Mohamed M. Fouad

Image Registration Under Illumination Variations Using Region-Based Confidence Weighted

We present an image registration model for image sets with arbitrarily shaped local illumination variations between images. Any nongeometric variations tend to degrade the geometric registration precision and impact subsequent processing. Traditional image registration approaches do not typically account for changes and movement of light sources, which result in interimage illumination differences with arbitrary shape. In addition, these approaches typically use a least-square estimator that is sensitive to outliers, where interimage illumination variations are often large enough to act as outliers. In this paper, we propose an image registration approach that compensates for arbitrarily shaped interimage illumination variations, which are processed using robust M-estimators tuned to that region. Each M-estimator for each illumination region has a distinct cost function by which small and large interimage residuals are unevenly penalized. Since the segmentation of the interimage illumination variations may not be perfect, a segmentation confidence weighting is also imposed to reduce the negative effect of mis-segmentation around illumination region boundaries. The proposed approach is cast in an iterative coarse-to-fine framework, which allows a convergence rate similar to competing intensity-based image registration approaches. The overall proposed approach is presented in a general framework, but experimental results use the bisquare M-estimator with region segmentation confidence weighting. A nearly tenfold improvement in subpixel registration precision is seen with the proposed technique when convergence is attained, as compared with competing techniques using both simulated and real data sets with interimage illumination variations.

M

In this paper, we focus on the sub-pixel geometric registration of images with arbitrarily-shaped local intensity variations, particularly due to shadows. Intensity variations tend to degrade the performance of geometric registration, thereby degrading subsequent processing. To handle intensity variations, we propose a model with illumination correction that can handle arbitrarily-shaped regions of local intensity variations. The approach is set in an iterative coarse-to-fine framework with steps to estimate the geometric registration with illumination correction and steps to refine the arbitrarily-shaped local intensity regions. The results show that this model outperforms linear scalar model by a factor of 6.8 in sub-pixel registration accuracy.

-Estimators

In geometric image registration, illumination variations that exist between image pairs tend to degrade the precision of the registration, which can negatively affect subsequent processing. In this paper, we present a model to improve the sub-pixel geometric registration precision of image pairs when there exists locally variant illuminations with arbitrary shape. This model extends on our previous work to include multiple local shading levels of arbitrary shape, where the ill-posed problem is conditioned by constraining the solution to an estimated number of shading levels. The proposed model is solved using least-squares estimation and is cast in an iterative coarse-to-fine framework, which allows a convergence rate that is similar to competing intensity-based image registration approaches. The primary advantage of the proposed approach is the nearly tenfold improvement in sub-pixel precision for the registration when convergence is obtained in this class of technique.

Geometric registration of images with arbitrarily-shaped local intensity variations from shadows

In this paper, we present a registration approach for images having arbitrarily-shaped locally variant illuminations. These variations tend to degrade the performance of geometric registration precision (GRP) and impact subsequent processing. Traditional registration approaches typically use a least-squares estimator that is sensitive to outliers. Instead, we propose using a robust bisquare M-estimator, as it differently penalizes the small and large residuals. The proposed approach shows clear improvements over competing approaches in terms of GRP and illumination correction, using simulated and real image pairs.

Geometric image registration under arbitrarily-shaped locally variant illuminations

In this paper, we extend our previous work on presenting a registration model for images having arbitrarily-shaped locally variant illuminations from shadows to multiple shading levels. These variations tend to degrade the performance of geometric registration and impact subsequent processing. Often, traditional registration models use a least-squares estimator that is sensitive to outliers. Instead, we propose using a robust Huber M-estimator to increase the geometric registration accuracy (GRA). We demonstrate the proposed model and compare it to other models on simulated and real data. This modification shows clear improvements in terms of GRA and illumination correction.

