AN EFFECTIVE IMAGE RETRIEVAL METHOD BASED ON KERNEL DENSITY ESTIMATION OF COLLAGE ERROR AND MOMENT INVARIANTS

In this paper, we propose a new method that combines collage error in fractal domain and Hu moment invariants for image retrieval with a statistical method - variable bandwidth Kernel Density Estimation (KDE). The proposed method is called CHK (KDE of Collage error and Hu moment) and it is tested on the Vistex texture database with 640 natural images. Experimental results show that the Average Retrieval Rate (ARR) can reach into 78.18%, which demonstrates that the proposed method performs better than the one with parameters respectively as well as the commonly used histogram method both on retrieval rate and retrieval time.

Shift Invariance Level Comparison of Several Contourlet Transforms and Their Texture Image Retrieval Systems

In this paper, we proposed a metric to measure the shift invariance of the three different contourlet transforms. And then, using the same structure texture image retrieval system which use subband coefficients energy, standard deviation and kurtosis features with Canberra distance, we gave a comparison of their texture description abilities. Experimental results show that contourlet-2.3 texture image retrieval system has almost retrieval rates with non-sub sampled contourlet system;the two systems have better retrieval results than the original contourlet retrieval system. On the other hand, for the relatively lower redundancy, we recommend using contourlet- 2.3 as texture description transform.

An Effective Image Retrieval Mechanism Using Family-based Spatial Consistency Filtration with Object Region

How to construct an appropriate spatial consistent measurement is the key to improving image retrieval performance. To address this problem, this paper introduces a novel image retrieval mechanism based on the family filtration in object region. First, we supply an object region by selecting a rectangle in a query image such that system returns a ranked list of images that contain the same object, retrieved from the corpus based on 100 images, as a result of the first rank. To further improve retrieval performance, we add an efficient spatial consistency stage, which is named family-based spatial consistency filtration, to re-rank the results returned by the first rank. We elaborate the performance of the retrieval system by some experiments on the dataset selected from the key frames of ＂TREC Video Retrieval Evaluation 2005 （TRECVID2005）＂. The results of experiments show that the retrieval mechanism proposed by us has vast major effect on the retrieval quality. The paper also verifies the stability of the retrieval mechanism by increasing the number of images from 100 to 2000 and realizes generalized retrieval with the object outside the dataset.

A flower image retrieval method based on ROI feature

Flower image retrieval is a very important step for computer-aided plant species recognition. In this paper, we propose an efficient segmentation method based on color clustering and domain knowledge to extract flower regions from flower images. For flower retrieval, we use the color histogram of a flower region to characterize the color features of flower and two shape-based features sets, Centroid-Contour Distance (CCD) and Angle Code Histogram (ACH), to characterize the shape features of a flower contour. Experimental results showed that our flower region extraction method based on color clustering and domain knowledge can produce accurate flower regions. Flower retrieval results on a database of 885 flower images collected from 14 plant species showed that our Region-of-Interest (ROI) based retrieval approach using both color and shape features can perform better than a method based on the global color histogram proposed by Swain and Ballard (1991) and a method based on domain knowledge-driven segmentation and color names proposed by Das et al.(1999).

A PERSONALIZED IMAGE RETRIEVAL BASED ON VISUAL PERCEPTION

A new scheme named personalized image retrieval technique based on visual perception is proposed in this letter, whose motive is to narrow the semantic gap by directly perceiving user's visual information. It uses visual attention model to segment image regions and eye-tracking technique to record fixations. Visual perception is obtained by analyzing the fixations in regions to measure gaze interests. Integrating visual perception into attention model is to detect the Regions Of Interest (ROIs), whose features are extracted and analyzed, then feedback interests to optimize the results and construct user profiles.

Image Retrieval Based on Objects’ Spatial Orientation Relationships

A new concept of characteristic scanning radial (CSR) is proposed for thesegmented image on the basis of two shape-specific points of its shape-objects. Subsequently, twocharacteristic attribute sequences (CAS) of relative distance and relative direction are derived torepresent the spatial orientation relationships among objects of the image. A novel image retrievalalgorithm is presented using these two CASs. The proposed retrieval approach not only satisfies thetransformational invariance, butalso attains the quantitative comparison of matching. Experimentsidentify the effectiveness and efficiency of the algorithm adequately.

