Abstract: An image processing apparatus includes a detector, an estimator and a determiner. The detector detects a candidate region of a captured image captured by a camera, the candidate region serving as a candidate for a water drop region affected by a water drop on the lens of the camera, based on an edge strength of each pixel in the captured image. The estimator estimates, based on the candidate region, a circle that includes the candidate region. The determiner determines whether or not the candidate region is part of the water drop region based on the edge strength of some of the pixels in the circle.

Abstract: A driving support device according to an embodiment includes an estimation unit, a rendering unit, and an erasing unit. The estimation unit estimates a running state of a movable body based on an image taken on the movable body. The rendering unit produces a prediction line that indicates a predicted direction of travel of the movable body based on the running state estimated by the estimation unit, and superimposes and renders the prediction line on the taken image. The erasing unit erases the prediction line rendered by the rendering unit in a case where the estimation unit estimates that the movable body is stopped.

Abstract: A water-droplet detecting apparatus according to an embodiment includes a setting unit, a first extraction unit, a second extraction unit, and a determination unit. The setting unit sets concentric circles having a center at arbitrary one point of a captured image of an image capturing unit. The first extraction unit extracts candidate pixels based on gradients of pixels on a circumference of each of the concentric circles. The candidate pixels are candidates for pixels that are estimated to indicate a water droplet adhered to the image capturing unit. The second extraction unit extracts one or more candidate circles based on the candidate pixels extracted by the first extraction unit. The one or more candidate circles are candidates for circles that indicate a shape of the water droplet. The determination unit determines whether or not the water droplet is adhered based on an extraction result of the second extraction unit.

Abstract: A driving support device according to an embodiment includes an estimation unit, a rendering unit, and an erasing unit. The estimation unit estimates a running state of a movable body based on an image taken on the movable body. The rendering unit produces a prediction line that indicates a predicted direction of travel of the movable body based on the running state estimated by the estimation unit, and superimposes and renders the prediction line on the taken image. The erasing unit erases the prediction line rendered by the rendering unit in a case where the estimation unit estimates that the movable body is stopped.

Abstract: There is provided an object alarm device and an object alarm method, which can surely alarm a driver of a high-risk target object. The object alarm device according to the embodiment includes a detecting unit, a selecting unit and, an alarm information generating unit. The detecting unit detects one or more objects from images captured on a movable body. The selecting unit selects a moving object and a still object from among the objects detected by the detecting unit, in which the still object has a contact possibility with the movable body. The alarm information generating unit generates alarm information on the objects selected by the selecting unit.

Abstract: A water-droplet detecting apparatus according to an embodiment includes a setting unit, a first extraction unit, a second extraction unit, and a determination unit. The setting unit sets concentric circles having a center at arbitrary one point of a captured image of an image capturing unit. The first extraction unit extracts candidate pixels based on gradients of pixels on a circumference of each of the concentric circles. The candidate pixels are candidates for pixels that are estimated to indicate a water droplet adhered to the image capturing unit. The second extraction unit extracts one or more candidate circles based on the candidate pixels extracted by the first extraction unit. The one or more candidate circles are candidates for circles that indicate a shape of the water droplet. The determination unit determines whether or not the water droplet is adhered based on an extraction result of the second extraction unit.

Abstract: An image processing apparatus includes a detector, an estimator and a determiner. The detector detects a candidate region of a captured image captured by a camera, the candidate region serving as a candidate for a water drop region affected by a water drop on the lens of the camera, based on an edge strength of each pixel in the captured image. The estimator estimates, based on the candidate region, a circle that includes the candidate region. The determiner determines whether or not the candidate region is part of the water drop region based on the edge strength of some of the pixels in the circle.

Abstract: There is provided an object alarm device and an object alarm method, which can surely alarm a driver of a high-risk target object. The object alarm device according to the embodiment includes a detecting unit, a selecting unit and, an alarm information generating unit. The detecting unit detects one or more objects from images captured on a movable body. The selecting unit selects a moving object and a still object from among the objects detected by the detecting unit, in which the still object has a contact possibility with the movable body. The alarm information generating unit generates alarm information on the objects selected by the selecting unit.

Abstract: A driving support device according to an embodiment includes an estimation unit, a rendering unit, and an erasing unit. The estimation unit estimates a running state of a movable body based on an image taken on the movable body. The rendering unit produces a prediction line that indicates a predicted direction of travel of the movable body based on the running state estimated by the estimation unit, and superimposes and renders the prediction line on the taken image. The erasing unit erases the prediction line rendered by the rendering unit in a case where the estimation unit estimates that the movable body is stopped.