The use of paired-comparison psychophysical experiments is an important technique that is used widely in imaging studies. It is sometimes difficult to compare every stimulus with every other; the number of paired comparisons for n stimuli becomes prohibitive for large values of n. Thus, experiments are often designed by missing some pairs. However, the effect on the accuracy of the estimations of the scale values is not clear. Similarly, if more resources are available, would it be better to recruit more observers making the same paired comparisons or to have the original observers carry out additional paired comparisons? This work seeks to develop a framework for addressing these practical questions surrounding incomplete paired-comparison experiment design. A Monte Carlo computational simulation is carried out with an ideal-observer model. Results suggest that the proportion of paired comparisons is more critical than the number of observers with small numbers of stimuli.
Long-exposure shots and flash photography are normally used to acquire images under low-light conditions. However, flash photography often induces color distortion and creates a red-eye effect, while long-exposure shots are prone to motion blur due to camera shake or subject motion. Thus, multi-spectral flash imaging has recently been introduced to overcome the limitations of traditional low-light photography. Multi-spectral flash imaging combines invisible and visible spectrum images, but most multi-spectral flash approaches result in color distortion due to the lower accuracy of the invisible spectrum image. Accordingly, this article presents a multi-spectral flash imaging algorithm using optimization with a weight map to improve the color accuracy and brightness of the image. First, UV/IR and visible spectrum images are captured. To compensate the luminance values under low-light conditions, adaptive tone reproduction is performed using the Naka–Rushton equation. Next, to discriminate uniform regions from detail regions, a weight map is generated using a Canny operator. Finally, the optimization process takes account of the likelihood of the visible light image, the sparsity of the image gradients, and the spectral constraints of the IR and UV channels. The performance of the proposed method is subjectively evaluated using a z-score, and the resulting images are confirmed to have an improved color accuracy and lower noise when compared with the results of other methods.
The non-local means (NLM) method is an effective and robust image denoising method. However, the existing NLM methods cannot guarantee the global optimum solution because they use a globally fixed bandwidth parameter when computing the similarity weight function. To address this problem, this article proposes an adaptive NLM method that can achieve a good trade-off between edge preservation and noise reduction. First, the difference curvature indicator is used to identify the local characteristics of each pixel. Then, depending on the properties of the difference curvature indicator, an adaptive bandwidth parameter is constructed. As a result, the bandwidth parameter depends continuously on the local characteristics of each pixel, which pixel-wise realize the selection of bandwidth. Experimental results show the effectiveness of our proposed method when compared with the mainstream methods.
As a measure against abnormal jetting caused by the nozzle of the print head drying after evaporation of the ink, the authors attempted to refresh the ink by installing an ink circulation path in addition to the conventional nozzle shaking which mixes the ink in the nozzle by means of piezo vibrations. On understanding the advantages of nozzle shaking and the ink circulation path on this structure, the use of both measures provided the authors with a complementary and highly reliable method for preventing any increase in viscosity. Using a hyper-speed camera they also observed an unusual behavior where the nozzles in non-jetting conditions begin to jet again as the non-jetting time increases. This phenomenon affects printing stabilization and can be well explained by the assumption that inkjetting occurs because of advection and lowered viscosity of the ink.
Although the n-color printing process increases the color gamut, it creates a challenge in generating color separations. This article evaluates different methods of implementing the inverse printer model to obtain the color separation for n-color printing of discrete (spot) colors. The constrained optimization and lookup table based inversion methods were evaluated for n-color printing processes. The colorant space was divided into sectors of four inks and the inverse printer models were applied to each sector.
The results were found to be adequate, with the mean CIEDE2000 values between the measured colors and the model predicted colors below 1.5 for most of the models. The lookup table based inversion was faster than the constrained optimization approach. The nine-level lookup table model gave accurate prediction without costing excessive processing time. It can be used to replace spot colored inks with the seven-color printing process in packaging printing to achieve significant cost savings.
Dictionary based learning has emerged as a powerful approach to a large class of machine learning problems, especially face recognition. The development of face recognition methods for unconstrained environments is still a challenging problem. In this article the authors present a dictionary based approach that considers compact face features to define a cluster centroid using k-means clustering in conjunction with a sparse representation classifier. The varying environmental aspects of human face recognition, namely, illumination and facial expression, have been dealt with for images captured under controlled and uncontrolled settings. Face normalization using the gradient face method is employed to handle variations in illumination conditions. Facial expression is handled by the use of compact face features, generated using the popular rotation invariant uniform local binary pattern and the histogram of gradients. The efficiency of the proposed method is demonstrated using three large benchmark databases with vast variations, extended Yale B, CMU-PIE and IMFDB. It is encouraging to note that the proposed method has superior performance to popular face recognition algorithms.