Content-Based Image Retrieval：Near Tolerance Rough Set Approach

The problem considered in this paper is how to detect the degree of similarity in the content of digital images useful in image retrieval,i.e.,to what extent is the content of a query image similar to content of other images.The solution to this problem results from the detection of subsets that are rough sets contained in covers of digital images determined by perceptual tolerance relations(PTRs).Such relations are defined within the context of perceptual representative spaces that hearken back to work by J.H.Poincare on representative spaces as models of physical continua.Classes determined by a PTR provide content useful in content-based image retrieval(CBIR).In addition,tolerance classes provide a means of determining when subsets of image covers are tolerance rough sets(TRSs).It is the nearness of TRSs present in image tolerance spaces that provide a promising approach to CBIR,especially in cases such as satellite images or aircraft identification where there are subtle differences between pairs of digital images,making it difficult to quantify the similarities between such images.The contribution of this article is the introduction of the nearness of tolerance rough sets as an effective means of measuring digital image similarities and,as a significant consequence,successfully carrying out CBIR.

Application of fuzzy logic in content-based image retrieval

We propose a fuzzy logic-based image retrieval system, in which the image similarity can be inferred in a nonlinear manner as human thinking. In the fuzzy inference process, weight assignments of multi-image features were resolved impliedly. Each fuzzy rule was embedded into the subjectivity of human perception of image contents. A color histogram called the average area histogram is proposed to represent the color features. Experimental results show the efficiency and feasibility of the proposed algorithms.

Content-Based Image Retrieval with Feature Extraction and Rotation Invariance

Over recent years, Convolutional Neural Networks (CNN) has improved performance on practically every image-based task, including Content-Based Image Retrieval (CBIR). Nevertheless, since features of CNN have altered orientation, training a CBIR system to detect and correct the angle is complex. While it is possible to construct rotation-invariant features by hand, retrieval accuracy will be low because hand engineering only creates low-level features, while deep learning methods build high-level and low-level features simultaneously. This paper presents a novel approach that combines a deep learning orientation angle detection model with the CBIR feature extraction model to correct the rotation angle of any image. This offers a unique construction of a rotation-invariant CBIR system that handles the CNN features that are not rotation invariant. This research also proposes a further study on how a rotation-invariant deep CBIR can recover images from the dataset in real-time. The final results of this system show significant improvement as compared to a default CNN feature extraction model without the OAD.

HIERARCHICAL CONTENT-BASED IMAGE RETRIEVAL

A hierarchical structure method of content based image retrieval was proposed. During image preprocessing stage three semi automatic algorithms were used to extract image regions. String matching can be used to reduce image searching range. Smallest enclose rectangle(SER) and Hausdorff distance under region normalization were used to measure the similarity between trademark images while keeping invariant under transform(translation, rotation and scale) and noise tolerant. The experiment results show its efficiency.

Automated neurosurgical video segmentation and retrieval system

Medical video repositories play important roles for many health-related issues such as medical imaging, medical research and education, medical diagnostics and training of medical professionals. Due to the increasing availability of the digital video data, indexing, annotating and the retrieval of the information are crucial. Since performing these processes are both computationally expensive and time consuming, automated systems are needed. In this paper, we present a medical video segmentation and retrieval research initiative. We describe the key components of the system including video segmentation engine, image retrieval engine and image quality assessment module. The aim of this research is to provide an online tool for indexing, browsing and retrieving the neurosurgical videotapes. This tool will allow people to retrieve the specific information in a long video tape they are interested in instead of looking through the entire content.

