SIFT method is used to extract Keypoints of the imagein order to overcome the problems of matching between the satellite and aerial images, including: difference in scale, rotation, brightness intensity and the geometric shape. Unfortunately, SIFT method extracts several unfavorable Keypoints of satellite and aerial images because of the turbulence and the environmental factors which leads to unreliable matching and increasing complexity. In order to improve the quality of the extracted specific areas and the run time of the algorithm, first the edges of the original images are extracted by Sobel operator and thresholding, then by using the SIFT method, Keypoints are extracted from the edge image. After extracting Keypoints, using the rBREIF method, that have stability dependence with respect to atmospheric turbulence and rotation, descriptor for every point of the extracted points is created. Then by applying the bilateral image matching and the RANSAC method that removes the unfavorable adaptive points, the correct matching between the satellite and aerial images are found using the suggested method. The results of the proposed method on the real images show the superiority of this method in term of the accuracy and speed, compared to the some well-known matching methods such as SIFT.

Matching remote sensing images is a challenging issue in computer vision applications. Due to the very large dimensions, local destructions, radiometric distortions, and geometric changes in the input images, the existing matching algorithms such as Scale Invariant Feature Transform (SIFT) produce a large number of false matches. Moreover, due to the high dimensional images a big number of Keypoints are extracted in large-scale satellite images. A very large number of Keypoints increases the computational, memory and time complexity in the stages of feature description and matching the Keypoints. In this paper, the geometric relationships between the key points extracted from the input images, are used to improve the detection process of false corresponding points and also to increase the speed of the SIFT algorithm. The proposed false correspondence removal algorithm uses the histogram of the scale difference values and the two image rotation angle. In the following, two new algorithms which are based on the hierarchical strategy are proposed to increase the speed of the SIFT algorithm. The first proposed algorithm is based on finding the optimal octaves in the scale space of the SIFT algorithm and selecting their compared Keypoints. In the second method, the parameters of the affine transformation which are between the two images are calculated by performing an initial matching, and then this transformation is used to reduce the search space in the final matching stage of the Keypoints. Finally, to check the performance and accuracy of each of the proposed methods, a variety of simulated and real images have been used. Moreover, for the final evaluation of the proposed algorithms, the obtained results are compared with SIFT, SR-SIFT and SIFT-GSI methods. The experimental results confirm the accuracy, stability and high speed of the proposed methods in matching satellite images.

The Scale Invariant Feature Transform (SIFT) algorithm is one of the most widely used algorithms in the machine vision field on which researchers have extensively studied and improved. SIFT is one of the common local detectors used in image registration, image mosaicking, copy-move image forgery, and etc. In this review paper, along with introducing the SIFT algorithm, the applications, pros and cons, modifications, categories and new research approaches in this algorithm are discussed. In addition, via four experiments, different aspects of this algorithm have been evaluated. This paper can help image processing researchers when utilizing the SIFT algorithm. The aim of the authors was to explore all the aspects of this algorithm.

Targets and objects registration and tracking in a sequence of images play an important role in various areas. One of the methods in image registration is feature-based algorithm which is accomplished in two steps. The first step includes finding features of sensed and reference images. In this step, a scale space is used to reduce the sensitivity of detected features to the scale changes. Afterward, we attribute feature points that obtained in the first step, descriptions using brightness value around the feature points. In this paper, a new algorithm is proposed based on Binary Robust Invariant Scalable Keypoints (BRISK) and Scale Invariant Feature Transform (SIFT) algorithms. The proposed algorithm uses the directional pattern to describe the edges which are around the Keypoints. This pattern is perpendicular to the direction of Keypoints which shows the direction of the edge and provides more useful information regarding brightness around the feature point to make descriptor vector. Furthermore, in the proposed algorithm, the output vector consists of multilevel values instead of binary values which means further useful information is involved in the descriptor vector. Also, levels of output vectors can be adjusted using a single parameter so that the processor with low computing ability can tune the output to a binary vector. Experimental results show that the proposed algorithm is more robust than the BRISK algorithm and the efficiency of the algorithm is about the same as BRISK algorithm.

