Image Content Retrieval Research Articles

Highly compressed image representation for classification and content retrieval

In this paper, we propose a new method of representing images using highly compressed features for classification and image content retrieval – called PCA-ResFeats. They are obtained by fusing high- and low-level features from the outputs of ResNet-50 residual blocks and applying to them principal component analysis, which leads to a significant reduction in dimensionality. Further on, by applying a floating-point compression, we are able to reduce the memory required to store a single image by up to 1,200 times compared to jpg images and 220 times compared to features obtained by simple output fusion of ResNet-50. As a result, the representation of a single image from the dataset can be as low as 35 bytes on average. In comparison with the classification results on features from fusion of the last ResNet-50 residual block, we achieve a comparable accuracy (no worse than five percentage points), while preserving two orders of magnitude data compression. We also tested our method in the content-based image retrieval task, achieving better results than other known methods using sparse features. Moreover, our method enables the creation of concise summaries of image content, which can find numerous applications in databases.

Saliency-Enhanced Content-Based Image Retrieval for Diagnosis Support in Dermatology Consultation: Reader Study.

Previous research studies have demonstrated that medical content image retrieval can play an important role by assisting dermatologists in skin lesion diagnosis. However, current state-of-the-art approaches have not been adopted in routine consultation, partly due to the lack of interpretability limiting trust by clinical users. This study developed a new image retrieval architecture for polarized or dermoscopic imaging guided by interpretable saliency maps. This approach provides better feature extraction, leading to better quantitative retrieval performance as well as providing interpretability for an eventual real-world implementation. Content-based image retrieval (CBIR) algorithms rely on the comparison of image features embedded by convolutional neural network (CNN) against a labeled data set. Saliency maps are computer vision-interpretable methods that highlight the most relevant regions for the prediction made by a neural network. By introducing a fine-tuning stage that includes saliency maps to guide feature extraction, the accuracy of image retrieval is optimized. We refer to this approach as saliency-enhanced CBIR (SE-CBIR). A reader study was designed at the University Hospital Zurich Dermatology Clinic to evaluate SE-CBIR's retrieval accuracy as well as the impact of the participant's confidence on the diagnosis. SE-CBIR improved the retrieval accuracy by 7% (77% vs 84%) when doing single-lesion retrieval against traditional CBIR. The reader study showed an overall increase in classification accuracy of 22% (62% vs 84%) when the participant is provided with SE-CBIR retrieved images. In addition, the overall confidence in the lesion's diagnosis increased by 24%. Finally, the use of SE-CBIR as a support tool helped the participants reduce the number of nonmelanoma lesions previously diagnosed as melanoma (overdiagnosis) by 53%. SE-CBIR presents better retrieval accuracy compared to traditional CBIR CNN-based approaches. Furthermore, we have shown how these support tools can help dermatologists and residents improve diagnosis accuracy and confidence. Additionally, by introducing interpretable methods, we should expect increased acceptance and use of these tools in routine consultation.

Handwritten Digits and Optical Characters Recognition

The process of transcribing a language represented in its spatial form of graphical characters into its symbolic representation is called handwriting recognition. Each script has a collection of characters or letters, often known as symbols, that all share the same fundamental shapes. Handwriting analysis aims to correctly identify input characters or images before being analysed by various automated process systems. Recent research in image processing demonstrates the significance of image content retrieval. Optical character recognition (OCR) systems can extract text from photographs and transform that text to ASCII text. OCR is beneficial and essential in many applications, such as information retrieval systems and digital libraries.

Efficient and secure content-based image retrieval with deep neural networks in the mobile cloud computing

Smart devices offer a variety of more convenient forms to help us record our lives and generate a large amount of data in this process. Limited by the local storage capacity, many users outsource their image data directly to the cloud server. However, images stored in plaintext on the cloud server are very insecure, resulting in image privacy information can be easily leaked. Therefore, users will encrypt the images and outsource them to the cloud server, but the encrypted images cannot be retrieved. Therefore, we proposed a secure and efficient ciphertext image retrieval scheme based on image content retrieval (CBIR) and approximate homomorphic encryption (HE). First, we used approximate homomorphic encryption to encrypt images after resizing and uploaded the ciphertext images to the cloud for feature extraction of ciphertext. At the same time, the large images (size, dimension, and resolution) would generate data inflation after using homomorphic encryption. Therefore, the original images are encrypted using the chaotic image encryption scheme to reduce ciphertext size and computation costs. Second, we proposed two deepening network depth optimization strategies that address the problem of insufficient neural network depth. Finally, reducing the dimensionality of the ciphertext feature vector using locally sensitive hashing (LSH) can accelerate the retrieval of ciphertext images. Compared with other literature, our ciphertext image retrieval scheme can significantly reduce the rounds of user-server communication.