Semi-supervised learning based probabilistic latent semantic analysis for automatic image annotation

In recent years,multimedia annotation problem has been attracting significant research attention in multimedia and computer vision areas,especially for automatic image annotation,whose purpose is to provide an efficient and effective searching environment for users to query their images more easily. In this paper,a semi-supervised learning based probabilistic latent semantic analysis( PLSA) model for automatic image annotation is presenred. Since it's often hard to obtain or create labeled images in large quantities while unlabeled ones are easier to collect,a transductive support vector machine( TSVM) is exploited to enhance the quality of the training image data. Then,different image features with different magnitudes will result in different performance for automatic image annotation. To this end,a Gaussian normalization method is utilized to normalize different features extracted from effective image regions segmented by the normalized cuts algorithm so as to reserve the intrinsic content of images as complete as possible. Finally,a PLSA model with asymmetric modalities is constructed based on the expectation maximization( EM) algorithm to predict a candidate set of annotations with confidence scores. Extensive experiments on the general-purpose Corel5k dataset demonstrate that the proposed model can significantly improve performance of traditional PLSA for the task of automatic image annotation.

全局感知与稀疏特征关联图像级弱监督病理图像分割

弱监督语义分割方法可以节省大量的人工标注成本,在病理全切片图像(WSI)的分析中有着广泛应用。针对弱监督多实例学习(MIL)方法在病理图像分析中存在的像素实例相互独立缺乏依赖关系,分割结果局部不一致和图像级标签监督信息不充分的问题,该文提出一种全局感知与稀疏特征关联图像级弱监督的端到端多实例学习方法(DASMob-MIL)。首先,为克服像素实例之间的独立性,使用局部感知网络提取特征以建立局部像素依赖,并级联交叉注意力模块构建全局信息感知分支(GIPB)以建立全局像素依赖关系。其次,引入像素自适应细化模块(PAR),通过多尺度邻域局部稀疏特征之间的相似性构建亲和核,解决了弱监督语义分割结果局部不一致的问题。最后,设计深度关联监督模块(DAS),通过对多阶段特征图生成的分割图进行加权融合,并使用权重因子关联损失函数以优化训练过程,以降低弱监督图像级标签监督信息不充分的影响。DASMob-MIL模型在自建的结直肠癌数据集YN-CRC和公共弱监督组织病理学图像数据集LUAD-HistoSeg-BC上与其他模型相比展示出了先进的分割性能,模型权重仅为14 MB,在YN-CRC数据集上F1 Score达到了89.5%,比先进的多层伪监督(MLPS)模型提高了3%。实验结果表明,DASMob-MIL仅使用图像级标签实现了像素级的分割,有效改善了弱监督组织病理学图像的分割性能。

边缘增强结合多尺度信息融合的皮肤病变分割

针对皮肤病灶大小不一、与周围皮肤对比度低、边界模糊不规则、存在伪影及毛发干扰等问题,本文提出一种边缘增强结合多尺度信息融合的皮肤病变分割算法,该算法由编码器、多尺度感知模块、边缘增强模块、轻量化解码器组成.首先在编码器中构建Transformer模块以提取全局信息,使用卷积操作以提取局部信息;其次设计多尺度感知模块,采用密集连接结构的门控空洞卷积金字塔模块来融合多尺度特征;接着构建边缘增强模块,利用深层特征促进对边缘特征的探索以更好的保留细节和边缘信息;最后设计轻量化解码器,采用CARAFE轻量化算子进行上采样,在使用相对较少参数的情况下保持较高的分割精度.在公开数据集ISIC2016和ISIC2018上做对比实验,结果表明,本文算法分割精度高于其他流行算法.

基于光照感知和特征增强的可见光-热红外图像语义分割

在智能光电设备中,基于人工智能的可见光-热红外(Red Greed Blue-Thermal, RGB-T)图像语义分割任务可以广泛应用于自动驾驶、无人机航拍、视频监控等。图像的光照信息能在一定程度上反映场景中图像局部区域信息的可靠性,利用光照先验信息有助于进一步提高语义分割的性能。基于此,提出一种基于光照感知和特征增强的RGB-T图像语义分割模型,通过挖掘光照先验信息并结合注意力机制,引导网络在多模态图像特征融合过程中更加关注可靠信息的提取,同时抑制干扰信息的引入。实验在MFNet数据集上与最新的12种方法进行了比较,相比于性能第2的模型,mAcc提高了5.4%,mIoU提高了1.0%。所提网络模型能够获得更准确的分割结果,并通过定性定量实验验证所提模型及各个模块的有效性。