Image mosaicing refers to stitching two or more images which have overlapping regions to a larger and more comprehensive image. Scale Invariant Feature transform (SIFT) is one of the most commonly used detectors previously used in image mosaicing. The defects of SIFT algorithm are the large number of redundant Keypoints and high execution time due to the high dimensions of classical SIFT descriptor, that reduces the efficiency of this algorithm. In this paper, to solve these problems a new four-step approach for image mosaicing is proposed. At first, the Keypoints of both reference and sensed images are extracted based on Redundant Keypoint Elimination-SIFT (RKEM-SIFT) algorithm to improve the mosaicing process. Then, to increase the speed of the algorithm, the 64-D SIFT descriptor for Keypoints description is used. Afterwards, the proposed RANdom SAmple Consensus (RANSAC) algorithm is used for removing mismatches. Finally, a new method for image blending is proposed. The details of the proposed steps are as follows. RKEM-SIFT algorithm has been proposed in [1] to eliminate redundant points based on redundancy index. In this paper, RKEM algorithm is used to extract Keypoints to improve the accuracy of image mosaicing. In the second stage, for each keypoint of the image, 64-D SIFT descriptor is computed. In this descriptor, unlike the 128-D SIFT descriptor, a smaller window is used which improves the accuracy of matching and reduces the running time. In the third stage, the proposed adaptive RANSAC algorithm is suggested to determine the adaptive threshold in the RANSAC algorithm to remove the mismatches and to improve the image mosaicing. Determining the appropriate threshold value in RANSAC is so important, because if an appropriate value is not chosen for this algorithm, the mismatches are not removed, and eventually there will be a serious impact on the outcome of the image mosaicing process. In this method, the threshold value is based on the median value of distances between matching points and their transformed model. Image blending in the mosaicing process is the final step which blends the pixels intensity in the overlapped region to avoid seams. The proposed method of blending is to combine the images based on the Gaussian weighting function, which the mean of this function is considered as the average of the data in the overlapped region of two images. The proposed blending method reduces artifacts in the image for better performance of the mosaicing process. Another advantage of this proposed method is the possibility to combine more than two images that are suitable for creating panoramic images. The simulation results of the proposed image mosaicing technique, which includes the RKEM-SIFT algorithm as feature detector, 64-D SIFT descriptor, proposed adaptive RANSAC algorithm and proposed image blending algorithm on different image databases show the superiority of the proposed method according to RMSE criteria, precision and running time.

Farsi font detection is considered as the first stage in the Farsi optical character recognition (FOCR) of scanned printed texts. To this aim, this paper proposes an improved version of the speeded-up robust features (SURF) algorithm, as the feature detector in the font recognition process. The SURF algorithm suffers from creation of several redundant features during the detection phase. Thus, the presented version employs the redundant keypoint elimination method (RKEM) to enhance the matching performance of the SURF by reducing unnecessary Keypoints. Although the performance of the RKEM is acceptable in this task, it exploits a fixed experimental threshold value which has a detrimental impact on the results. In this paper, an Adaptive RKEM is proposed for the SURF algorithm which considers image type and distortion, when adjusting the threshold value. Then, this improved version is applied to recognize Farsi fonts in texts. To do this, the proposed Adaptive RKEM-SURF detects the Keypoints and then SURF is used as the descriptor for the features. Finally, the matching process is done using the nearest neighbor distance ratio. The proposed approach is compared with recently published algorithms for FOCR to confirm its superiority. This method has the capability to be generalized to other languages such as Arabic and English.

Predicting pedestrians' intentions to cross paths with cars, particularly at intersections and crosswalks, is critical for autonomous systems. While recent studies have showcased the effectiveness of deep learning models based on computer vision in this domain, current models often lack the requisite confidence for integration into autonomous systems, leaving several unresolved issues. One of the fundamental challenges in autonomous systems is accurately predicting whether pedestrians intend to cross the path of a self-driving car. Our proposed model addresses this challenge by employing convolutional neural networks to predict pedestrian crossing intentions based on non-visual input data, including body pose, car velocity, and pedestrian bounding box, across sequential video frames. By logically arranging non-visual features in a 2D matrix format and utilizing an RGB semantic map to aid in comprehending and distinguishing fused features, our model achieves improved accuracy in pedestrian crossing intention prediction compared to previous approaches. Evaluation against the criteria of the JAAD database for pedestrian crossing intention prediction demonstrates significant enhancements over prior studies.

Given the increasing daily volume of videos generated by security cameras in personal and public spaces, monitoring the activities present in videos has become crucial. Many video surveillance systems are designed to verify performance accuracy and provide alerts during the occurrence of abnormal activities. In this regard, various intelligent models have been proposed for detecting activities in videos. Considering recent advances in artificial intelligence, particularly deep learning, this paper introduces a model based on the Transformer network. To reduce computational complexity, Keypoints of the human body are utilized in this approach. Fifteen key body points are input into the Transformer model, leveraging parallel processing during training and a self-attention mechanism. This enhances the speed and accuracy of the model. Experimental results on the JHMDB public database indicate an improvement in the accuracy of detecting abnormal activities compared to baseline models.

There are many applications for face recognition. Due to illumination changes, and pose variations of facial images, face recognition is often a challenging and a complicated process. In this paper, we propose an e ective and robust face recognition method. Firstly, we select those areas from the face (such as eyes, nose, and mouth), which are more informative in face recognition. Then SIFT (Scale Invariant Feature Transform) descriptor is utilized for feature extraction from the selected areas. SIFT descriptor detects Keypoints in the image and describes each keypoint with a feature vector with length 128. To speed up the proposed method, PCA (Principal Component Analysis) is applied on the SIFT feature vector to reduce the vector's length. Finally, Kepenekci matching method is used to assess similarity between the images. The proposed method is evaluated on the ORL, Extended Yale B, and FEI databases. Results show considerable performance of the proposed face recognition method in comparison with several state-of-the-arts.