Performing content-based image retrieval using rotated local binary pattern and multiple descriptors

Images have been viewed as a powerful medium for displaying visual information in numerous applications. Content-based retrieval and querying the indexed collections are required to access visual information. Content image retrieval is an efficient means of retrieving similar images from large repositories which require high speed and accuracy. The extraction of features is imperative for efficient results. Following research work aimed at colors and texture extraction for enhancing image metadata. The proposed technique was evaluated on WANG datasets using MATLAB. WANG consists of one thousand images of flowers, buildings and forest. Each category consists of one hundred images of size 187x126 or 126x187 in JPEG format. The features were extracted using RGB and HSV. Rotated Local Binary Patterns were used and distances between images was obtained using Euclidean algorithm. The results showed high accuracy and precision.

Computer Image Content Retrieval considering K-Means Clustering Algorithm

The traditional computer image content retrieval technology can only meet the specific requirements of customers; because of its general features, it cannot comply with the requirements of various environments, purposes, and time simultaneously. This study presents a computer image content retrieval method for a K-means clustering algorithm (KCA). The information collected by computer is preprocessed by K-means clustering algorithm, and the unacquired computer image is labeled based on the optimal learning order according to the KCA. The K-means clustering algorithm classifies the color, pattern, shape, and content of computer images and takes advantage of the invariance advantages of image content retrieval such as scale, rotation, illumination, and blur correction to effectively solve the recurring problems of computer images during retrieval and increase its accuracy. The results of the experiment indicate that the proposed K-means clustering algorithm can enhance the efficiency and performance effectively during image retrieval by computer as compared to the traditional content search algorithm and also help to quickly converge to the query content; it also shows that KCA has good robustness.

Fault Detection of Permanent Magnet Synchronous Motor Based on Image Content Retrieval

The motor fault detection based on data-driven approach has achieved some results. However, there are still some problems, such as complex signal processing, long training time and poor adaptability to different working conditions. A new fault detection method for permanent magnet synchronous motor based on image content retrieval is proposed in this paper, which does not require complex signal processing and model training, and has strong adaptability to conditions. First, the fault signals are converted into images by the symmetrized dot pattern method. Second, an image coding technology based on KAZE feature and bag of features is proposed to reduce the computational cost. Then, to improve adaptability of conditions, a cyclic update algorithm of image retrieval library based on image fusion is proposed. Finally, fault detection is realized through image retrieval of feature code matching. Extensive experiments demonstrate the superiority and effectiveness of the proposed method.

Application of image content feature retrieval based on deep learning in sports public industry

The image content retrieval can effectively promote the development of the entire industry. At present, sports competition is becoming more and more fierce, and the requirements for image content retrieval are getting higher and higher. In this paper, research has been carried out on image descriptor generation, image feature quantization and coding, accurate nearest neighbor cluster center fast search, multi-dimensional inverted index construction and fast retrieval. Moreover, based on deep learning, this paper constructed an effective detection algorithm for the characteristics of sports images, and compared the image shape and color as examples. It can be seen from the comparative study that the research method of this paper can effectively reduce the size of the candidate set of query results without affecting the accuracy of the query, which is of great significance for improving the speed of image query and has certain significance for promoting the development of sports public industry.