基于深度旋转不变特征图哈希的遥感图像检索

哈希技术采用紧致哈希码表示数据,因其高效性被广泛应用于大规模遥感图像检索任务。受卫星观测影响,同一地物在不同遥感图像中呈现不同角度,导致检索性能下降。为解决该问题,该文提出深度特征图旋转不变哈希方法(DRIFMH),包括特征提取、哈希量化2个模块。特征提取模块对特征图进行不同角度旋转,提出特征一致性损失,使不同旋转角度的图像特征保持一致,克服旋转带来的不利影响。哈希量化模块对图像特征进行二值量化,生成哈希码,引入分类交叉熵损失,提升哈希码的鉴别能力。该文选取经典遥感图像数据集AID、UCMD作为实验数据集,将DRIFMH与多个哈希方法进行实验对比,结果表明DRIFMH能够生成旋转不变的遥感图像特征,提升大规模遥感图像检索性能。

基于多尺度区域可靠性感知的半监督肺肿瘤分割

提出一种基于多尺度下一致性和区域可靠性感知的半监督学习方法,在使用少量标注数据的情况下结合未标记的数据来实现高性能的肺部肿瘤分割任务。首先,提出一种多尺度一致性均值教师框架,构建多尺度一致性损失并约束教师学生网络中多个尺度上的输出保持一致,使模型学习到更丰富的一致性知识。此外,提出一种区域可靠性感知方法使一致性学习之间的知识交换更加有效,使模型从无标注的数据中学习到更有效且可靠的知识。本文方法在医学图像分割十项全能比赛肺肿瘤数据集上进行充分的评估,与当前先进的半监督学习方法比较有更优越的性能,验证本文方法的有效性。

基于卷积神经网络的嵌入式视觉感知交互系统设计与实现

针对语音智能助理无法提供周围环境的视觉感知问题,该文设计并实现一个视觉感知交互系统。该系统的基本结构由语音识别、语音播放、图像采集、中央处理控制等模块组成,具备语音交互、目标检测等功能。该系统设计选择语音识别专用芯片、利用卷积神经网络技术完成识别,采用基于图分割截块的算法进行目标分割。实验结果表明,系统性能良好,能够实现对周围环境的视觉感知并进行语音交互。

游径感知视角下传统村落文旅空间景观质量分析

基于游径感知视角,以泉州市梧林传统古村落为例,通过灰色统计分析法对乡村景观要素进行初步排除,结合图像语义分割技术将景观要素定量分析,构建美景度多元线性回归模型,研究梧林传统古村落景观质量评价。结果表明:天空开阔度、土壤裸露度、传统乡村建筑占比、乡村传统建筑色彩占比、村落景观道路占比、绿视率、铺装占比这7个景观要素与梧林传统古村落景观美景度显著相关;美景度模型结果显示具有高贡献率的3个主要因子依次是天空开阔度(0.387)、土壤裸露度(0.235)、传统乡村建筑占比(0.233);高美景度评价场景特征要素占比为天空开阔度26%~50%、土壤裸露度2%~5%、传统乡村建筑10%~30%;低美景度评价场景最突出特征为植被种类单一、道路设施风格与周围景观不协调、空间尺度感过大的景观空间;梧林传统古村落文旅空间景观质量存在空间分异性,主要受到空间要素、自然要素、建筑要素的共同影响。

一种改进最优拼接线的图像拼接算法

图像拼接过程中存在着视角、颜色、纹理信息等差异,这些差异严重影响了图像拼接质量。针对此问题,论文提出了一种改进最优拼接线的图像拼接算法。在图像配准方面,基于超像素分割算法对局部单应性模型进行改进,提高了图像配准的精度,削弱了视差对图像拼接过程的影响;在接缝切割方面,基于原有平滑项中考虑了色差与梯度差信息,在平滑项中添加了显著性感知权重与纹理性信息。实验证明,该算法提高了图像配准的精度,保护了视觉显著物体结构,避免了错位、重影现象的发生,能够拼接生成高清自然的宽视域图像。