Showing papers on "Image conversion published in 2021"

Image synthesis of monoenergetic CT image in dual-energy CT using kilovoltage CT with deep convolutional generative adversarial networks

Abstract: Purpose To synthesize a dual-energy computed tomography (DECT) image from an equivalent kilovoltage computed tomography (kV-CT) image using a deep convolutional adversarial network. Methods A total of 18,084 images of 28 patients are categorized into training and test datasets. Monoenergetic CT images at 40, 70, and 140 keV and equivalent kV-CT images at 120 kVp are reconstructed via DECT and are defined as the reference images. An image prediction framework is created to generate monoenergetic computed tomography (CT) images from kV-CT images. The accuracy of the images generated by the CNN model is determined by evaluating the mean absolute error (MAE), mean square error (MSE), relative root mean square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and mutual information between the synthesized and reference monochromatic CT images. Moreover, the pixel values between the synthetic and reference images are measured and compared using a manually drawn region of interest (ROI). Results The difference in the monoenergetic CT numbers of the ROIs between the synthetic and reference monoenergetic CT images is within the standard deviation values. The MAE, MSE, RMSE, and SSIM are the smallest for the image conversion of 120 kVp to 140 keV. The PSNR is the smallest and the MI is the largest for the synthetic 70 keV image. Conclusions The proposed model can act as a suitable alternative to the existing methods for the reconstruction of monoenergetic CT images in DECT from single-energy CT images.

Identification of Freshness of Marine Fish Based on Image of Hue Saturation Value and Morphology

Abstract: Euthynus is one of the fish that is widely consumed for the enjoyment of the people of Indonesia or abroad, because of its very soft quality, easy to obtain, and contains a lot of essential protein amino acids that are good for the body. This research aims to identify the freshness of the fish purchased based on the eyes and fish gills. The initial process of identifying the freshness of fish uses several methods. Image input process through image object taking using a cell phone camera. The image object is used to determine the value of the RGB image object. RGB color extraction clarifies the value obtained from the image object before proceeding to the next process. Image resize is the process of cutting the image on the desired object part. Image conversion using the HSV method was used to determine the freshness of fish in the gills. The Local Binary Pattern method is used to determine the freshness of the fisheye. The next step is to refine the RGB image into Morphology. The KNN (K-Nearest Neighbor Method) method is used to group objects based on learning data closest to the object. The journal analysis results on the comparison of methods, after 45 trials for each method, found that the Hue Saturation Value method obtained the highest success by 90% and for the texture value obtained 85% success.

IICNet: A Generic Framework for Reversible Image Conversion

Abstract: Reversible image conversion (RIC) aims to build a reversible transformation between specific visual content (e.g., short videos) and an embedding image, where the original content can be restored from the embedding when necessary. This work develops Invertible Image Conversion Net (IICNet) as a generic solution to various RIC tasks due to its strong capacity and task-independent design. Unlike previous encoder-decoder based methods, IICNet maintains a highly invertible structure based on invertible neural networks (INNs) to better preserve the information during conversion. We use a relation module and a channel squeeze layer to improve the INN nonlinearity to extract cross-image relations and the network flexibility, respectively. Experimental results demonstrate that IICNet outperforms the specifically-designed methods on existing RIC tasks and can generalize well to various newly-explored tasks. With our generic IICNet, we no longer need to hand-engineer task-specific embedding networks for rapidly occurring visual content. Our source codes are available at: this https URL.

Innovative research on the visual performance of image two-dimensional animation film based on deep neural network

