Search

Toward realizing Automatic High Beam, we are working on a study to classify the beam mode of headlights as "High" and "Low" from night-time in-vehicle camera images using deep learning. Deep learning requires a large amount of training data, and the more effective data are used for training, the better the accuracy of the model is. However, since creating training data is very time consuming, it is often not possible to prepare sufficient data. One of the approaches to perform any task with limited data is to use the learned weights of another task as initial weights of the model and finetune the model with actual limited data. In the image classification, the weights learned on large-scale dataset such as ImageNet are used as initial weights to perform another task. However, in such initialization, if the dataset domain differs from the target domain, there is a report that it is not a factor of the accuracy improvement and it is necessary to learn a model on the pre-training dataset which is suitable to the target domain. Therefore, in this study, we propose a method to create a pre-training dataset easily suitable for the target domain by utilizing public dataset. Experiment results show that proposed method improves the classification accuracy of "High" and "Low" of headlights beam from nighttime in-vehicle camera image.

Visual saliency detection is a useful technique for predicting, which regions humans will tend to gaze upon in any given image. Over the last several decades, numerous algorithms for automatic saliency detection have been proposed and shown to work well on both synthetic and natural images. However, two key challenges remain largely unaddressed: 1) How to improve the relatively low predictive performance for images that contain large objects and 2) how to perform saliency detection on a wider variety of images from various categories without training. In this paper, we propose a new saliency detection algorithm that addresses these challenges. Our model first detects potentially salient regions based on multiscale extrema of local perceived color differences measured in the CIELAB color space. These extrema are highly effective for estimating the locations, sizes, and saliency levels of candidate regions. The local saliency candidates are further refined via two global extrema-based features, and then a Gaussian mixture is used to generate the final saliency map. Experimental validation on the extensive CAT2000 data set demonstrates that our proposed method either outperforms or is highly competitive with prior approaches, and can perform well across different categories and object sizes, while remaining training-free.

The Helmholtz–Kohlrausch (H-K) effect is the phenomenon in which two color stimuli have the same luminance but different chroma in a certain hue, so the perceived brightness induced by the two stimuli are different. In expanding gamut, it is necessary to consider the H-K effect. A quantification of the H-K effect is required in order to evaluate and develop display devices for which the change of perceived brightness of gamut expansion must be considered. For quantification of the H-K effect, prediction equations that can derive the equivalent luminance in a single color image have been proposed in previous studies. However, these equations have not been applied to natural images that are important. Therefore, the purpose of this study is to quantify the H-K effect by deriving calculated values for natural images. For this purpose, first, we conducted the quantification of the H-K effect in natural images by deriving the equivalent luminance as calculated values expanding the three prediction equations proposed in previous studies. Next, we carried out a subjective evaluation experiment by varying image's chroma and luminance. We then verified the effectiveness of the calculated values by comparing them with the result from the experiment.

The color gamut is one of the critical parameters that dictate the image quality of displays. The liquid crystal displays using white color light-emitting diodes (LEDs) as the backlight, though having been widely employed recently, are not very satisfactory in terms of their color gamut because of the broad spectrum inherent to white LEDs. This prompted the authors to develop improved liquid crystal displays using an edge-lit wide color gamut backlight that used red laser diodes and cyan LEDs. Generating laser beams with high color purity, the laser diodes are light sources with a significant effect on expanding the color gamut. However, laser diodes, red ones in particular, have unfavorable thermal characteristics. To cope with this shortcoming, the authors clearly defined the restrictive criteria for laying out two kinds of light source on the edge-lit backlight and made a prototype 55-type laser backlight for performance evaluation.

The purity discrimination threshold of monochromatic light and mixed light was investigated in the mesopic condition. Monochromatic light and mixed light were first resolved into their components of pure color light of dominant wavelength and white light. By Using the definition of colorimetric purity, the luminance ratios of the pure color light to white light were obtained. The pure color light component in mixed light was calculated Using the luminance ratios. The purity discrimination threshold of monochromatic and mixed light is expressed by the ratio of the pure color component to the white light of a reference field. Therefore, this purity discrimination threshold can be expressed in a unified manner.