Abstract: Deep convolutional neural networks (CNNs) have improved RGB-induced salient object detection due to their superior feature learning. Cluttered backgrounds, low light, and lighting variations make such detections difficult. This paper uses thermal infrared images…
Abstract: Salient object detection has advanced rapidly thanks to deep convolutional neural networks, especially fully convolutional neural networks (FCNs). Due to their heavy model weights, FCNs-based methods cannot generate high-resolution saliency maps. This paper proposes…
Abstract: Understanding the relationship between vascular changes and many diseases requires complex network topology estimation. Ophthalmologists can diagnose and treat eye disease by automatically classifying retinal vascular trees into arteries and veins. Their projective ambiguity…
Abstract: Multi-scale strategies in fully convolutional neural networks (FCNs) improve deep learning-based salient object detection methods. Integrating all scale-specific predictions yields the result. Multi-scale-based methods have several drawbacks: 1) Directly learning discriminative features and filters…
Abstract: We often work with discrete-domain analog image samples due to digital imaging systems. All imaging devices blur the input image before sampling to form output pixels due to physical constraints. This paper reconstructs binary…
Abstract: For compression, we propose an overcomplete multiview imagery representation. We use rate-distortion (R-D) to decompose multiview datasets into additive diffuse and specular content. We use an R-D inspired cost function to drive the compressibility-based…
Abstract: Deep neural networks are now used for image generation and other tasks. GANs can create realistic images for image compression. Like standard compression, generative image perceptual quality should be automatically assessed to monitor and…
Abstract: Due to their speed, deep learning-based deformable image registration methods are appealing. These methods fail to estimate larger displacements in complex deformation fields, requiring a multi-resolution strategy. To solve this problem, we propose progressively…
Abstract: Deep learning with large benchmarks drives object detection development. For real-world applications, collecting fully-annotated data is difficult or expensive, limiting deep neural network power. Humans can use prior knowledge to identify new objects with…
Abstract: Intracoronary imaging helps diagnose coronary disease by showing coronary artery tissue morphologies. Computer-aided disease diagnosis can benefit from intracoronary vessel border detection (VBDI). Existing VDBI methods struggle with vessel-environment variability (high intra- and inter-subject…
Abstract: We present an almost lossless affine 2D image transformation method. We extend the theory of the well-known Chirp-z transform to allow fully affine transformation of general n-dimensional images. We also provide a practical spatial…
Abstract: We propose an ultrasound speckle filtering method that preserves edge features and filters tissue-dependent complex speckle noises. Phase asymmetry (PAS), a phase congruence-based edge significance measure, takes 0 in non-edge smooth regions, 1 at…
WhatsApp us