Abstract: Deep neural network (DNN) has gone through more than forty years from the simulation stage to the embryonic stage, to the conception and the initial formation of the theory, until further simple applications. Firstly, deep convolution neural network is used to extract features, and then the extracted features are coded by Gaussian aggregation. Finally, the encoded features are input into the full connection layer to classify the image. The experiment in this paper focuses on the two-dimensional facial expression animation technology under DNN. This experiment mainly introduces the two important steps of image conversion, spatial mapping and resampling technology, as well as the detailed definition of the MPEG-4 standard. By using the feature points obtained in the experiment as the feature points set in the face definition parameters, in the deformation algorithm based on Delaunay triangulation, the inverse mapping technology and the quadratic linear interpolation technology are combined with the face and combined with facial animation. The parameter step can realize the conversion of facial expressions, and the generated facial expressions are more natural and delicate.The experimental data show that the facial animation parameters (FAP) as a complete set of basic facial movements can recognize the most subtle facial expressions under DNN. Change, different FAP combinations can form different facial expressions. The experimental results show that the 76 facial feature points located in the experiment have been tested for grid generation. When the segmentation threshold is larger, the merge method is closer to the sub-point insertion method; otherwise, it is closer to the divide and conquer method. When the threshold is 20, the algorithm belongs to the divide and conquer method; when the threshold is 80, the algorithm belongs to the point insertion method. Through research, this article finds that the innovation of DNN on the visual performance of animated films can provide more help for the production of sophisticated animation works, and can continuously improve the visual performance and cultural connotation of animated films.

A Comparison between AttnGAN and DF GAN: Text to Image Synthesis

Abstract: Nowadays conversion from text to high resolution image is a challenging task due to its wide variety of application area. For text to image conversion almost all systems use Generative Adversarial Networks as the basic part of the system and GAN guarantees semantic consistency between the text input and the generated image output. In this paper we are comparing two algorithms that is used for generating image from text. The first algorithm is the AttnGAN and the second one is the DF-GAN. AttnGAN builds on top of StackGAN by using attention network which allows it to capture word level information along with the broader sentence level information. The second algorithm is the DF-GAN, which uses single generator and discriminator model to synthesize high resolution images and also uses Matching-Aware Gradient Penalty (MA-GP) to get real images with real description. The model contains a Deep text-image Fusion Block (DFBlock) to generate image features from text. Both algorithms work efficiently for image generation from text but DF-GAN generates the perfect output than AttnGAN. The AttnGAN always focus on the textual part to generate output image but DF-GAN also focuses on background of image.

Real-time Lane Detection Based on Image Edge Feature and Hough Transform

Abstract: Background and Objectives: Lane detection systems are an important part of safe and secure driving by alerting the driver in the event of deviations from the main lane. Lane detection can also save the lifes of car occupants if they deviate from the road due to driver distraction.Methods: In this paper, a real-time and illumination invariant lane detection method on high-speed video images is presented in three steps. In the first step, the necessary preprocessing including noise removal, image conversion from RGB colour to grey and the binarizing input image is done. Then, a polygon area as the region of interest is chosen in front of the vehicle to increase the processing speed. Finally, edges of the image in the region of interest are obtained with edge detection algorithm and then lanes on both sides of the vehicle are identified by using the Hough transform.Results: The implementation of the proposed method was performed on the IROADS database. The proposed method works well under different daylight conditions, such as sunny, snowy or rainy days and inside the tunnels. Implementation results show that the proposed algorithm has an average processing time of 28 milliseconds per frame and detection accuracy of 96.78%.Conclusion: In this paper a straightforward method to identify road lines using the edge feature is described on high-speed video images. ======================================================================================================Copyrights©2021 The author(s). This is an open access article distributed under the terms of the Creative Commons Attribution (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, as long as the original authors and source are cited. No permission is required from the authors or the publishers.======================================================================================================

Implementation of HSV- based Thresholding Method for Iris Detection

Abstract: Image thresholding is one of the most frequently used methods in image processing to perform digital image processing. Image thresholding has a technique that can separate the image object from its background. This is a technique that is quite good and effective for segmenting love. In this study, the threshold method used will be combined with the HSV mode for color detection. The threshold method will separate the object and the image background, while HSV will help improve the segmentation results based on the Hue, Saturation, Value values to be able to detect objects more accurately. Segmentation is carried out using the original input image without pre-processing or direct segmentation. As we know that in digital image processing, there are steps that are usually done to get a good input image, namely pre-processing. In this pre-processing stage, processes such as image conversion and image intensity changes are carried out so that the input image is better. Therefore, even though the input image is used without going through the pre-processing stage, the object can be segmented properly based on the color type of the object. The results of this segmentation can later be used for recognition and identification of image objects. The results of the test method for object segmentation achieved a color similarity level of 25%, with an accuracy rate of 75% in detecting uniform color objects. So that this method can be one of the most effective methods in segmenting image objects without pre-processing or direct thresholding

