Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784151
E. Palma, I. Tabus
In this paper we propose a near-lossless encoder for sensor images acquired by plenoptic cameras, and we investigate its usage for encoding in an archive all information needed for reconstructing high quality versions of the light field (LF) array of views(AoV). The near-lossless encoding of the plenoptic camera sensor image is realized by a modified version of the recently published sparse relevant regressors and contexts (SRRC) encoder. The lossy reconstruction is obtained in two nested loops: the outer one operates over the sensor image patches (each patch corresponding to a microlens image), and the inner loop operates over the pixels in the patch. In the latter, we enforce the SRRC predictors to use the already reconstructed lossy version of the sensor image. Then, we examine the usage of the near-lossless SRRC (NL-SRRC) codec as a building block for an archiving scheme including all information needed for running the plenoptic processing pipeline and obtaining the LF-AoV. Finally, we replace in the archiving scheme the NL-SRRC codec with other state of the art lossy codecs and compare the results, which show that NL-SRRC based archiving scheme achieves better performance for the range of high bitrates.
{"title":"Near-Lossless Coding of Plenoptic Camera Sensor Images for Archiving Light Field Array of Views","authors":"E. Palma, I. Tabus","doi":"10.1109/IPTA54936.2022.9784151","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784151","url":null,"abstract":"In this paper we propose a near-lossless encoder for sensor images acquired by plenoptic cameras, and we investigate its usage for encoding in an archive all information needed for reconstructing high quality versions of the light field (LF) array of views(AoV). The near-lossless encoding of the plenoptic camera sensor image is realized by a modified version of the recently published sparse relevant regressors and contexts (SRRC) encoder. The lossy reconstruction is obtained in two nested loops: the outer one operates over the sensor image patches (each patch corresponding to a microlens image), and the inner loop operates over the pixels in the patch. In the latter, we enforce the SRRC predictors to use the already reconstructed lossy version of the sensor image. Then, we examine the usage of the near-lossless SRRC (NL-SRRC) codec as a building block for an archiving scheme including all information needed for running the plenoptic processing pipeline and obtaining the LF-AoV. Finally, we replace in the archiving scheme the NL-SRRC codec with other state of the art lossy codecs and compare the results, which show that NL-SRRC based archiving scheme achieves better performance for the range of high bitrates.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"147 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124430325","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784129
A. Goumeidane, D. Ziou, Nafaa Nacereddine
This paper presents a new elliptical structure detection method, combining the advantages of the multiscale Hessian, and the scale space Radon transform (SSRT) for an ellipse. The advantage of the former is twofold: highlighting the lines defining the ellipses present in the image and reducing the search space for these ellipses in the SSRT space, which will discard the false SSRT maxima. The subsequent application of the SSRT permits, in turn, to alleviate the computation load and to obtain, moreover, a good detection of thick ellipses when they are not threadlike. Experiments carried out on synthetic and real images have shown good detection of thick ellipses, with low computational overhead compared to the Elliptical Radon transform.
{"title":"Scale Space Radon Transform for Non Overlapping Thick Ellipses Detection","authors":"A. Goumeidane, D. Ziou, Nafaa Nacereddine","doi":"10.1109/IPTA54936.2022.9784129","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784129","url":null,"abstract":"This paper presents a new elliptical structure detection method, combining the advantages of the multiscale Hessian, and the scale space Radon transform (SSRT) for an ellipse. The advantage of the former is twofold: highlighting the lines defining the ellipses present in the image and reducing the search space for these ellipses in the SSRT space, which will discard the false SSRT maxima. The subsequent application of the SSRT permits, in turn, to alleviate the computation load and to obtain, moreover, a good detection of thick ellipses when they are not threadlike. Experiments carried out on synthetic and real images have shown good detection of thick ellipses, with low computational overhead compared to the Elliptical Radon transform.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121063571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784143
P. Coupeau, Jean-Baptiste Fasquel, M. Dinomais
This paper addresses the fundamental task of semantic image segmentation by exploiting structural information (spatial relationships between image regions). To perform such task, we propose to combine a deep neural network (CNN) with inexact “many-to-one-or-none” graph matching where graphs encode efficiently class probabilities a nd structural information related to regions segmented by the CNN. In order to achieve node classification, a basic 2 -layer graph neural network (GNN) based on the edge-conditioned convolution operator (ECConv), managing both node and edge attributes, is considered. Prelim-inary experiments are performed on both a synthetic dataset and a public dataset of face images (FASSEG). Our approach is shown to be resilient to small training datasets that often limit the performance of deep learning thanks to a preprocessing task of graph coarsening. Results show that the proposal reaches a perfect accuracy on synthetic dataset and improves performance of the CNN by 6% (bounding box dice index) on FASSEG. Moreover, it enhances by 27% the initial Hausdorff distance (i.e. with CNN only) using the entire training dataset and by 41% with only 75% of training samples.