Specific category region proposal network for text detection in natural scene

Natural scene text usually carries considerable abstract semantic information, which is closely related to the surrounding environment. Thus, natural scene text detection plays a vital role in image content retrieval and understanding. In this study, the authors propose a novel specific category region proposal network (SCRPN) based on maximally stable extremal regions (MSER) and fully convolutional network (FCN) for natural scene text detection. First, FCN for pixel-level recognition is utilised to obtain the text saliency map and MSER is used to obtain oversegmented regions. Then, the multiple features of oversegmented regions and text saliency map are used for region aggregation. Next, single-linkage clustering method is adopted to cluster the segmentation regions to obtain a hierarchical structure of text region proposals. Finally, for the top-ranking region proposals, SCRPN built an end-to-end pipeline for scene text detection directly. Experiments on street view text and international conference on document analysis and recognition (ICDAR) 2013 have demonstrated the effectiveness of SCRPN for generating the text proposals. SCRPN could work with various two-stage text detection networks; thus, faster region convolutional neural network was used as the text detection framework to evaluate the performance of SCRPN in the ICDAR 2015 and MSRA-TD500 benchmarks. The experimental results confirmed that SCRPN makes text detection more robust in complex scenarios.

Using Color and Texture Feature Extraction Technique to Retrieve Image

This paper proposes a content image retrieval using the texture and the color feature of the images. Although for extraction of texture feature, the “gray level co-occurrence matrix (GLCM) algorithm” is used and for extracting color feature the color histogram is used. The presented system is tested on the WANG database that contains a thousand color images with ten different classes by the help of three various type of distances

Color–Texture Pattern Classification Using Global–Local Feature Extraction, an SVM Classifier, with Bagging Ensemble Post-Processing

Many applications in image analysis require the accurate classification of complex patterns including both color and texture, e.g., in content image retrieval, biometrics, and the inspection of fabrics, wood, steel, ceramics, and fruits, among others. A new method for pattern classification using both color and texture information is proposed in this paper. The proposed method includes the following steps: division of each image into global and local samples, texture and color feature extraction from samples using a Haralick statistics and binary quaternion-moment-preserving method, a classification stage using support vector machine, and a final stage of post-processing employing a bagging ensemble. One of the main contributions of this method is the image partition, allowing image representation into global and local features. This partition captures most of the information present in the image for colored texture classification allowing improved results. The proposed method was tested on four databases extensively used in color–texture classification: the Brodatz, VisTex, Outex, and KTH-TIPS2b databases, yielding correct classification rates of 97.63%, 97.13%, 90.78%, and 92.90%, respectively. The use of the post-processing stage improved those results to 99.88%, 100%, 98.97%, and 95.75%, respectively. We compared our results to the best previously published results on the same databases finding significant improvements in all cases.

Efficient retrieval algorithm for multimedia image information

The research on the retrieval of multimedia image data information is of great significance for increasing the retrieval rate of multimedia image information. Due to the certain similar characteristics of massive multimedia image information, the picture information features are confused. The traditional image retrieval method mainly uses the image information feature to classify and retrieve. When the picture information is disordered, it is impossible to classify the mass multimedia image information features, resulting in slow retrieval speed and low accuracy. A new high-efficiency retrieval algorithm for massive multimedia image information is proposed and optimized. Based on the theory of granular computing, an image region similarity measurement method for content retrieval is proposed. The image feature information table is transformed into an ordered matrix form. By studying the ordered matrix, the concept of image feature granules and granule granules is introduced, the importance of image features is analyzed from different granularity levels, and the order relationship between regions in the image feature information table is maintained, and the weight of the theoretical image feature is calculated based on the granularity for implementing the image region similarity measurement method. The example shows that the similarity measure method can measure the degree of similarity between image regions objectively and effectively, and provides a new idea and method for the research of granular computing theory in multimedia image content retrieval.

Image Feature Layered Indexing Method for Transparent Access

In the data transparent access process, it is often necessary to query the data accurately to find the location of the data that the user needs to access. This paper studies the image feature organization and indexing of image data based on image content retrieval, and uses the image plane position constraint relationship to propose an image feature layered index method. In the process of image transparent access, data storage management needs to detect and extract local feature information of the image (such as SIFT feature points), and then design appropriate data structures to organize and index massive high-dimensional feature vectors, and in a specific storage structure. Design an efficient and accurate feature search algorithm. The paper mainly includes two parts: LSH high-dimensional indexing algorithm for thumbnails and KD-Tree high-dimensional indexing algorithm based on improved position information.

