An online visual measurement method for workpiece dimension based on deep learning

MSFCN-multiple supervised fully convolutional networks for the osteosarcoma segmentation of CT images

Compared with traditional High Efficiency Video Coding (HEVC), 3D-HEVC introduces multi-view coding and depth map coding, which leads to significant increase in coding complexity. In this paper, we propose a low complexity intra coding algorithm for depth map based on end-to-end edge detection network. Firstly, we use Holistically Nested Edge Detection (HED) network to determine the edge location of the depth map. Secondly, we use Ostu method to divide the output of the HED into foreground region and background region. Finally, the CU size and the candidate list of intra mode are determined according to the region of coding tree unit (CTU). Experimental results demonstrate that the proposed algorithm can reduce the encoding time by 39.56% on average under negligible degradation of coding performance.

Cellular Learning Automata (CLA) has been used in many fields of image processing such as noise elimination, smoothing, retrieval, fractionated and extraction of the content Characteristics of the images. The edge detection in images and methods if edge detection, have a great role in machine vision and cognizance systems. This method uses operands for analyzing images and digital image processing. Many studios here been conducted till now in edge detection algorithms of various conditions. In this study a new method for edge detection in images with the use of CLA is recommended. The proposed method of edge detection in images was tested with different sizes and the results were compared with Sobel edge detector classic method. The result show that the method based on CLA has a desirable performance in edge detection and compares the images with a more uniformity during a minimum period of time.

Gliomas are the most frequent primary brain tumors in adults. Improved quantification of the various aspects of a glioma requires accurate segmentation of the tumor in magnetic resonance images (MRI). Since the manual segmentation is time-consuming and subject to human error and irreproducibility, automatic segmentation has received a lot of attention in recent years. This paper presents a fully automated segmentation method which is capable of automatic segmentation of brain tumor from multi-modal MRI scans. The proposed method is comprised of a deeply-supervised neural network based on Holistically-Nested Edge Detection (HED) network. The HED method, which is originally developed for the binary classification task of image edge detection, is extended for multiple-class segmentation. The classes of interest include the whole tumor, tumor core, and enhancing tumor. The dataset provided by 2017 Multimodal Brain Tumor Image Segmentation Benchmark (BraTS) challenge is used in this work for training the neural network and performance evaluations. Experiments on BraTS 2017 challenge datasets demonstrate that the method performs well compared to the existing works. The assessments revealed the Dice scores of 0.86, 0.60, and 0.69 for whole tumor, tumor core, and enhancing tumor classes, respectively.

The stability of arc bubble is a crucial indicator of underwater wet welding process. However, limited research exists on detecting arc bubble edges in such environments, and traditional algorithms often produce blurry and discontinuous results. To address these challenges, we propose a novel arc bubble edge detection method based on deep transfer learning for processing underwater wet welding images. The proposed method integrates two training stages: pre-training and fine-tuning. In the pre-training stage, a large source domain dataset is used to train VGG16 as a feature extractor. In the fine-tuning stage, we introduce the Attention-Scale-Semantics (ASS) model, which consists of a Convolutional Block Attention Module (CBAM), a Scale Fusion Module (SCM) and a Semantic Fusion Module (SEM). The ASS model is further trained on a small target domain dataset specific to underwater wet welding to fine-tune the model parameters. The CBAM can adaptively weight the feature maps, focusing on more crucial features to better capture edge information. The SCM training method maximizes feature utilization and simplifies training by combining multi-scale features. Additionally, the skip structure of SEM effectively mitigates semantic loss in the high-level network, enhancing the accuracy of edge detection. On the BSDS500 dataset and a self-constructed underwater wet welding dataset, the ASS model was evaluated against conventional edge detection models—Richer Convolutional Features (RCF), Fully Convolutional Network (FCN), and UNet—as well as state-of-the-art models LDC and TEED. In terms of Mean Absolute Error (MAE), accuracy, and other evaluation metrics, the ASS model consistently outperforms these models, demonstrating edge detection capabilities that are both effective and stable in detecting arc bubble edges in underwater wet welding images.

An edge detection algorithm based on improved Rotating Kernel Transformation, IRKT edge detection method (IRKTE), is proposed in this paper. The algorithm adopt the line detection approach RKT, and defines a new model of edge detection according to the direction difference between edge and smooth regions. Simultaneously, an accurate edge location approach based on edge normal direction is presented to overcome the wide width caused by a large scale kernel in IRKT. Furthermore, the improvement of previous IRKT with weight edge detection (IRKTEW) is proposed to improve the ability to resist the noise effectively. A series of experiments are carried out through the picture libraries with ground truth and the performance is analyzed with ROC curves. The experimental results show that the proposed method can effectively detect the edge under the strong noises, and the performance of edge detection is improved with the proposed approach.