{"title":"On the relevance of edge-conditioned convolution for GNN-based semantic image segmentation using spatial relationships","authors":"P. Coupeau, Jean-Baptiste Fasquel, M. Dinomais","doi":"10.1109/IPTA54936.2022.9784143","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784143","url":null,"abstract":"This paper addresses the fundamental task of semantic image segmentation by exploiting structural information (spatial relationships between image regions). To perform such task, we propose to combine a deep neural network (CNN) with inexact “many-to-one-or-none” graph matching where graphs encode efficiently class probabilities a nd structural information related to regions segmented by the CNN. In order to achieve node classification, a basic 2 -layer graph neural network (GNN) based on the edge-conditioned convolution operator (ECConv), managing both node and edge attributes, is considered. Prelim-inary experiments are performed on both a synthetic dataset and a public dataset of face images (FASSEG). Our approach is shown to be resilient to small training datasets that often limit the performance of deep learning thanks to a preprocessing task of graph coarsening. Results show that the proposal reaches a perfect accuracy on synthetic dataset and improves performance of the CNN by 6% (bounding box dice index) on FASSEG. Moreover, it enhances by 27% the initial Hausdorff distance (i.e. with CNN only) using the entire training dataset and by 41% with only 75% of training samples.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128180085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/ipta54936.2022.9784150
{"title":"Special Session 3: Visual Computing in Digital Humanities","authors":"","doi":"10.1109/ipta54936.2022.9784150","DOIUrl":"https://doi.org/10.1109/ipta54936.2022.9784150","url":null,"abstract":"","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114230944","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784116
Bianca Janse van Rensburg, W. Puech, Jean-Pierre Pedeboy
3D objects play an essential part in many differ-ent domains. They are considered to be important assets and therefore need to be secured. Sometimes, certain users may have the right to access only a part of the 3D object, for example the shape of the object, but not the content. Recently, selective encryption schemes have been proposed in order to allow this. 3D objects are often stored on the cloud or need to be rendered in real time, and therefore compression is important. In this paper, we propose a selective crypto-compression method for 3D objects. This is based on the Draco compression method, which was developed by Google for 3D objects. Our method is format compliant, and to the best of our knowledge, we are the first to propose a selective cry to-compression method for 3D objects.
