Color Image Database Research Articles

A comparative study of steganography using watermarking and modifications pixels versus least significant bit

This article presents a steganography proposal based on embedding data expressed in base 10 by directly replacing the pixel values from images red, green blue (RGB) with a novel compression technique based on watermarks. The method considers a manipulation of the object to be embedded through a data compression triple process via LZ77 and base 64, watermark from low-quality images, embedded via discrete wavelet transformation-singular value decomposition (DWT-SVD), message embedded by watermark is recovered with data loss calculated, the watermark image and lost data is compressed again using LZ77 and base 64 to generate the final message. The final message is embedded in portable network graphic (PNG) images taken from the Microsoft common objects in context (COCO), ImageNet and uncompressed color image database (UCID) datasets, through a filtering process pixel of the images, where the selected pixels expressed in base 10, and the final message data is embedded by replacing units’ position of each pixel. In experimentation results an average of 40 dB in peak signal noise to ratio (PSNR) and 0.98 in the similarity structural index metric (SSIM) evaluation were obtained, and evasion steganalysis rates of up to 93% for stego-images, the data embedded average is 3.2 bpp.

Image and Texture Independent Deep Learning Noise Estimation Using Multiple Frames

In this study, a novel multiple frame based image and texture independent Convolutional Neural Network (CNN) noise estimator is introduced. Noise estimation is a crucial step for denoising algorithms, especially for ones that are called “non-blind”. The estimator works for additive Gaussian noise for varying noise levels. The noise levels studied in this work have a standard deviation equal to 5 to 25 increasing 5 by 5. Since there is no database for noisy multiple images to train and validate the network, two frames of synthetic noisy images with a variety of noise levels are created by adding Additive White Gaussian Noise (AWGN) to each clean image. The proposed method is applied on the most popular gray level images besides the color image databases such as Kodak, McMaster, BSDS500 in order to compare the results with the other works. Image databases comprise indoor and outdoor scenes that have fine details and richer texture. The estimator has an accuracy rate of 99 % for the classification and favourable results for the regression. The proposed method outperforms traditional methods in most cases. And the regression output can be used with any non-blind denoising method.

Robust mean shift filter for mixed Gaussian and impulsive noise reduction in color digital images

Noise reduction is one of the most important topics of digital image processing and despite the fact that it has been studied for a long time it remains the subject of active research. In the following work, we present an extension of the Mean Shift technique, which is efficiently reducing the Gaussian noise, so that it is able to cope with the impulsive disturbances. Furthermore, the elaborated technique can be applied to enhance the images corrupted by a mixture of strong Gaussian and impulsive noise, severely decreasing the quality of color digital images. By means of our approach, which is based on a novel similarity measure between a pixel and a patch located in the center of the processing block, even heavily disturbed images can be effectively restored, which enables the success of further stages of the image processing pipeline. We evaluate the efficiency of the proposed method using a publicly available database of test color images and compare the restored images applying a set of standard quality metrics with the results delivered by state-of-the-art denoising methods. Additionally, we compare our method with the Medoid and Quick Shift techniques, accelerating the original Mean Shift algorithm, in terms of objective quality criteria and computational complexity. The results of the performed experiments indicate that the proposed technique is superior to the widely used denoising techniques and can be used as a robust extension of the Mean Shift procedure. In the paper, a particular emphasis is placed on the ability of the presented algorithm to preserve and enhance image edges. The performed experiments evaluated with the use of the Pratt’s index, quantitatively confirm the superiority of the proposed design over the Mean Shift and standard denoising methods. The preservation of edges and even their sharpening is a very important feature of our algorithm whereas the final goal is segmentation, playing a crucial role in various computer vision tasks. The proposed algorithm is intended for the mixed noise reduction in color images, but it can be also applied in multispectral imaging and clustering of multidimensional data. To enable the comparison of our method with the standard denoising techniques and to help applying it in other image processing fields, we made its code freely available.