Remote sensing image retrieval using object-based, semantic classifier techniques

Data captured by satellites is increasing exponentially for agriculture and crop management, health, climate changes and future prediction of plants. Nowadays, it is required to have reliable, automated, satellite image classification and recovering system. Every day there is a massive amount of remotely-sensed data being collected and sent by satellite. Many retrieval systems are proposed for image content and information retrieval. However, the output of these approaches is generally not up to expectation. In this work, a new approach, remote sensing image retrieval scheme by content base image retrieval with grid computing and advanced database concepts are used and this will help to speed up both input processing and system response time. This paper presents the idea of the parallel processing of input data, queries, and storing images in the database using advanced database concept like B+ or BST trees.

Multi resolution feature combined with ODBTC technique for robust CBIR system

Content based image retrieval (CBIR) is a system that retrieves a set of images that most resembles the query image. The technology is used in many applications. Currently used image content retrieval method is ordered-dither block truncation coding (ODBTC). This method is used to produce image content descriptors. In this system, it gives only an average accuracy of 70.5%. Our aim is to create a more robust and accurate system for CBIR. For this purpose in addition to colour cooccurrence feature (CCF) and bit pattern features (BPF), contourlet and wavelet features from the query image is extracted for image retrieval process. In our experiment the system is first tested with ODBTC and wavelet and then ODBTC and contourlet. The results obtained with ODBTC and contourlet is more accurate and produced accuracy 91.5%. The dataset used for our experiment is CorelDB.

Content image retrieval using combination of dominant and LBP based features

Content image retrieval using combination of dominant and LBP based features

SABC-SBC: a hybrid ontology based image and webpage retrieval for datasets

In the recent days, web mining is the one of the most widely used research area for finding the patterns from the web page. Similarly, web content mining is defined as the process of extracting some useful information from the web pages. For this mining, a Block Acquiring Page Segmentation (BAPS) technique is proposed in the existing work, which removes the irrelevant information by retrieving the contents. Also, the Tag-Annotation-Demand (TAD) re-ranking methodology is employed to generate the personalized images. The major disadvantage of these techniques is that it fails to retrieve both the images and web page contents. In order to overcome this issue, this paper focused to integrate the TAD and BAPS techniques for the image and web page content retrieval. There are two important steps are involved in this paper, which includes, server database upload and content extraction from the database. Furthermore, the databases are applied on the Semantic Annotation Based Clustering (SABC) for image and Semantic Based Clustering (SBC) for webpage content. The main intention of the proposed work is to accurately retrieve both the images and web pages. In experiments, the performance of the proposed SABC technique is evaluated and analyzed in terms of computation time, precision and recall.

Mobile Application for Archaeological Site Image Content Retrieval and Automated Generating Image Descriptions with Neural Network

This paper presents a novel algorithm for generating descriptions of stupa image such as stupa's era, stupa's architecture and other description in mobile application by using key points generated from SIFT algorithms and learning stupa description from the generated key points with artificial neural network. Neural network was used for being the classifier for generating the description. We have presented a new approach to feature extraction based on analysis of key points and descriptors of an image. The algorithms were tested with stupa image dataset in Phra Nakhon Si Ayutta province, Sukhothai province and Bangkok. The experimental results show that the proposed framework can efficiently give the correct descriptions to the stupa image compared to using the traditional method.

Remote Sensing Image Content Retrieval Based on Frequency Spectral Energy

There are different research and processing methods for Image Content Retrieval. This article is described the retrieval of remote sensing images method, which is based on the image texture features of the Fourier power spectrum. Firstly, a fixed size remote sensing image unit is selected from the remote sensing image, which is converted to a spectral image by FFT, and extract the eigenvalues in the frequency domain to constitute a searchable image eigenvalue database. Secondly, the similarity between the feature and the database features are examined by the Angular Cosine Algorithm. Lastly, similarity of the resulting image are sorted to select the best match. This method converts the remote sensing image into a spectral image, and makes the selection of the eigenvalue more simple. The Angle Cosine similarity detection method with weight is used to make the similarity detection more accurate.

An Efficient QBIR System Using Adaptive Segmentation and Multiple Features

Query by Image Content Retrieval abbreviated as QBIR, has become new thirst now a days. By using this systems, user can retrieve the similar images of an already existed image (or) a rough sketch (or) a symbolic representation. To make more efficient and user friendly QBIR multiple features areemployed. This paper proposes a novel approach for image retrieval using adaptive k-means clustering and shape, texture features. The experimental results portraystheperformance of the proposed retrieval system in terms of better precision. To evaluate the proposed method COIL and MPEG-7 shape 1 datasets are used.