We propose an automated platform for extra-coronary calcification detection on low dose CT scans. We utilize faster regional convolutional neural networks (R-CNN) to directly detect calcifications at the lesion-level without performing vessel extraction. To segment detected calcifications at the voxel-level, we employ holistically nested edge detection (HED). CT scans of 112 vasculitis patients and 3219 images with labeled calcifications were used to develop and evaluate our method. By employing a two-class faster R-CNN, the average precision (AP) increased from 49.2% to 84.4% for calcification detection. In addition, sensitivity of 85.0% at 1 false positive per image was observed. The Dice Similarity Coefficient (DSC) for calcification segmentation using HED (0.83±0.08) was significantly better (p≪0.01) than the traditional threshold-based method (0.59±0.26).

Parapapillary atrophy (PPA) is a kind of atrophic abnormality of eyes, which lies on the retinal fundus adjacent to optic disc (OD). Clinically, PPA is correlated with several pathologies of eyes. For example, the area of PPA has a strong association with the severity of myopia. In this work, we developed a new segmentation algorithm aimed at PPA in retinal images. The frame work was derived from Holistically-Nested Edge Detection (HED), which originally deals with edge detection. Firstly, extracting region of interest (ROI) and normalizing datasets were performed in preprocessing stage. Next, an end-to-end model began to acquire rich hierarchical features with deep supervision on side outputs. We applied a new balance parameter to loss function for realization of the area segmentation. Finally, the holistic image predictions (PPA area segmented) were performed through full convolutional neural layers. The model training on retinal images with annotation from professional ophthalmologist achieves great performance on both adults and pupils datasets. Experiential results present the average F-score of 0.7910 on adults' test data (200 images) and 0.7478 on pupils' (50 images). MIoU of the segmentation reaches 0.7237 of adults and 0.6892 of pupils.

Seismic interpolation via multi-scale HU-Net

Based on the shortcomings of the traditional edge detection algorithm easily lead to the brink of fracture and discontinuous, a new improved Pulse-coupled neural network (PCNN) is proposed to image segmentation and edge detection. Using Gaussian filter pre-processing, gray image is segmented into binary image. Based on this binary image, the edge detection is accomplished using improved PCNN, which can increase contrast of image and enhance the edge of image. Compared with correlative result using Canny operator and others in the existing references, simulation results show that our methods have both preferable results of gray image edge detection and fine applicability.

Edge detection plays a crucial role in image processing and computer vision, and is widely used in tasks such as object recognition and image segmentation. Traditional edge detection algorithms perform well in many applications, but there are still some shortcomings in terms of real-time performance and processing efficiency. To address this issue, a highly efficient image edge detection model combining Sobel algorithm and field programmable gate array technology was proposed. YCbCr color space conversion was performed on the image, then Sobel operator was utilized to calculate the image gradient, and adaptive thresholding method was applied to determine the edges. Finally, the model was implemented and optimized on a field programmable gate array. The experimental results showed that when the dataset size was 1000, the information retention rate of the proposed image preprocessing model was 0.89, and the structural information loss was 0.05. When the data volume was 100, the accuracy of the proposed image edge detection model was 0.90, and the root mean square error value was 0.16. The research results indicate that the proposed image edge detection model based on field programmable gate arrays has significant advantages in edge detection performance and processing efficiency. The model has high accuracy and speed in image edge recognition, which can provide certain guidance for research in the field of image edge detection

Edge detection is a dynamic area of research and it gives support to higher level image analysis. Edge detection is a method of image processing used to make points in a digital image with discontinuities, sharp modifiers in the image brightness. Edge detection methods used in this work are Sobel, Prewitt, Roberts, Laplacian of Gaussian and Canny. The performance evaluation measures such as accuracy and specificity calculated from the Berkeley Segmentation dataset. Edge detection is used for object detection which offers various applications like medical field, industrial field, biometrics etc.

Edge detection is regularly used as a fundamental operation for correctly identifying and measuring some required features. In a grey-level image, an area where the grey-level value moves from a high value to a low value or vice versa is considered an edge. Edges are indicative of a boundary between an object and a background or between two objects. Consequently, edge detection in earth sciences is an important tool for locating geological features and determining their shapes and sizes. Edge detection usually forms a part of the geophysical interpretation or inversion procedure. Seismic tomography is a straightforward field of applying edge detection because the tomogram can be directly considered as an image. In the tomographic reconstruction of seismic travel time data, care must be taken to keep the propagation of data errors to the model space under control. The noise - especially the outliers in the data sets - can cause appreciable distortions in the tomographic imaging. To reduce the noise sensitivity well-developed tomography algorithms can be used. On the other hand, the quality of the tomogram can further be improved by using image processing tools. This is especially important in edge detection, as it is extremely sensitive to noise. In the paper, we present two ways to find robust edge detection. At first, remaining in the framework of traditional image processing a robust Cauchy–Steiner filter is used to improve the quality of edge detection in tomographic images. In the second part of the paper Deep Learning algorithm developed for edge detection is shown and investigating its noise sensitivity the robustness of the method is demonstrated.