Research on feature point generation and matching method optimization in image matching algorithm

Abstract: Image matching is a basic problem in image processing and pattern recognition. It is used to calculate the visual similarity between images taken in the same scene with different sensors, different perspectives or at different times. In addition to image adjustment, it is an indispensable step in image analysis and digital photogrammetry. It is also important for applications such as automatic navigation, image processing, medical image analysis, and motion estimation. The current image adjustment technology can be divided into three categories: domain-based image conversion technology, gray-scale-based technology, and performance-based technology. Among them, the feature-based matching algorithm directly matches the features of the image, so it greatly improves the calculation efficiency and is easy to adapt to complex image transformations, such as geometric distortion, different resolutions, and image transformations at different angles. Image matching refers to the process of using effective matching algorithms to find the same or similar cue points for two or more image data. In applications such as medical image processing and analysis, remote sensing monitoring, weapon movement and image processing, image matching technology is an important step. Images have strong structural features, such as corners, edges, statistics, and textures. These functions play an important role in image matching and scanning technology. The key to many image matching problems depends on selection, detection and expression. For different image matching problems, different functions are selected, and the matching results may be very different.

Method and apparatus for image conversion

Abstract: A method and apparatus for image conversion according to an embodiment of the present disclosure includes receiving original image data, separating the original image data into a front view image and a back view image for performing 3D conversion processing of the original image data, and generating a converted 3D image by restoring a background space between the front view image and the back view image using a 3D conversion processing neural network. The 3D conversion processing neural network according to the present disclosure may be a deep neural network generated by machine learning, and input and output of images may be performed in an Internet of things environment using a 5G network.

An automatic 3d image reconstruction process from real-world 2d images

Abstract: The invention relates to a method of converting a two-dimensional (2D) image into a three-dimensional (3D) image using an image conversion system having at least one processor and at least one memory, the method comprising: extracting a 2D RGB (Red, Green, Blue) object image attribute from a 2D object image; uploading the extracted 2D RGB object image attribute to a cloud computing system (or service), wherein developed algorithms are located; calculating a 3D mesh object image attribute based on the uploaded and extracted 2D RGB object image attribute; texturing the estimated 3D mesh object from the calculated 3D mesh object image attribute; and displaying the textured 3D mesh object on a display device.

Image conversion device and image decoding device

Abstract: The present invention provides an image conversion device for dividing an image into blocks of a prescribed size and converting an encoded picture including a slice that includes an integral number of rectangular areas into which the blocks are combined into one or more block lines. If a slice boundary and a rectangular area boundary are the same when determining whether the encoded picture is to apply a loop filter, the device converts the encoded picture so as to reconstruct the relationship of the slice and the rectangular area included in the encoded picture, and if the slice boundary and the rectangular area boundary are not the same when determining whether the encoded picture is to apply a loop filter, the device converts the encoded picture so as not to change the relationship of the slice and the rectangular area included in the encoded picture.

Driving controller, display device and driving method including a lower frequency mode and an image transition mode

Abstract: A driving controller includes an image conversion circuit configured to convert an image signal to an image data signal including active data and blank data, a still image determination circuit configured to output a flag signal of an active level when the image signal is a still image, an operation mode determination circuit configured to output an operation mode signal indicating a low frequency mode when the flag signal is the active level, and to output an operation mode signal indicating an image transition mode when the flag signal is changed from the active level and an inactive level, and a blank voltage determination circuit configured to output a blank voltage signal corresponding to a first gray scale during the low frequency mode, and a blank voltage signal corresponding to a second gray scale during the transition mode, wherein the blank data corresponds to the blank voltage signal.