Image registration under local illumination variations using robust bisquare M-estimation

In this paper, a new parallel scheme for the deblocking filter (DF) in high efficiency video coding (HEVC) is proposed. This scheme is based on a parallel-straight processing order that improves the performance of HEVC DF. One of the challenges in HEVC is coding time due to computational complexity. Deblocking in HEVC is responsible for nearly 15% of the time consumed while performing video compression. As such, a parallel-straight processing order is introduced that allows improved concurrency for deblocking multiple horizontal and vertical edges. For our examined 4-core case, the approach achieves full utilization of all cores with fewer number of DF steps (i.e., two edges or more) by 27% compared to recent techniques. A four-core parallel architecture is also proposed. This new parallel scheme is implemented on a graphical processing unit (GPU) rather than a CPU to further speed up coding time. Experimental results demonstrate the ability to achieve decoded frame processing times as low as 5.0 ms for All-Intra filtering and 3.0 ms for Low-Delay filtering, corresponding to speedup factors as high as 12 and 7, respectively, compared to the HEVC reference.

Geometric image registration under locally variant illuminations using Huber M-estimator

In this paper, we present a super-resolution (SR) technique for images having arbitrarily-shaped local illumination changes. These variations tend to degrade the performance of the image registration, and hence impact the SR image reconstruction. Conventional SR techniques focus on enhancing the reconstruction step assuming aligned images. In this paper, we exploit our recent registration approach of images having illumination variations using a robust bisquare M-estimation. Then, we extend a bounded total variation-based approach for upsampling single frames to super-resolving multi-frames in order to reconstruct the unknown high-resolution (HR) frame. The proposed SR technique shows clear improvements over competing techniques in terms of objective metrics using simulated and real image pairs with illumination variations.

High throughput parallel scheme for HEVC deblocking filter

Surface fatigue failure occurs in geared transmission systems due to factors such as high contact stress, and monitoring its progression is vital if the eventual failure of the tooth flank is to be prevented. Techniques involving the analysis of vibration, acoustic emission and oil debris have been used to successfully monitor the progression of the different phases of gear surface fatigue failure. However, during monitoring a suitable assessment method is required to correlate the characteristics of the signal with the condition of the gear surface. The most commonly used such methods are gear flank profile scanning, replica sample analysis and conventional image analysis, with each technique having both advantages and disadvantages. Responding to the demand for effective assessment methods, this study experimentally evaluates the development of micro-pitting in gears using a new image registration technique in an online health monitoring system. Given a set of captured images of gear surface degradation with different exposure times and geometric deformations, an image registration approach is proposed to cope with inter-image illumination changes of arbitrary shape. Then correlation between the resulting aligned images is compared to a reference one before testing obtained. The results validate the systems capabilities to detect early gear defects and reliably identify the gradual development of micro-pitting in gears, so that it could be used in predictive health monitoring (PHM) systems.

Two-step super-resolution technique using bounded total variation and bisquare M-estimator under local illumination changes

In HEVC, deblocking filtering (DF) is responsible for about 20% of the time consumed to perform video compression. In a typical parallel DF scheme, a set of horizontal and vertical edges are processed using deblocking filters. In conventional parallel DF schemes, deblocking filters could be applied to the same edges more than once. Moreover, some edges are assigned to cores to be filtered even though those edges are not designated to be filtered. Accordingly, the used parallel hardware architecture requires more on-chip memory modules. Those challenges negatively affect HEVC performance resulting in an increase in computational complexity. In this paper, an optimized parallel DF scheme is proposed for HEVC using graphical processing units (GPUs). The proposed scheme outperforms competing ones in terms of reducing the decoding time of all frames of video sequences by average speed-up factors of 2.83 and 2.45 using the all-intra and low-delay video coding configuration modes, respectively. The proposal does not change the rate-distortion between the decoded video sequences and their original sequences.

Application of image registration methods in monitoring the progression of surface fatigue failures in geared transmission systems

In this paper, the implementation method for encoding the high resolution videos using high efficiency video coding (HEVC) standard is introduced with a new approach. The HEVC standard, successor to the H.264/AVC standard, is more efficient than the H.264/AVC standard in the encoding high resolution videos. HEVC has been designed to focus on increasing video resolution and increasing the use of parallel processing architectures. Therefor, this approach merging all traditional configuration files used in the encoding process into only one configuration file without removing any parameters used in the traditional methods. Improvements are shown using the proposed approach in terms of encoding time as opposed to the traditional methods by reducing the access time by half which resulting from reducing the data exchange between the configuration files used in this process and without changing the rate-distortion (RD) performance or compression ratio.