Several problems are encountered in modeling and processing of many life science applications such like tissue engineering, bioacoustics, and food processing detection. Recently, various image processing techniques have been developed in an attempt to solve these problems. Due to the rapid advance and high number of researches included in such field, there is always a room for improvement. In this work, advanced image processing approaches are presented and applied to different life science applications. The first approach utilizes image processing in the generation of micro-scaffolds for tissue engineering modeling. Furthermore, a spectrogram enhancement approach is developed to improve the pattern detection of weevils existence in grain storage, based on their detected and identified sounds. Finally, a novel edge detection method applied on intramuscular fat recognition is presented based on the energy and skewness as two smoothed versions of the meat image. Results, in comparison to experimental data, offer a maximum error of 7.81% in the specified detection process. 1. Material and Method 1.1 Generation of micro-porous approach The difficulty of modeling such complex micro-structures always lies in retrieving the geometry into the solver as edges. Edge detection algorithms are based upon highlighting discontinuities in images and were first developed for satellite images. The technique has become well-known and was further developed and widely used for other image processing applications. During the conventional detection procedures, usually all the image size is processed, which leads to redundant calculations of areas containing no edge information. In this study a redundant avoiding algorithm is developed which searches for the threshold value in the image and then propagates around it in a 2-D lattice in eight directions figure 1.1, until an edge is hit. Figure 1.1 The scanning procedure for gradient changes propagates in an eight directional lattice until an edge is hit 1.2 Spectrogram enhancement in bioacoustics This approach is based on detection of the sound pattern edges in the crest factor image, as shown in figure 1.2 and 1.3, which behaves as a smoothed version of the spectrogram image and avoids the use of edge detectors prior smoothing filters and their scaling constraints. Figure 1.2 Spectrogram with 40 dB limited dynamic range Figure 1.3 Crest factor distributions (left), the scaled crest factor image of the spectrogram (right). 1.3 Edge detection for intramuscular fat recognition A novel edge detection method applied on intramuscular fat is presented based on the energy and skewness as two smoothed versions of the meat image, as shown in the flow chart given in figure 1.4. Figure 1.4 flow chart edge detection 2. Results 2.1.Generation of micro-porous Approach The conventional approach for edge detection is based upon either thresholding the first derivative (gradient) of the image or locating zero crossings of its second derivative (Laplacian).The sample used, figure 2.1, for this simulation was of size 3 mm, 350 μm and 350 μm in z-direction, x-direction and y-direction, respectively. The geometry was uploaded on the solver by collective segmenting and edge detection techniques. The method has proven time saving and as well enables the possibility of parallel processing of the image, as each threshold value can be scanned and dealt with independently. After the predictions of the edges, the 2-d binary edge images are assembled forming a 3-D binary file of the scaffolds figure 2.1. Figure 2.1 From the micro-CT images 430 hundred 2-D binary edge images are assembled forming a 3-D binary file of the scaffolds 2.2.Spectrogram enhancement in bioacoustics The spectrogram enhancement approach is compared to other enhancement approaches, including multiband spectral subtraction and wavelet packet decomposition, for different structure bioacoustics calls including bats (ultrasound range) and birds (acoustic range), with additive white Gaussian noise added across a range of signal to noise ratios. Results measured subjectively and objectively, indicate that the proposed method is promising for efficiently spectrogram enhancing while preserving its temporal and spectral accuracy, as shown in figure 2.2. Figure 2.2 (a) One of Sitta Canadensis bird calls at SNR = 20 dB, and its enhanced spectrogram by (b) BP, (c) MBSS, (d) WF, (e) WPD, and (f) the proposed method approaches. 2.3.Edge detection for intramuscular fat recognition The generation of micro-scaffolds approach is applied on a sample meat image, which is scanned line by line and every noticeable variance in the color values or the grey levels is captured and the local thresholds values are estimated according to the highest drop gradient value. The local threshold values are averaged to estimate the global threshold for the whole image, as shown in figure 2.1. Figure 2.3 an image for a meat with fat (a), the x y scan of the image which does not show all the fat details (b) and the lattice scan of the image showing more details of the fat, as circled by red (c). For intramuscular fat, the developed edge detection approach is used to predict the fat content in marbled meat images, as shown in figure 2.3. Figure 2.4 (a) original marbled meat images, (b) edges images using the proposed method (c) detected fat regions, and (d) the percentage of fat content Conclusion As observed in the different approaches there is always a room for improvement that minimize the error observed in the specified detection process. The methods have proven time saving and as well enables the possibility of parallel processing of the image that leads to a more efficient performance highly desirable for industrial applications.

Image edge detection based tool condition monitoring with morphological component analysis