Detection of leukemia in microscopic images using digital image processing tool in matlab

Abstract: This research describes leukemia detection techniques. Leukemia is a type of cancer affecting blood-forming tissues of the spleen, bone marrow, and lymph system. Its symptoms are headache, fatigue, weakness, mouth sores, sternal tenderness, gingival hyperplasia, minimal hepatosplenomegaly, and lymphadenopathy. To avoid the rapid progression of immature hematopoietic cells, it is very important to detect leukemia at an early stage. Existing methods of diagnosis are –medical history and physical examination, complete blood count, bone marrow aspiration, cytogenetic analysis, and immunohistochemistry. These methods are time-consuming, not cost-effective, and totally dependent on medical personnel [1]. To get rid of these problems a digital image processing tool in MATLAB is used. So many methods of image processing are used for the identification of red blood cells and immature or mature white blood cells and different diseases like anemia, malaria, etc. can also be diagnosed by using digital image processing methods[2]. The main objective of this research work is to detect leukemia cells and count their area, perimeter, roundness, and standard deviation. For detection of immature blast cells, a number of methods are used like histogram equalization, linear contrast stretching, ostu thresholding, some morphological techniques like area opening, area closing, dilation, and erosion they also help to identify the acute or chronic stages of leukemia. Previously microscopic images were inspected by hematologists and it’s really very time-consuming, but the technique now is used is digital image processing by which neither patient nor doctor has to wait for his/her report. Blood is taken from the patient’s arm and slides are prepared then microscopic slides images are captured and uploaded in digital image processing tool MATLAB software then it gives accurate results that whether the patient has acute or chronic leukemia or he/she does not have leukemia [3]. Microscopic images are processed using image processing techniques such as image enhancement, segmentation, feature extraction, and classification. The initially RGB image is read then RGB image to gray image conversion after the image is converted into binary image and segmentation is also done which will segregate white blood cells from all other blood components i.e. erythrocytes and platelets. The image conversion into the binary image area opening is done. Then hole filling and after that boundary is detected for each and every cell, for this purpose Sobel operator is used and it also differentiates between overlapped and non-overlapped cell, with the value of major axis, minor axis, convex hull, and standard deviation.[4] This whole process is performed using a digital image processing tool in MATLAB. ‘Region props’ properties are used to find area, roundness, centroid, the major and minor axis of cells. In this research shape-based features are used because it is very easy for the detection of white mature and immature cells and the shape of a cell. The early and fast detection of leukemia is very important because it helps aid in providing treatment better. The proposed result gives promising results for varying image quality and even so any images can be detected. This research will be helpful for those who cannot afford fees for detection of leukemia and for those also who have less time and due to this process treatment of leukemia can be done earlier.[5]

Apparatus for imaging power data and method therefor

Abstract: The present invention discloses an apparatus for imaging power data, and a method therefor. The apparatus for imaging power data according to the present invention is characterized by comprising: a data input unit that receives time series three-phase power data; an image conversion unit that converts the domain of the three-phase power data input from the data input unit, expresses the converted domain as a grayscale image, and creates a color image by allocating an RGB channel to the grayscale image; and an output unit that outputs the image generated by the image conversion unit.

System and method for simulating visual field observed upon insertion of artificial lens

Abstract: Disclosed is a device for image-converting and simulating a visual field observed upon insertion of an artificial lens. A device for image-converting and simulating a visual field observed upon insertion of an artificial lens may include: a collection unit which receives each of a first image acquired through a monofocal artificial lens and a second image acquired through a multifocal artificial lens; a learning unit which analyzes the correlation between the first image and the second image through machine-learning; an input unit to which a third image acquired through a monofocal lens is input; and an image conversion unit which converts the third image into a virtual fourth image obtained with the multifocal artificial lens on the basis of the correlation analyzed by the learning unit.

Color to Grayscale Image Conversion Using High Dimensional Model Representation

Abstract: Color images are converted to grayscale when texture information is needed more than color information when it is desired to reproduce or when it is used for artistic purposes.The methods used to convert color images to grayscale should both produce perceptually acceptable grayscale results and retain as much information as possible about the original color image. In this study, an alternative color-to-grayscale image conversion algorithm has been developed for decolarization using the High Dimensional Model Representation (HDMR) method. In order to evaluate the efficiency of the proposed algorithm, normalized cross correlation, color contrast preservation ratio, E-score and color content fidelity ratio have been used within the scope of standard objective measures.

