Search

Stabilizing videos and detecting moving objects are important tasks in many computer vision applications, though it becomes challenging because of the presence of atmospheric turbulence that causes random pixel shifting and blurring of the videos. This paper proposes an improved method for correcting geometrical distortions of videos degraded by atmospheric turbulence while keeping moving objects unaltered. In this method, three different techniques are used to generate three different masks, which are then combined together to generate a more accurate mask. This mask is employed to properly detect the moving objects and finally fusing with the background a stabilized video output is obtained. The performance of this method is tested by applying it on different real-world datasets. A comparison with an existing method shows that the proposed method gives better detection of moving objects and improved stabilization of the degraded videos.

Atmospheric turbulence causes non-uniform geometric deformation of images due to random fluctuation of refractive index throughout the imaging path. The identification of moving objects in a turbulent medium is a fundamental preprocessing step in computer vision. In this paper, a new mask generation process is proposed that removes misdetection due to turbulence in the medium. In this regard, the first step is to estimate the background frame from the video and determine the difference images from the input frames with respect to the background frame. Then three different thresholding techniques: Otsu thresholding, maximum entropy based thresholding, and standard deviation and mean based thresholding, are applied on these difference images to generate three different masks. Finally, a refined mask is generated using these three masks which detects the moving objects from the degraded video. In simulation experiment, qualitative comparisons are conducted to evaluate the performance of the proposed method with a previous one and higher accuracy is obtained.

This paper presents an improved method to detect moving objects from videos distorted by atmospheric turbulence. The method is based on generating an accurate mask from the changing properties of pixel intensities from frame to frame. The background frame is estimated by calculating the median from a sufficient number of input frames. Three different masks are generated by thresholding the difference image and pixel shiftmap of each input frame with respect to the background. A final mask is then obtained by combining all these three masks, which is more accurate than the individual ones. The performance of the proposed method is compared with that of an existing method by applying them on real-world videos. Results show that the proposed method provides better detection of moving objects than the compared method.

Predicting upcoming bands of hyperspectral images is an important task in modern image compression algorithms. This paper proposes a new algorithm to predict the band-wise correlation of hyperspectral images based on a generalized regression neural network (GRNN). The proposed algorithm uses the intensity values of the previous bands to train the GRNN and approximates the correlation between them. The next band is then predicted using the trained network and the immediately previous band. This algorithm works on a pixel-by-pixel basis and does not involve any mathematical modeling or any previous knowledge of the images. The performance of the proposed algorithm is evaluated by applying it to several Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) reflectance datasets. Simulation results show that the proposed algorithm provides substantial accuracy in the prediction of upcoming bands.

This paper presents a novel approach in image restoration based on calculating the pixel shift-maps of geometrically distorted images in a video sequence and using these maps to obtain non-distorted frames by de-warping the originals. In the similar previous approaches, the reference frame in the image registration process for pixel shift-maps calculation is selected either using a blind image quality metric or simply using the first frame. However, the former technique may not estimate the best frame accurately in the sense of geometric distortion. As for the latter, there is always a possibility that the first frame is one of the most distorted ones in the sequence. In the proposed method, an optimal reference frame search algorithm is introduced to improve the quality of restored frames and to yield a stable video output. The proposed method was applied both to synthetic and real data sets, and simulation results show significant improvement in comparison to the state-of- the-art methods.

This paper addresses the problem of stabilizing underwater videos with non-uniform geometric deformations or warping due to a wavy water surface. It presents an improved method to correct these geometric deformations of the frames, providing a high-quality stabilized video output. For this purpose, a non-rigid image registration technique is employed to accurately align the warped frames with respect to a prototype frame and to estimate the deformation parameters, which in turn, are applied in an image dewarping technique. The prototype frame is chosen from the video sequence based on a sharpness assessment. The effectiveness of the proposed method is validated by applying it on both synthetic and real- world sequences using various quality metrics. A performance comparison with an existing method confirms the higher efficacy of the proposed method.

In this paper, we present a new approach to track moving objects in videos having a dynamic background. At first, we apply an object detection algorithm that deals with the detection of real objects in a degraded video by separating them from turbulence-induced motions using a two-level thresholding technique. Then, a generalized regression neural network is used to track the detected objects throughout the frames in the video. The proposed approach utilizes the features of centroid and area of moving objects and creates the reference regions instantly by selecting the objects within a circle. The performance of the proposed approach is compared with that of an existing approach by applying them to turbulence degraded videos, and competitive results are obtained.