Efficient color image retrieval method using deep stacked sparse autoencoder

The recent advancement in deep learning-based approaches vastly outperforms the traditional image descriptors. Deep learning models, such as residual networks (ResNet), are well known for finding salient features. Although effective, high-level description often has a high dimensionality that increases computational overhead. The autoencoders find the useful approximation of the input data without losing critical information. Considering this, we propose a content-based image retrieval system for natural color images using a deep stacked sparse autoencoder (DSSA). The DSSA model learns latent features in an unsupervised way from the high-level description obtained using ResNet. The DSSA model achieves a nearly 50% reduction in size compared with the full-length features for the simple distance-based retrieval approach while increasing accuracy. The image retrieval efficacy of the learned latent features is also evaluated for two classifier-based methods using a Softmax classifier. Further, this study investigates the impact of unsupervised feature learning on retrieval using three benchmark natural color image databases of varying complexities, viz., Corel-1K, Corel-10K, and Canadian Institute for Advanced Research (CIFAR)-10. The latent features learned by the DSSA model with the fuzzy class membership-based retrieval method achieve promising improvements and yield a highly competitive retrieval performance with the large-size CIFAR-10 database.

A Statistical Modeling Framework for DCT Coefficients of Tampered JPEG Images and Forgery Localization

Various manipulations on JPEG images introduce single and multiple compression artifacts for forged and unmodified areas respectively. Based on the statistical analysis of JPEG compression cycle and on the finite mixture paradigm, we propose in this paper a modeling framework for AC DCT coefficients of such tampered JPEG images. Its accuracy is numerically assessed using the Kullback-Leibler divergence on the basis of a tampered JPEG image dataset built from six well-known uncompressed color image databases. To illustrate the framework utility, an application in image forgery localization is proposed. By formulating the localization as a clustering problem, we use the plug-in Bayes rule combined with a simple EM algorithm to distinguish between forged and unmodified areas. Numerous experiments show that, when the quality factor of final JPEG compression is high enough, the proposed modeling framework yields higher localization performances in terms of <i>F</i>₁-score than prior art regardless of divers local manipulations.

A SIFT Description Approach for Non-Uniform Illumination and Other Invariants

The new improvement keypoint description technique of image-based recognition for rotation, viewpoint and non-uniform illumination situations is presented. The technique is relatively simple based on two procedures, i.e., the keypoint detection and the keypoint description procedure. The keypoint detection procedure is based on the SIFT approach, Top-Hat filtering, morphological operations and average filtering approach. Where this keypoint detection procedure can segment the targets from uneven illumination particle images. While the keypoint description procedures are described and implemented using the Hu moment invariants. Where the central moments are being unchanged under image translations. The sensitivity, accuracy and precision rate of data sets were evaluated and compared. The data set are provided by color image database with variants uniform and non-uniform illumination, viewpoint and rotation changes. The evaluative results show that the approach is superior to the other SIFTs in terms of uniform illumination, non-uniform illumination and other situations. Additionally, the paper demonstrates the high sensitivity of 100%, high accuracy of 83.33% and high precision rate of 80.00%. Comparisons to other SIFT approaches are also included.

Two advanced classics exploiting DNA and RNA characteristics to encrypt a color image

This article tracks the development of a new algorithms supported by the abundant Vigenere lap, followed by two deeply improved Feistel lathes, acting on arbitrary-sized nucleotide blocks calculated from the chaos map used. This step can be ensured by adapting the basic gWith the progress of the theory of some mathemenetic operator to encrypt the color image. At the end of the second round, mutations and genetic crosses will be applied on the global image. Finally, a passage to the codons writing with the intention of modification of the released protein, extracting the Exons, Introns and Stop codons. Testing on a large number of different sizes images and formats randomly selected from the color image database ensures that our system can endures any known attacks.

Research on blur discrimination thresholds of three-Channel for color image

On the basis of modern theory of color vision, the blur discrimination threshold of color image with three channels (luminance, red-green and the yellow-blue channel) in the opponent color space (DKL space) is studied through visual experiments in this paper. The visual evaluation experiment is designed by the psychophysical method based on Quest algorithm. Finding the blur discrimination threshold of color image according to the DKL space is of great significance for helping us understand color perception of image. It also has a wide range of applications in areas such as color image compression during color image transmission. Two (flowers and artificial object) typical images in the McGill standard color image database were chosen as the experimental images. After converting the color space of the image from RGB to DKL space, Gaussian low-pass filtering is used to establish the fuzzy image database. Quest algorithm which is a psychological physical method is used to design the visual evaluation experiment. Eleven volunteers with normal color vision were invited to participate in the visual evaluation experiment for each of the three channels of the two images. Our main conclusions are: (1) Human eyes are more sensitive to luminance information than color information. (2) The larger the contrast of the image and the smaller the uniformity of the image, the more sensitive the human eye is to changes in image blur.