{"title":"Draco-Based Selective Crypto-Compression Method of 3D objects","authors":"Bianca Janse van Rensburg, W. Puech, Jean-Pierre Pedeboy","doi":"10.1109/IPTA54936.2022.9784116","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784116","url":null,"abstract":"3D objects play an essential part in many differ-ent domains. They are considered to be important assets and therefore need to be secured. Sometimes, certain users may have the right to access only a part of the 3D object, for example the shape of the object, but not the content. Recently, selective encryption schemes have been proposed in order to allow this. 3D objects are often stored on the cloud or need to be rendered in real time, and therefore compression is important. In this paper, we propose a selective crypto-compression method for 3D objects. This is based on the Draco compression method, which was developed by Google for 3D objects. Our method is format compliant, and to the best of our knowledge, we are the first to propose a selective cry to-compression method for 3D objects.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"192 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114838654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784123
Kallil M. C. Zielinski, L. C. Ribas, Leonardo F. S. Scabini, O. Bruno
Since the 1960s, texture has become one of the most-studied visual attribute of images for analysis and classification tasks. Among many different approaches such as statistical, spectral, structural and model-based, there are also methods that rely on analyzing the image complexity and learning techniques. These recent approaches are receiving attention for its promising results in the past few years. This paper proposes a method that combines complex networks and randomized neural networks. In the proposed approach, the texture image is modeled as a complex network, and the information measures obtained from the topological properties of the network are then used to train the RNN in order to learn a representation of the modeled image. Our proposal has proven to perform well in comparison to other literature approaches in two different texture databases. Our method also achieved a high performance in a very challenging biological problem of plant species recognition. Thus, the method is a promising option for different tasks of image analysis.
{"title":"Complex Texture Features Learned by Applying Randomized Neural Network on Graphs","authors":"Kallil M. C. Zielinski, L. C. Ribas, Leonardo F. S. Scabini, O. Bruno","doi":"10.1109/IPTA54936.2022.9784123","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784123","url":null,"abstract":"Since the 1960s, texture has become one of the most-studied visual attribute of images for analysis and classification tasks. Among many different approaches such as statistical, spectral, structural and model-based, there are also methods that rely on analyzing the image complexity and learning techniques. These recent approaches are receiving attention for its promising results in the past few years. This paper proposes a method that combines complex networks and randomized neural networks. In the proposed approach, the texture image is modeled as a complex network, and the information measures obtained from the topological properties of the network are then used to train the RNN in order to learn a representation of the modeled image. Our proposal has proven to perform well in comparison to other literature approaches in two different texture databases. Our method also achieved a high performance in a very challenging biological problem of plant species recognition. Thus, the method is a promising option for different tasks of image analysis.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"131 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114362360","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784114
L. C. Ribas, Leonardo F. S. Scabini, O. Bruno
Fish otolith recognition is an essential task to study the evolution and food chains in paleontological and ecological sciences. One of the approaches to this problem is to automatically analyze the shape of otolith contour present in images. In this paper, we explore a state-of-the-art shape analysis method called “angular descriptors of complex networks (ADCN)” applied to the classification of otolith images for fish species recognition. The ADCN method models the otolith contour as a graph, or complex network, and computes angular properties from its connections for shape characterization. The ADCN method is evaluated in an otolith image dataset composed of 14 fish species from three families. Up to 95.71% of accuracy is achieved, which surpasses other literature methods and confirms that the ADCN method can be an important tool for such biological problems.
{"title":"A complex network approach for fish species recognition based on otolith shape","authors":"L. C. Ribas, Leonardo F. S. Scabini, O. Bruno","doi":"10.1109/IPTA54936.2022.9784114","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784114","url":null,"abstract":"Fish otolith recognition is an essential task to study the evolution and food chains in paleontological and ecological sciences. One of the approaches to this problem is to automatically analyze the shape of otolith contour present in images. In this paper, we explore a state-of-the-art shape analysis method called “angular descriptors of complex networks (ADCN)” applied to the classification of otolith images for fish species recognition. The ADCN method models the otolith contour as a graph, or complex network, and computes angular properties from its connections for shape characterization. The ADCN method is evaluated in an otolith image dataset composed of 14 fish species from three families. Up to 95.71% of accuracy is achieved, which surpasses other literature methods and confirms that the ADCN method can be an important tool for such biological problems.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130534150","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784118
Yuhao Luo, Hengjing Zhang, Hengchang Liu
Intimate contact recognition has gained more attention in academia field in recent years due to the outbreak of Covid-19. However, state of the art solutions suffer from either inefficient accuracy or high cost. In this paper, we propose a novel method for COVID-19 intimate contact recognition in public spaces through video camera networks (CCTV). This method leverages distance detection and re-Identification algorithms, so pedestrians in close contact are re-identified, their identity information is obtained and stored in a database to realize contact tracing. We compare different social distance detection algorithms and the Faster-RCNN model outperforms other al-ternatives in terms of running speed. We also evaluate our Re-Identification model on two types of indicators in the PETS2009 dataset: mAP reaches 85.1%; rank-1, rank-5, and rank-10 reach 97.8%, 98.9%, and 98.9%, respectively. Experimental results demonstrate that our solution can be effectively applied in public places to realize fast and accurate automatic contact tracing.