Ultrasound (US) imaging is widely used in clinical diagnostics as it is an economical, portable, painless, comparatively safe, and non-invasive real-time tool. However, the image quality of US imaging is severely affected by the presence of speckle noise during the acquisition process. It is essential to achieve speckle-free high resolution US imaging for better clinical diagnosis. In this paper, we propose a speckle and blur reduction algorithm for US imaging based on artificial neural networks (ANNs). Here, speckle noise is modelled as a multiplicative noise following a Rayleigh distribution, whereas blur is modelled as a Gaussian blur function. The noise and blur variances are estimated by a cascade-forward back propagation (CFBP) neural network using a set of intensity and wavelet features of the US image. The estimated noise and blur variances are then used for speckle reduction by solving the inverse Rayleigh function, and for de-blurring, using the Lucy-Richardson algorithm. The proposed approach gives improved results for both qualitative and quantitative measures.

This paper proposes a high accuracy and fast image restoration approach to restore a sequence of atmospheric turbulence degraded frames of a remote object or scene. A coarse-to-fine optical flow technique is employed to estimate the dense motion fields of the frames against a reference frame. The First Register Then Average And Subtract (FRTAAS) method is used to correct the geometric distortions and restore a high quality sequence. Finally, a non-local means filter is applied to extract noise from each frame of the sequence. A performance comparison is presented between the proposed restoration method and an earlier method in terms of computational time and accuracy. The effectiveness of the proposed approach is demonstrated on both synthetic and real-life videos.

This paper presents a comparative study between field oriented control (FOC) and direct torque control (DTC), two most popular control strategies for inverter fed interior permanent magnet synchronous motor (IPMSM) drives. The comparison is done in four switch three phase (FSTP) inverter scheme instead of six switch three phase (SSTP) inverter scheme. The FSTP inverter scheme is better than SSTP inverter scheme because of the reduction in price, switching losses and the complexity of the control board. The comparison is based on various conditions such as normal operating condition, sudden change in load torque, speed reversal and change in stator resistance. To validate the effectiveness of the drive systems, simulation is carried out in MATLAB environment. The simulation results and comparative study have been found quite satisfactory.

Restoration of a sequence of images influenced by atmospheric turbulence is a challenging task. A new approach for geometrical corrections and noise cancelations of the turbulence degraded frames of a video is presented. The time-averaged frame of the video is used to overcome the geometric deformations through an iterative robust image registration technique. The results of the registration are geometrically corrected frames which still contain noise. A non-local means filter is used to extract the noise from the individual frames. A performance comparison between our proposed restoration approach and the earlier First Register Then Average And Subtract (FRTAAS) approach is presented in terms of restoration accuracy. The effectiveness of our approach is demonstrated on both synthetic and real-world surveillance videos.

This paper proposes a new atmospheric warp estimation method based on Artificial Neural Network (ANN). We employed a Generalized Regression Neural Network (GRNN) for a-priori estimation of the upcoming warped frames using history of the previous frames. A non-rigid image registration technique is used for determining pixel shifts of the captured frames with respect to the reference frame. The proposed method is independent of the pixel-wander model. The performance of the method is evaluated using various quality metrics. Simulation results show that the proposed method provides substantial estimation of the upcoming frames with considerable errors.

A high precision and fast image restoration method is proposed to restore a geometrically corrected image from the atmospheric turbulence degraded video sequence of a static scenery. In this approach, we employ an optical flow technique to register all the frames of the distorted video to a reference frame and determine the flow fields. We use the First Register Then Average And Subtract-variant (FRTAASv) method to correct the geometric distortions using the computed flow fields. We present a performance comparison between our proposed restoration method and earlier Minimum Sum of Squared Differences (MSSD) image registration based FRTAASv method in terms of computational time and accuracy. Simulation experiments show that our proposed method provides higher accuracy with quicker processing time.

In this paper, a fast image restoration method is proposed to restore the true image from an atmospheric turbulence degraded video. A non-rigid image registration algorithm is employed to register all the frames of the video to a reference frame and determine the shift maps. The First Register Then Average And Subtract-variant (FRTAASv) method is applied to correct the geometric distortion of the reference frame. A performance comparison is presented between the proposed restoration method and the earlier Minimum Sum of Squared Differences (MSSD) image registration based FRTAASv method, in terms of processing time and accuracy. Simulation results show that the proposed method requires shorter processing time to achieve the same geometric accuracy.

A position sensorless Surface Permanent Magnet Synchronous Motor (SPMSM) drive based on flux angle is presented in this paper. The motor equations are written in rotor fixed d-q reference frame. A PID controller is used to process the speed error to generate the reference torque current keeping the magnetizing current fixed. The estimated stator flux using Recurrent Neural Network (RNN) is used to find out the rotor position. The flux angle and the reference current phasor angle are used in vector rotator to generate the reference phase currents. Hysteresis current controller block controls the switching of the 3-phase inverter to apply voltage to the motor stator. Simulation studies on different operating conditions indicate the acceptability of the drive system. The drive system only requires a speed transducer and is free from position sensor requirement. The proposed control scheme is robust under load torque disturbances and motor parameter variations. It is also simple and low cost to implement in a practical environment.