3 3 Method and Apparatus for Generating 3D Fluorescent Label Image of Label-Free using 3D Refractive Index Tomography and Deep Learning

Abstract: Disclosed are a method and apparatus for generating a three-dimensional (3-D) molecular image based on a label-free method using a 3-D refractive index image and deep learning. The apparatus for generating a 3-D molecular image based on a label-free method using a 3-D refractive index image and deep learning may include a 3-D refractive index cell image measurement unit configured to measure a 3-D refractive index image of a cell to be monitored and a 3-D refractive index and fluorescence molecule staining image conversion unit configured to input a measured value of the 3-D refractive index image to a deep learning algorithm and to output a 3-D fluorescence molecule staining cell image of the cell.

Image processing device, image processing method, and program

Abstract: A memory control unit 12 divides input image data into: a first block having the pixel data of a prescribed number of pixels for a plurality of lines in a line direction; a second block having the pixel data of a prescribed number of pixels for a plurality of lines, including some lines of the first block, that follow the first block in the line direction; and a third block having the pixel data of a prescribed number of pixels for a plurality of lines, including lines in the first block that are different from lines included in the second block, that follow in the line direction. The memory control unit 12 causes the divided blocks to be stored in a memory unit 13. An arithmetic processing unit 15 calculates an interpolation position that is a position before image conversion corresponding to a pixel position after image conversion. The memory control unit 12 reads the first to third blocks of pixel data, including the pixel data of pixels surrounding the interpolation position, from the memory unit 13. An interpolation processing unit 14 generates pixel data at the interpolation position through interpolation processing using the surrounding pixels that have been read out. The processing speed of image processing can be improved.

Segmentasi Citra Menggunakan Metode Watershed Transform Berdasarkan Image Enhancement Dalam Mendeteksi Embrio Telur

Abstract: Image processing can be applied in the detection of egg embryos. The egg embryos detection is processed using a segmentation process. The segmentation divides the image according to the area that is divided. This process requires improvement of the image that is processed to obtain optimal results. This study will analyze the detection of egg embryos based on image processing with image enhancement and the concept of segmentation using the watershed method. Image enhancement in preprocessing in image improvement uses a combination of Contrast Limited Adaptive Histogram Equalization (CLAHE) and Histogram Equalization (HE) methods. The grayscale egg image is corrected using the CLAHE method, and the results are reprocessed using HE. The image improvement results show that the CLAHE-HE combination method gives a clear picture of the object area of the egg image that has an embryo. The segmentation process using image conversion to black and white image and watershed segmentation can clearly show the object of a chicken egg that has an embryo. The results of segmentation can divide the area of the egg having embryos in a real and accurate way with a percentage \approx 98\%.

Method and apparatus for generating 3-d molecular image based on label-free method using 3-d refractive index image and deep learning

Abstract: Disclosed are a method and apparatus for generating a three-dimensional (3-D) molecular image based on a label-free method using a 3-D refractive index image and deep learning. The apparatus for generating a 3-D molecular image based on a label-free method using a 3-D refractive index image and deep learning may include a 3-D refractive index cell image measurement unit configured to measure a 3-D refractive index image of a cell to be monitored and a 3-D refractive index and fluorescence molecule staining image conversion unit configured to input a measured value of the 3-D refractive index image to a deep learning algorithm and to output a 3-D fluorescence molecule staining cell image of the cell.

