We propose a deep bilinear model for blind image quality assessment (BIQA) that handles both synthetic and authentic distortions. Our consists of two convolutional neural networks (CNN), each which specializes in one distortion scenario. For distortions, we pre-train CNN to classify type level, where enjoy large-scale training data. adopt pre-trained classification. The features from the CNNs are pooled bilinearly into unified representation final prediction. then fine-tune entire on target...

An auto fabric defect detection system via computer vision is used to replace manual inspection. In this paper, we propose a hardware accelerated algorithm based on small-scale over-completed dictionary (SSOCD) sparse coding (SC) method, which realized parallel platform (TMS320C6678). order reduce computation, the image patches projections in training SSOCD are taken as features and proposed more robust, exhibit obvious advantages results computational cost. Furthermore, introduce ratio...

No-reference image quality assessment (NR-IQA) aims to measure the without reference image. However, contrast distortion has been overlooked in current research of NR-IQA. In this paper, we propose a very simple but effective metric for predicting contrast-altered images based on fact that high-contrast is often more similar its enhanced Specifically, first generate an through histogram equalization. We then calculate similarity original and one by using structural-similarity index (SSIM) as...

Abstract. RANSAC algorithm is a robust method for model estimation. It widely used in the extraction of geometry primitives and 3D reconstruction. However, there has been relatively little comprehensive evaluation RANSAC-based approach plane extraction. In order to provide reference improving quality on roof facets or segmentation, this paper focuses analysis classical algorithm. Airborne LIDAR data from test Area 1 2 Vaihingen (German) used. 33 buildings (4 with flat roofs 29 slope roofs)...

This study proposed a tracking algorithm based on oversaturated sub‐region classifiers. Compared with the compressive (CT), tracker can reduce influence of occlusion and improve stability accuracy result. First, target region is divided into sub‐regions randomly, then some classifiers are adaptively selected their confidence. Each classifier find candidate position. At last, place maximum positions’ distribution density final location target. Experiments different videos demonstrate that has...

This article presents a new method based on video images to measure the speed of vehicles. The working principle is taking two instantaneous tandem photographs. Then location vehicle and CCD, with reference advance calibration lines, we can obtain vehicle. Further, this paper analyses measurement errors, discusses how choose picture filming location. Simulations verify effectiveness given method.

Fine-grained image recognition, a computer vision task filled with challenges due to its imperceptible inter-class variance and large intra-class variance, has been drawing increasing attention. While manual annotation can be utilized effectively enhance performance in this task, it is extremely time-consuming expensive. Recently, Convolutional Neural Networks (CNN) achieved state-of-the-art classification. We propose fine-grained recognition framework by exploiting CNN as the raw feature...

We propose an end-to-end deep fusion-based approach to enhance the quality of images acquired in weak illumination environment. The proposed fusion network (DFN), without estimating explicitly, uses a convolutional neural (CNN) generate confidence maps as spatial weighting factors fuse created by multiple base image enhancement techniques that complement each other content-dependent manner. Our tests on both synthetic and real weakly illuminated show DFN delivers superior performance terms...

Deep convolutional neural networks have recently achieved great success on image aesthetics assessment task. In this paper, we propose an efficient method which takes the global, local and scene-aware information of images into consideration exploits composite features extracted from corresponding pretrained deep learning models to classify derived with support vector machine. Contrary popular methods that require fine-tuning or training a new model scratch, our training-free directly...

Heap overflow is one of the most widely exploited vulnerabilities, with a large number heap instances reported every year. It important to decide whether crash caused by can be turned into an exploit. Efficient and effective assessment exploitability crashes facilitates identify severe vulnerabilities thus prioritize resources. In this paper, we propose first metrics assess based on both attack aspect feasibility aspect. We further present HCSIFTER, novel solution automatically under our...

Successfully detecting, analyzing, and reasoning about collective anomalies is important for many real-life application domains (e.g., intrusion detection, fraud analysis, software security). The primary challenges to achieving this goal include the overwhelming number of low-risk events their multimodal relationships, diversity by various data anomaly types, difficulty in incorporating domain knowledge experts. In paper, we propose novel concept faceted High-Order Correlation Graph (HOCG)....

Image aesthetics assessment is an interesting yet challenging topic which can be applied on numerous scenarios such as high quality image retrieval or recommendation systems. We propose a hierarchical features fusion aesthetic (HFFAA) model for this task. HFFAA two-stream convolutional neural network (CNN) composed of two branches with heterogeneous and complementary perceptual abilities. learns the mapping from deep representation into their ground-truth labels (good bad) in end-to-end...

Abstract. Three-dimensional modelling plays a vital role in indoor 3D tracking, navigation, guidance and emergency evacuation. Reconstruction of models is still problematic, part, because spaces provide challenges less-documented than their outdoor counterparts. Challenges include obstacles curtailing image point cloud capture, restricted accessibility wide array objects, each with unique semantics. environments can be achieved through photogrammetric approach, e.g. by using frames, aligned...

This paper introduces an efficient feature learning framework via sparse coding for no-reference image quality assessment. The important part of the proposed is based on extraction from a representation matrix, which computed using algorithm. Image patches extracted salient regions unlabeled images are used to learn dictionary coding. ℓ1-norm taken as penalty term in process and computing representation. A detector adopts together with max-pooling results matrix output features obtain...

Fingerprint image quality assessment is important because the good performance of minutiae-based matching algorithm heavily dependent on fingerprint images with high quality. Many efforts have been made in existing methods, but most methods either use full or local areas and involve subjective judgments. Unlike previous proposed method considers both global assessments. Local feature vectors are extracted from block for hierarchical clustering, results used as target outputs back-propagation...

Although diagnostic expert systems using a knowledge base which models decision-making of traditional experts can provide important information to non-experts, they tend duplicate the errors made by experts. Decision-Theoretic Model (DTM) is therefore very useful in system since prev ent from incorrect reasoning under uncertainty. For system, corresponding DTM and arithmetic are studied sequential decision-theoretic model based on Bayesian Network given. In model, alternative features...

Most vision-language (VL) trackers rely on coarse-grained information from sentences to achieve multi-modal alignment. However, this is insufficient for accurately describing the target in each frame due inherent ambiguity, summarization, and invariance of sentences, thereby making alignment challenging. This paper introduces TTCTrack, a novel VL tracker that employs textual token classification address challenge. Specifically, we exploit cross-relations classify tokens into various types...