Luminance approximated vector quantization algorithm to retain better image quality of the decompressed image

Some compressed images using Vector Quantization algorithm suffers from blocking artifacts which degrades the visual appeal of the image. Present study proposes a hybrid vector quantization method applicable on de-correlated color model. As luminance channel carries image information and loss of image information results in degradation of the visual appeal of an image, so aim of this study is focused on retaining more image information during compression process. For luminance channel compression, a new four level quantization based compression method is developed. Luminance channel is partitioned into smaller blocks. Then for each block, four level quantization is applied which local to the current block only. This results many level luminance value effectively for the whole image. It helps to retain better information. Chrominance channels are compressed using conventional Vector Quantization. This hybrid compression method improves visual quality of the decompressed image reasonably compared to VQ. The proposed method is applied on many standard images found in literature and images of UCIDv.2 color image database. Results are analyzed in terms of Peak Signal to Noise Ratio, Structure Similarity Index and space requirement reduction for compressed image using the method. Experimental results show that proposed method retains better quality of image in terms of PSNR and SSIM than Vector Quantization and Modified Vector Quantization. This method reduces storage space requirement for the compressed images in the range of 84% to 89%.

Classification and Indirect Weighing of Sweet Lime Fruit through Machine Learning and Meta-heuristic Approach

ABSTRACT In the past few decades, both academicians and industries have shown interest toward the agricultural post-harvest operation aiming to reduce the post-harvest losses. In order to assist farmers in post-harvest decision-making some effective and innovative methodological frameworks are required. The fruit weight measurement is of prime importance in many food processing industries during sorting, grading, and packaging. In this work, different Support vector machine (SVM) classifiers as well as weighing models developed using the optimized adaptive neuro-fuzzy inference system (ANFIS) coupled with a computer vision system are proposed. More precisely, the weighing models based on the hybrid ANFIS approach using two well-known optimization algorithms are analyzed. In the first approach, a series of GA-ANFIS models have been evaluated for different population size. In the later approach, different PSO-ANFIS models have been evaluated by varying the most influential parameters. The comprehensive self-built color image database has been used for both calibration and validation of the models. From an economic point of view, this indirect way of weighing fruits may be useful to fruit growers and traders in deciding the market depending on the fruit size and weight before packaging. The result shows the higher reliability and prediction capability of the proposed meta-heuristics (GA-ANFIS) model in estimating the weight of sweet lime fruit.

Improved Single Feistel Circuit Supporter by A Chaotic Genetic Operator

This document outlines a new color image encryption technology development. After splitting the original image into 240-bit blocks and modifying the first block by an initialization vector, an improved Feistel circuit is applied, sponsored by a genetic crossover operator and then strong chaining between the encrypted block and the next clear block is attached to set up the confusion-diffusion and heighten the avalanche effect, which protects the system from any known attack. Simulations carried out on a large database of color images of different sizes and formats prove the robustness of such a system.

Retinal blood vessel segmentation from fundus image using an efficient multiscale directional representation technique Bendlets.

The improper circulation of blood flow inside the retinal vessel is the primary source of most of the optical disorders including partial vision loss and blindness. Accurate blood vessel segmentation of the retinal image is utilized for biometric identification, computer-assisted laser surgical procedure, automatic screening, and diagnosis of ophthalmologic diseases like Diabetic retinopathy, Age-related macular degeneration, Hypertensive retinopathy, and so on. Proper identification of retinal blood vessels at its early stage assists medical experts to take expedient treatment procedures which could mitigate potential vision loss. This paper presents an efficient retinal blood vessel segmentation approach where a 4-D feature vector is constructed by the outcome of Bendlet transform, which can capture directional information much more efficiently than the traditional wavelets. Afterward, a bunch of ensemble classifiers is applied to find out the best possible result of whether a pixel falls inside a vessel or non-vessel segment. The detailed and comprehensive experiments operated on two benchmark and publicly available retinal color image databases (DRIVE and STARE) prove the effectiveness of the proposed approach where the average accuracy for vessel segmentation accomplished approximately 95%. Furthermore, in comparison with other promising works on the aforementioned databases demonstrates the enhanced performance and robustness of the proposed method.