{"title":"Towards Fast and Accurate Intimate Contact Recognition through Video Analysis","authors":"Yuhao Luo, Hengjing Zhang, Hengchang Liu","doi":"10.1109/IPTA54936.2022.9784118","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784118","url":null,"abstract":"Intimate contact recognition has gained more attention in academia field in recent years due to the outbreak of Covid-19. However, state of the art solutions suffer from either inefficient accuracy or high cost. In this paper, we propose a novel method for COVID-19 intimate contact recognition in public spaces through video camera networks (CCTV). This method leverages distance detection and re-Identification algorithms, so pedestrians in close contact are re-identified, their identity information is obtained and stored in a database to realize contact tracing. We compare different social distance detection algorithms and the Faster-RCNN model outperforms other al-ternatives in terms of running speed. We also evaluate our Re-Identification model on two types of indicators in the PETS2009 dataset: mAP reaches 85.1%; rank-1, rank-5, and rank-10 reach 97.8%, 98.9%, and 98.9%, respectively. Experimental results demonstrate that our solution can be effectively applied in public places to realize fast and accurate automatic contact tracing.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114606186","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-19DOI: 10.1109/IPTA54936.2022.9784128
Wael Saideni, F. Courrèges, D. Helbert, J. Cances
This paper presents a novel reconstruction algorithm for video Snapshot Compressive Imaging (SCI). Inspired by recent research works on Transformers and Self-Attention mechanism in computer vision, we propose the first video SCI reconstruction algorithm built upon Transformers to capture long-range spatio-temporal dependencies enabling the deep learning of feature maps. Our approach is based on a Spatiotempo-ral Convolutional Multi-head Attention (ST-ConvMHA) which enable to exploit the spatial and temporal information of the video scenes instead of using fully-connected attention layers. To evaluate the performances of our approach, we train our algorithm on DAVIS2017 dataset and we test the trained models on six benchmark datasets. The obtained results in terms of PSNR, SSIM and especially reconstruction time prove the ability of using our reconstruction approach for real-time applications. We truly believe that our research will motivate future works for more video reconstruction approaches.
{"title":"End-to-End Video Snapshot Compressive Imaging using Video Transformers","authors":"Wael Saideni, F. Courrèges, D. Helbert, J. Cances","doi":"10.1109/IPTA54936.2022.9784128","DOIUrl":"https://doi.org/10.1109/IPTA54936.2022.9784128","url":null,"abstract":"This paper presents a novel reconstruction algorithm for video Snapshot Compressive Imaging (SCI). Inspired by recent research works on Transformers and Self-Attention mechanism in computer vision, we propose the first video SCI reconstruction algorithm built upon Transformers to capture long-range spatio-temporal dependencies enabling the deep learning of feature maps. Our approach is based on a Spatiotempo-ral Convolutional Multi-head Attention (ST-ConvMHA) which enable to exploit the spatial and temporal information of the video scenes instead of using fully-connected attention layers. To evaluate the performances of our approach, we train our algorithm on DAVIS2017 dataset and we test the trained models on six benchmark datasets. The obtained results in terms of PSNR, SSIM and especially reconstruction time prove the ability of using our reconstruction approach for real-time applications. We truly believe that our research will motivate future works for more video reconstruction approaches.","PeriodicalId":381729,"journal":{"name":"2022 Eleventh International Conference on Image Processing Theory, Tools and Applications (IPTA)","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-04-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122072721","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：