A Novel Method of PEM Fuel Cell Fault Diagnosis Based on Signal-to-Image Conversion

Abstract: This paper proposes an image processing-based approach for the fault diagnosis of polymer electrolyte membrane (PEM) fuel cells As more abundant information is contained in the image than that that in 1D signal, features representing PEM fuel cell faults could be better highlighted with the image Experimental data from a PEM fuel cell system at different states, including flooding and dehydration scenarios, is used to validate the proposed method By converting the PEM fuel cell voltage signal into a 2D grey image, several features are extracted from the image, their performance in discriminating different PEM fuel cell states is investigated, and two optimal features are determined for fault diagnosis Moreover, the diagnostic performance of optimal features from grey image is compared with features from PEM fuel cell voltage Results demonstrate that better diagnostic performance could be obtained with the proposed method

Obstacle recognition device

Abstract: Provided is an obstacle recognition device capable of verifying a camera posture parameter to be used for image conversion and capable of defining a relationship between a device camera and the current environment, thereby achieving improved obstacle recognition reliability by simultaneously improving both the accuracy of obstacle detection and accuracy in calculating the distance to the detected object even if there is a change in environmental conditions in which a vehicle is traveling. This invention comprises: acquiring a first difference (∆A) relating to the posture of a camera based on road surface shapes at different time instants; acquiring a second difference (∆R) relating to the amount of movement of characteristic points within images captured at said different time instants; and verifying, by employing the first difference (∆A) and the second difference (∆R), a camera posture parameter to be used for image conversion for calculating a differential image.

Image conversion model training method, heterogeneous face recognition method, device and apparatus

Abstract: An image conversion model training method, a heterogeneous face recognition method, device and apparatus, wherein the heterogeneous face recognition method comprises the following steps: obtaining a face sketch map to be recognized, and cutting the face sketch map to be recognized to obtain a face region sketch map (S31); inputting the face region sketch into a pre-trained image conversion model for processing to generate a second face composite image (S32); performing feature extraction on the second face composite image to obtain a feature vector of the second face composite image (S33); and matching the feature vector of the second face composite image with the feature vectors of a plurality of real face images stored in a database to obtain a face recognition result (S34). In this way, the face region is cut from the face sketch map to be recognized, so as to reduce the interference of the background area on face recognition, and meanwhile, the preset image conversion model is adopted for image conversion, and the heterogeneous face recognition is converted into homogeneous face recognition, so as to improve the accuracy of heterogeneous face recognition.

Application of Internet of Things Technology in Artificial Intelligence Image Detection System

Abstract: In order to solve the disadvantages of traditional methods, this paper studies the application of Internet of Things (IOT) technology in artificial intelligence (AI) image detection system. System usually adopts intelligent artificial pixel characteristic collection technology, point by point on the image feature extraction, using Internet of rich data resources and processing capacity, the analysis of characteristics of image pixels on feedback, feedback signals by AI image synthesis module, the signal to do image conversion processing complete results of the analysis and the output image detection, AI image detection system design; The simulation test proves that the recognition accuracy rate is up to 94.71%, which has provided a new design idea for the research and development of image detection system.

Image display system, image transmission device, image display device, and image display system connection method

Abstract: One embodiment of this invention is an image display system comprising an image transmission device and a plurality of image display devices which are communicably connected via a network, the image display system being characterized by: the image transmission device including an image conversion unit that performs, on the basis of a detection image which is an image displayed by and input from an image display device to be connected, from among the plurality of image display devices, a conversion to first image data representing features of the image, and a communication unit that transmits the first image data to the plurality of image display devices; and the plurality of image display devices each including an image conversion unit that converts a displayed image to second image data, an image data determination unit that compares the first image data received from the image transmission device and the second image data to determine if the first image data and the second image data obtained by converting the displayed image match, and a communication unit that, in accordance with the determination results determined by the image data determination unit, transmits to the image transmission device connection permission information for connecting to the image transmission device.

Image recognition device, image recognition method, and image recognition program

Abstract: Provided are an image recognition device, an image recognition method, and an image recognition program that require few teaching images when learning is carried out, and that have a high recognition accuracy and have a low calculation load, due to the use of state information of an imaging device at the time of imaging. The image recognition device includes: a recognition unit which is a machine learning model trained on the basis of teaching images captured by an imaging device in a first state expressed by first state information; an input image acquisition unit that acquires an input image captured by the imaging device in a second state, and second state information expressing the second state; and an image conversion unit that, on the basis of the second state information, generates a converted image by applying, to the input image, conversion processing so as to approach the image captured in the first state.