Noise-robust color edge detection using anisotropic morphological directional derivative matrix

In this paper, a color edge detector using the anisotropic morphological directional derivatives (AMDDs) is presented to detect edges in color images corrupted by Gaussian or impulsive noise. The AMDD matrix, robust to impulsive noise owing to the underlying morphological operators, is constructed to represent edge information at each pixel of a color image. The color edge strength map and color edge direction map of a color image are extracted by spatial and directional matched filtering and singular value decomposition of the AMDD matrices. Embedding them in the route of the Canny detector yields a noise-robust color edge detector. Moreover, a color image database with groundtruths (GTs) of edges are built. The GT of a color image is generated in three steps. First, the contours and results of multiple color edge detectors are fused into a candidate edge map (CEM). Next, the CEM, the original image, and a special software for edge modification are sent to twenty experienced observers to modify the CEM. Finally, their results are used to yield the edge pixels, non-edge pixels, and don't care regions in the GT by the voting rule. The proposed detector is compared with existing color edge detectors on the database.

Classification of image distortion based on the generalized Benford’s law

Distortion classification is an important step in blind image quality assessment. In this paper, a new image distortion classification algorithm is presented. Classification is based on features extracted from the distribution of the first digit of transform coefficients of the image. The generalized Benford’s law is used to model the distribution. The discrete cosine transform with three different patch sizes and the wavelet transform have been tested. Features, such as distribution data and model parameters, are extracted from an image. A kernel support vector machine is trained using these features. The LIVE database is used for both training and testing, while other four databases, namely, TID2008, CSIQ, Waterloo exploration database and McGill calibrated colour image database, are used for validation. Experimental results show that the performance of the proposed algorithm outperforms state-of-the-art algorithms in terms of classification accuracy.

Image encryption scheme based on a new secure variant of Hill cipher and 1D chaotic maps

This article suggests a new image encryption algorithm based on a secure variant of Hill Cipher (HC) and three improved one-dimensional (1D) chaotic maps. The proposed scheme fully satisfies the two basic concepts of security, namely confusion and diffusion. The confusion is ensured by the product of a vector consisting of the key-pixel couple and a 2 × 2 Hill matrix on the one hand, and the addition of another pseudo-random translation vector on the other hand. And the diffusion is ensured by a strong avalanche effect that links each encrypted pixel to its adjacent. The chaotic sequences used come from 1D chaotic maps with excellent statistical proprieties. The proposed algorithm aims to encrypt pixel-by-pixel all types of images, even those with a high correlation of adjacent pixels or a black background, and to make the encryption system more secure. So, the simulation results performed on a database of color and grayscale images show that the histogram of a cipher image produced a uniform distribution, a higher entropy, a correlation among image pixels that was significantly decreased; Also, the proposed algorithm has great performance in terms of sensitivity, UACI and NPCR.

Counter JPEG Anti-Forensic Approach Based on the Second-Order Statistical Analysis

The forensic investigation of JPEG compression generally relies on the analysis of first-order statistics based on image histogram. The JPEG compression detection methods based on such methodology can be effortlessly circumvented by adopting some anti-forensic attacks. This paper presents a counter JPEG anti-forensic method by considering the second-order statistical analysis based on the co-occurrence matrices (CMs). The proposed framework comprises three stages: selection of the target difference image, evaluation of CMs, and generation of second-order statistical feature based on CMs. In the first stage, we explore the effects of dithering operation of JPEG anti-forensics by analyzing the variance inconsistencies along the diagonals. Afterward, CMs are evaluated in the second stage to highlight the effects of grainy noise introduced during the dithering operation. The third stage is devoted to generate an optimal second-order statistical feature which is fed to the SVM classifier. The experimental results based on the uncompressed color image database and BOSSBase dataset images demonstrated that the proposed forensic detector based on CM is very efficient even in the presence of anti-forensic attacks. Moreover, the experimental results also confirm the competency of the proposed method in counter median filtering and contrast enhancement anti-forensics. The proposed scheme also provides satisfactory results in detecting other image processing operations such as mean filtering, Gaussian filtering, Weiner filtering, scaling, and rotation, thereby revealing its multi-purpose nature.

A proposed multi-image compression technique

Image compression is a process that reduces memory space required to store an image. The image compression techniques are broadly classified into two categories a) Lossless technique b) Lossy technique. Lossy compression technique achieves higher result, but due to data loss it may cause spatial in consistency, blocking artifact, quantization noise in the decompressed image degrading the image quality. Most of the existing compression techniques are applicable on single image separately. In this study an image compression method is proposed where multiple images of the same size are combined together and compressed to achieve higher compression ratio while keeping image quality as close as standard image compression techniques. Luminance channel of each image is compressed separately using Vector Quantization (VQ) algorithm while two chrominance channels, Cb and Cr of all images are combined into a three dimensional matrix that forms training vector. Clustering is applied on the training vector to get the initial color representatives. Thus, for the chrominance channels of n number of images, the proposed method generates one index matrix and one centroid matrix of size 256×2× n where 256 is the number of clusters. This centroid matrix contains one 256×2 dimensional centroid matrix for each individual image. This 256×2 matrix contains centroid of each cluster. The centroids of each and every cluster of an image are updated individually using optimization technique to get a better centroid (Cb, Cr) pair. This process updates the color representative pair of Cb and Cr further. This method has been applied on standard images in literature and images collected from UCID v. 2 color image database. Experimental results are analyzed in terms of PSNR and space reduction. Experimental results show that the proposed method achieves a higher compression ratio retaining almost similar image information as other standard lossy compression algorithm.

