Abstract:Characteristics of noise (type, statistics, spatial correlation) are nowadays exploited in many image denoising and enhancement methods. However, these characteristics are often unknown, and they have to be extracted from an image at hand. There are many powerful and accurate blind methods for noise variance estimation for the cases of additive and multiplicative noise models. However, more complicated noise models containing a mixture of signal-independent (SI) and signal-dependent (SD) components are often m… Show more
“…In practice, inferring data labels and removing the least critical objects from the LiDAR point cloud before transmission guarantee quasi real-time operations as encoding and decoding times never exceed the LiDAR inter-frame time. Finally, even in the most aggressive, space-saving configuration (i.e., HSC-2), the PSNR is still guaranteed to be above 50 dB (a typical acceptable value for wireless transmission quality loss at which distortions in compressed frames can be hardly noticed [26]) when BPP > 5. This is also validated in Fig.…”
In a fully autonomous driving framework, where vehicles operate without human intervention, information sharing plays a fundamental role. In this context, new network solutions have to be designed to handle the large volumes of data generated by the rich sensor suite of the cars in a reliable and efficient way. Among all the possible sensors, Light Detection and Ranging (LiDAR) can produce an accurate 3D point cloud representation of the surrounding environment, which in turn generates high data rates. For this reason, efficient point cloud compression is paramount to alleviate the burden of data transmission over bandwidth-constrained channels and to facilitate real-time communications. In this paper, we propose a pipeline to efficiently compress LiDAR observations in an automotive scenario. First, we leverage the capabilities of RangeNet++, a Deep Neural Network (DNN) used to semantically infer point labels, to reduce the channel load by selecting the most valuable environmental data to be disseminated. Second, we compress the selected points using Draco, a 3D compression algorithm which is able to obtain compression up to the quantization error. Our experiments, validated on the Semantic KITTI dataset, demonstrate that it is possible to compress and send the information at the frame rate of the LiDAR, thus achieving real-time performance.
“…In practice, inferring data labels and removing the least critical objects from the LiDAR point cloud before transmission guarantee quasi real-time operations as encoding and decoding times never exceed the LiDAR inter-frame time. Finally, even in the most aggressive, space-saving configuration (i.e., HSC-2), the PSNR is still guaranteed to be above 50 dB (a typical acceptable value for wireless transmission quality loss at which distortions in compressed frames can be hardly noticed [26]) when BPP > 5. This is also validated in Fig.…”
In a fully autonomous driving framework, where vehicles operate without human intervention, information sharing plays a fundamental role. In this context, new network solutions have to be designed to handle the large volumes of data generated by the rich sensor suite of the cars in a reliable and efficient way. Among all the possible sensors, Light Detection and Ranging (LiDAR) can produce an accurate 3D point cloud representation of the surrounding environment, which in turn generates high data rates. For this reason, efficient point cloud compression is paramount to alleviate the burden of data transmission over bandwidth-constrained channels and to facilitate real-time communications. In this paper, we propose a pipeline to efficiently compress LiDAR observations in an automotive scenario. First, we leverage the capabilities of RangeNet++, a Deep Neural Network (DNN) used to semantically infer point labels, to reduce the channel load by selecting the most valuable environmental data to be disseminated. Second, we compress the selected points using Draco, a 3D compression algorithm which is able to obtain compression up to the quantization error. Our experiments, validated on the Semantic KITTI dataset, demonstrate that it is possible to compress and send the information at the frame rate of the LiDAR, thus achieving real-time performance.
“…Кроме того, эти помехи могут быть в значительной степени пространственно коррелированными [10 -12]. Несмотря на достаточно большое количество методов автоматического оценивания характеристик помех, разработанных к настоящему моменту [10 -16], среди них нет универсального решения, способного обеспечить приемлемую точность оценивания [17] и быстродействие во всех практических ситуациях, поэтому задача разработки новых методов и усовершенствования уже существующих решений не утрачивает своей актуальности.…”
The article deals with the scatter-plot method of automatic evaluation of mixed noise characteristics in multi-channel images. The aim is to solve the problematic issues associated with the adaptation of this method to hyperspectral images processing. The tasks to be solved are: to investigate the influence of the formation method of the jointly processed multichannel groups of images on the accuracy and stability of the aforementioned method; to formulate the recommendations on the choice of jointly processed images and the method of combination. The applied methods are the following: robust estimation of signal parameters, spectral, correlation and regression analysis. The following results were obtained. Three possible ways of groups forming of three channel images were considered: 1) joint processing of images belonging to adjacent channels; 2) joint processing of images with the highest cross-correlation coefficients; 3) joint processing of images with the lowest cross-correlation coefficients. It was defined that if the cross-correlation coefficients of images in the group are low, and the images have of complex structure, it is possible a significant reduction of the method accuracy, up to a complete loss of its working capacity. The method demonstrates sufficiently high accuracy and stability when the groups are formed of the neighbor channel images or of the images with the highest cross-correlation coefficients, and the values of the estimated noise parameters for these issues have no significant differences. Conclusions. The group formation method significantly affects not only the accuracy, but also the operability of the considered estimation method, and in order to increase the reliability of the method, it is appropriate to form groups of images with rather high levels of inter-channel correlation. However, since the accuracy of the method when groups are formed of the neighbor images and of images with the highest levels of cross-correlation have no significant differences, in order to maintain the high performance of the method, it is recommended to form groups of jointly processed images applying the images obtained in the neighbor spectral zones
“…Meanwhile, HVS-based metric does it better but also not perfectly [35]. Our experience with the metric M HVS PSNR [41] shows that improvement of visual quality becomes noticeable if M HVS PSNR increases by, at least, 0.5…1 dB. To make the final conclusions, consider simulation data for the metric M HVS IPSNR .…”
International audienceTextures or high-detailed structures as well as image object shapes contain information that is widely exploited in pattern recognition and image classification. Noise can deteriorate these features and has to be removed. In this paper, we consider the influence of textural properties on efficiency of image enhancement by noise suppression for the posterior treatment. Among possible variants of denoising, filters based on discrete cosine transform known to be effective in removing additive white Gaussian noise are considered. It is shown that noise removal in texture images using the considered techniques can distort fine texture details. To detect such situations and to avoid texture degradation due to filtering, filtering efficiency predictors, including neural network based predictor, applicable to a wide class of images are proposed. These predictors use simple statistical parameters to estimate performance of the considered filters. Image enhancement is analysed in terms of both standard criteria and metrics of image visual quality for various scenarios of texture roughness and noise characteristics. The discrete cosine transform based filters are compared to several counterparts. Problems of noise removal in texture images are demonstrated for all of them. A special case of spatially correlated noise is considered as well. Potential efficiency of filtering is analysed for both studied noise models. It is shown that studied filters are close to the potential limits
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.