The main purpose of this paper is to provide a noise-cancellation method based on a autoencoder deep neural network that can be used to restore distorted and illegible QRs to readability.<br />To create noise and distortion, unlike other articles that used added simulated noise to the image, the challenge of extracting QR coded into a color image was used to collect more realistic data by collecting real-world dataset. therefore we Have more reliable estimation of proposed QRs noise-canceling method. As a result, we created a comprehensive data set of distorted QRs from three different watermark extraction approaches after the screen-camera attack. For the noise reduction process, three independent MCNN networks ( which is an upgrade from the U-net network) are used for each of the three extraction approaches,Todays, barcodes play a significant role in various industries, and among the two-dimensional barcodes, the most famous one is QR code (Quick Response code) that has grown widely.<br />The main purpose of this paper is to provide a noise-cancellation method based on a autoencoder deep neural network that can be used to restore distorted and illegible QRs to readability.<br />To create noise and distortion, unlike other articles that used added simulated noise to the image, the challenge of extracting QR coded into a color image was used to collect more realistic data by collecting real-world dataset. therefore we Have more reliable estimation of proposed QRs noise-canceling method. As a result, we created a comprehensive data set of distorted QRs from three different watermark extraction approaches after the screen-camera attack. For the noise reduction process, three independent MCNN networks ( which is an upgrade from the U-net network) are used for each of the three extraction approaches,https://jmvip.sinaweb.net/article_144810_df9bf0022b382c38383fc3601e2d3874.pdf