Hybrid Scheme of Image’s Regional Colorization Using Mask R-CNN and Poisson Editing

Image colorization is a creative process of reasonably adding colors on gray-scale images to generate well-pleasing colorized images. The existing colorization methods normally require user-supplied hints of color points and doodles, or handpicked color reference images for transferring colors, or diverse color images for predicting the colorized results; but the final colorized results generated by most of them may seem unnatural as a consequence of the unprofessional users' skills, inaccurately color transferring, or limited scale of color image collection. To overcome these limitations, a hybrid scheme consisting of two modules is proposed for images' region colorization by combining semantic segmentation and seamless fusion techniques in this paper. In the first module, the masks and category of input image's regions and background are derived from a Mask R-CNN model, and the corresponding reference images of each region are selected from a pre-classified color image database. In the second module, the background and various regions of an image are colorized by a U-Net model and a VGG model respectively. Then, the Poisson editing technique is applied for fusing all the colorized results to generate the final whole colorized image. The experiments show that our scheme can not only flexibly select the appropriate reference images for different regions of the image according their semantic information, but also effectively merge colorized results to generate a plausible colorful image. By reasonably combining different CNN-based models according to their superiority, our scheme avoids the limitation of failures caused by single-step methods, and achieves richer artistic visual effect compared with other existing methods.

Early Detection of High Blood Pressure and Diabetic Retinopathy on Retinal Fundus Images Using CBRIR Based on Lifting Wavelets

In this paper we present a lifting wavelet based CBRIR image retrieval system that uses color and texture as visual features to describe the content of a retinal fundus images. Our contribution is of three directions. First, we use lifting wavelets 9/7 for lossy and SPL5/3 for lossless to extract texture features from arbitrary shaped retinal fundus regions separated from an image to increase the system effectiveness. This process is performed offline before query processing, therefore to answer a query our system does not need to search the entire database images; instead just a number of similar class type patient images are required to be searched for image similarity. Third, to further increase the retrieval accuracy of our system, we combine the region based features extracted from image regions, with global features extracted from the whole image, which are texture using lifting wavelet and HSV color histograms. Our proposed system has the advantage of increasing the retrieval accuracy and decreasing the retrieval time. The experimental evaluation of the system is based on a db1 online retinal fundus color image database. From the experimental results, it is evident that our system performs significantly better accuracy as compared with traditional wavelet based systems. In our simulation analysis, we provide a comparison between retrieval results based on features extracted from the whole image using lossless 5/3 lifting wavelet and features extracted using lossless 9/7 lifting wavelet and using traditional wavelet. The results demonstrate that each type of feature is effective for a particular type of disease of retinal fundus images according to its semantic contents, and using lossless 5/3 lifting wavelet of them gives better retrieval results for almost all semantic classes and outperform 4-10% more accuracy than traditional wavelet.

A Database with Reference for Image Dehazing Evaluation

In this article, the authors introduce a new color image database, CHIC (Color Hazy Images for Comparison), devoted to haze model assessment and dehazing method evaluation. For three real scenes, they provide two illumination conditions and several densities of real fog. The main interest lies in the availability of several metadata parameters such as the distance from the camera to the objects in the scene, the image radiance and the fog density through fog transmittance. For each scene, the fog-free (ground-truth) image is also available, which allows an objective comparison of the resulting image enhancement and potential shortcomings of the model. Five different dehazing methods are benchmarked on three intermediate levels of fog using existing image quality assessment (IQA) metrics with reference to the provided fog-free image. This provides a basis for the evaluation of dehazing methods across fog densities as well as the effectiveness of existing dehazing dedicated IQA metrics. The results indicate that more attention should be given to dehazing methods and the evaluation of metrics to meet an optimal level of image quality. This database and its description are freely available at the web address http.;thic.ii-bourgogne.fr. (C) 2018 Society for Imaging Science and Technology