The enhancement of noisy images has been playing a key role in improving the visual effect and the performance of image processing. Traditional methods for image enhancement are mainly focusing on eliminating noise, which cannot acquire good effect under low peak-signal-to-noise ratio (PSNR) conditions. Stochastic resonance (SR), on the contrary, is a technique using noise to enhance signal. Owing to the unique feature of SR, a novel binary image enhancement scheme based on aperiodic SR (ASR) technique is proposed. In this study, the authors take the improvement in PSNR as a measure of the ASR-based binary image enhancement system, which provides a guideline for the realisation of the ASR system. On this basis, they obtain the PSNR expression of the ASR-based binary image enhancement system. Simulation results show that the proposed method is superior to the traditional binary image enhancement methods both in visual effect and PSNR performance.

With rapidly advancing of contemporary medicine, it is necessary to help people identify various kinds of pills to prevent the adverse pill events. In this study, a high-accuracy automatic pill recognition system is proposed for accurate and automatic pill recognition. As pill imprint is main distinction between different pills, this system proposes algorithms on both imprint extraction and description parts to make use of imprint information. First, proposed modified stroke width transform is adopted to extract the imprint by detecting coherent strokes of imprint on the pill. Moreover, image segmentation by Loopy belief propagation is also added on printed imprint pills to solve the incoherent and coarse stroke problem. Second, a new descriptor named two-step sampling distance sets is proposed for accurate imprint description and successfully cut down the noise on extracted imprint. This strategy is based on the imprint partition – partitions the imprint on the basis of separated strokes, fragments and noise points. Recognition experiments are applied on extensive databases and result shows 90.46% rank-1 matching accuracy and 97.16% on top five ranks when classifying 12 500 query pill images into 2500 categories.

Super resolution (SR) reconstruction is often considered to be an inverse problem in the sense that unknown high resolution images are sought for giving low resolution images. Recent studies have shown that the sparsity regularisation used in compressed sensing (CS) reconstruction improves the performance of SR reconstruction. Furthermore, under the assumption that mutually similar regions exist within a natural image, non-local (NL) estimation produces accurate estimates for given degraded images. The incorporation of this NL estimation in SR reconstruction has been shown to yield better reconstructions. In this study, the authors propose the use of block matching and three-dimensional filtering with sharpening estimation as the regularisation constraint under the CS-based SR framework. This estimation collects similar blocks and adaptively filters them by the shrinkage of the transform coefficients. It recovers detailed structures while attenuating ringing artefacts. In addition, a sharpening technique used in the estimation also emphasises edges. As a result, the proposed SR algorithm searches for the solution that is similar to this enhanced estimate from among all feasible solutions. The experimental results demonstrate that the proposed method provides high-quality SR images, both numerically and subjectively.

In this study, the authors describe an objective smoothness assessment method for volume data. The metric can predict the extent of the difference in smoothness between a reference model, which may not be of perfect quality, and a distorted version. The proposed metric is based on the wavelet characterisation of Besov function spaces. The comparison of Besov norms between two models can resolve the global and local differences in smoothness between them. Experimental results from volume datasets with smoothing and sharpening operations demonstrate its effectiveness. By comparing direct volume rendered images, the experimental results show that the proposed smoothness index correlates well with human perceived vision. Finally, the metric can help the analyse compression distortions when they compare volume data with different smoothness.

Recently, computer vision-based methods have started to replace conventional sensor-based fire detection technologies. In general, visible band image sequences are used to automatically detect suspicious fire events in indoor or outdoor environments. There are several methods which aim to achieve automatic fire detection on visible band images, however, it is difficult to identify which method is the best performing as there is no fire image dataset which can be used to test the different methods. This study presents a benchmarking of state of the art wildland fire colour segmentation algorithms using a new fire dataset introduced for the first time. The dataset contains images of wildland fire in different contexts (fuel, background, luminosity, smoke etc.). All images of the dataset are characterised according to the principal colour of the fire, the luminosity, and the presence of smoke in the fire area. With this characterisation, it has been possible to determine on which kind of images each algorithm is efficient. Also a new probabilistic fire segmentation algorithm is introduced and compared to the other techniques. Benchmarking is performed in order to assess performances of 12 algorithms that can be used for the segmentation of wildland fire images.

The reference-table-based (RTB) steganography employs two pixels to conceal a secret digit, according to a predetermined reference table. Most of the RTB steganographic schemes discussed in the literature do not consider the non-uniformity of the secret digits and the locating numbers indicated by the pixel pairs. In contrast, the non-uniformity of the locating numbers exists for most of host images. When a non-uniform secret sequence is embedded, the quality control of the stego image is quite challengeable when compared with a uniform secret sequence. To the best of the authors knowledge, most of the existing algorithms focus on the uniform sequences. In this study, they propose a multi-round dynamic swap optimisation (MRDSO) algorithm for RTB steganography, which is more suitable for non-uniform secret sequences. By optimising the permutation of the reference table, the proposed algorithm increases the scenarios that the locating number indicated by the pixel pair is equal to the secret digit, which means that these host pixels can carry the secret digit without any distortion. The experimental results show that the proposed MRDSO algorithm increases the quality of the stego image by 0.1–1.7 dB.

This study presents an image processing algorithm capable of generating high dynamic range (HDR) images from a single frame. On the basis of liquid crystal display (LCD) backlight theory, the proposed algorithm uses an inverted local pattern approach to adjust the histogram of lighting distribution. To improve processing efficiency, sampled images are classified into three types according to exposure characteristics, which are used to obtain suitable processing factor. The inverting factor is mixed with the original sample to darken the bright areas and brighten the dark areas in the images. Brightness enhancement and auto-gain control are then added to expand the range of the grey levels. Authors’ results demonstrate the efficacy of the proposed HDR algorithm in improving shadow details. In addition, single-pass processing is used to reduce computational complexity, making the algorithm applicable for low-power portable cameras and video recorders.

In this study, the author presents an algorithm for approximating the contour of a digital planar image by cubic splines. In the authors’ method, a subset of points (called corners) from the contour is selected. These corners are used to segment the contour and each segment is then approximated by a cubic spline. Parameters of the fitted splines are estimated by optimisation methods. The novelty of the proposed approach lies in the way the corners are selected. An initial set of corners are first selected using a process which is called as iterative points-suppression. This initial set is further reduced by a novel technique termed spline-suppression. The result is a very compact cubic spline representation of the contour using few corners on the contour. The effectiveness of the proposed method is demonstrated on two large databases: MPEG7_CE-Shape-1_Part_B database and a database of handwritten characters.

Active contour models are effective image segmentation methods. However, they are very time-consuming, and their convergence depends upon the choice of initial contour. To overcome the two drawbacks, in the study, the authors suggest a signal-walking-driven active contour model. By walking a signal, they construct a forest of object evolution. Each tree grows from a root object, and child node contains its shrunk or/and split version. The merit value of an object is a composite metric from the colour, edge, or/and shape properties. The merit function plays an important role in tree construction and the goodness of object evolution. The objects are selected and added to the tree in the levels the merit function reaches the local maxima. After the forest of object evolution is constructed, by traversing each tree branch in post-order, the objects corresponding to maximum merit values are extracted as the final segmentation. Experimental results on a set of oil-sand images indicate the proposed signal-walking-driven active contour model outperforms Chan and Vese's model and adaptive thresholding.