Search

Deep-learning-based target recognition in synthetic aperture radar (SAR) images has been actively studied in recent years. However, it is very costly to collect large numbers of labeled SAR images, especially measured SAR target images of various classes, to train high-performance classification networks. To solve the problem of insufficient SAR data, electromagnetic computational tools have often been developed and used to synthesize the measured SAR target images from data modeling. However, despite the use of sophisticated SAR image modeling, there is a large domain gap between synthetic SAR images and measured images such that networks trained with synthetic SAR images tend to show poor classification performance when tested on measured SAR target images. In this article, we propose a novel transformer-based synthetic-to-measured SAR target image translation network, referred to as SAR-SMT Net, to bridge the gap between synthetic and measured SAR target images. SAR-SMT Net takes synthetic SAR target images as input and estimates the latent representational features of their corresponding measured SAR images to faithfully adjust the global context and scattering characteristics of the input synthetic SAR target images to the corresponding measured SAR values. In addition, we propose five challenging experimental scenarios that can validate SAR image translation performance outcomes. Experimentally, SAR-SMT Net as proposed here outperforms previous state-of-the-art methods in the experiment scenarios, demonstrating feasible generalization ability when used to translate synthetic SAR target images into their corresponding measured SAR target images with a high level of fidelity, even for unseen target classes at unseen azimuth angles.

Satellite synthetic aperture radar (SAR) images are immensely valuable because they can be obtained regardless of weather and time conditions. However, SAR images have fatal noise and less contextual information, thus making them harder and less interpretable. Thus, translation of SAR to electro-optical (EO) images is highly required for easier interpretation. In this article, we propose a novel coarse-to-fine context-aware SAR-to-EO translation (CFCA-SET) framework and a misalignment-resistant (MR) loss for the misaligned pairs of SAR-EO images. With our auxiliary learning of SAR-to-near-infrared translation, CFCA-SET consists of a two-stage training: 1) the low-resolution SAR-to-EO translation is learned in the coarse stage via a local self-attention module that helps diminish the SAR noise and 2) the resulting output is used as guidance in the fine stage to generate the SAR colorization of high resolution. Our proposed auxiliary learning of SAR-to-NIR translation can successfully lead CFCA-SET to learn distinguishable characteristics of various SAR objects with less confusion in a context-aware manner. To handle the inevitable misalignment problem between SAR and EO images, we newly designed an MR loss function. Extensive experimental results show that our CFCA-SET can generate more recognizable and understandable EO-like images compared to other methods in terms of nine image quality metrics. Our CFCA-SET surpasses the state-of-the-art methods for two (QXS and CASET) datasets with the improvements: PSNR (3.6%, 29%), ERGAS (7.4%, 30%), SSIM (15%, 15%), SAM (21%, 38%), $D_{S}$ (16%, 13%), QNR (1.5%, 3.1%), CHD (18%, 12%), LPIPS (4.2%, 8%), and FID (9.0%, 33%).

Video quality assessment is an important tool of guaranteeing video services in a required level of quality. Although subjective quality assessment is more reliable due to the reflection of Human Visual System (HVS) than objective quality assessment, it is a time-consuming and very expensive approach, and is not appropriate for real-time applications. Therefore, much research has been made for objective video quality assessment instead of subjective video quality assessment. Among three kinds of objective assessment approaches which are full-reference, reduced-reference and no-reference methods, no-reference method has drawn much attention because it does not require any reference. The encoding parameters are good features to use for no-reference model because the encoded bitstreams carry plenty of information about the video contents and it is easy to extract some coding parameters to assess visual quality. In this paper, we propose a no-reference quality metric using two kinds of coding parameters in H.264/AVC: quantization and block mode parameters. These parameters are extracted and computed from H.264/AVC bitstreams, without relying on pixel domain processing. We design a linear quality metric composed of these two parameters. The weight values of the parameters are estimated using linear regression with the results of subjective quality assessment which are obtained based on the DSIS (Double Stimulus Impairment Scale) method of ITU-R BT.500-11.

Object-based coding is one of the distinct features of the MPEG-4 standard. Hence, video object segmentation is an indispensable process for MPEG-4 coding. A method of localising moving objects for video object segmentation is introduced which utilises a statistical hypothesis test without the need for true statistical parameters to be known a priori.

In this paper, a novel distortion model based on a mixture of Laplacian distributions is presented for the transform coefficients of predicted residues in quadtree coding. The mixture Laplacian distribution is made on the coding structure with different quadtree coding unit (CU) depth. Moreover, for intra-coded CU, the distortion model is asymptotically simplified based on the signal characteristics of the transform coefficient. The proposed mixture model of multiple Laplacian distributions is tested for the High Efficiency Video Coding (HEVC) Test Model (HM) with quadtreestructured Coding Unit (CU) and Transform Unit (TU). The experimental results show that the proposed model achieves more accurate results of distortion estimation than the single probability models.

We introduce a novel model capturing user preference using the Bayesian approach for recommending users’ preferred multimedia content. Unlike other preference models, our method traces the trend of a user preference in time. It allows us to do online learning so we do not need exhaustive data collection. The tracing of the trend can be done by modifying the frequency of attributes in order to force the old preference to be correlated with the current preference under the assumption that the current preference is correlated with the near future preference. The modification is done by partitioning usage history data into smaller sets in a time axis and then weighting the frequencies of attributes to be computed from the partitioned sets of the usage history data in order to differently reflect their significance on predicting the future preference. In the experimental section, the learning and reasoning on user preference in genres are performed by the proposed method with a set of real TV viewers’ watching history data collected from many real households. The reasoning performance by the proposed method is also compared with that by a typical method without training in order to show the superiority of our proposed method. © 2005 SPIE and IS&T.