JP7345106B2 - Visibility support image generation device and image conversion program - Google Patents

Description

TECHNICAL FIELD The present invention relates to a device that generates a visibility support image to be displayed to a vehicle occupant, and an image conversion program.

Some vehicle door mirrors use an aspherical mirror, which is an optical mirror, in order to reduce blind spots in the lateral direction (left and right directions as seen from the driver of the vehicle, directions away from the vehicle). FIG. 1 is a schematic diagram of an optical aspherical mirror according to existing technology, in which a door mirror M includes a normal mirror portion M1 and an aspherical portion M2. The aspherical portion M2 changes its curvature so that a wider range of images can be seen in the lateral direction.

Even in a CMS (camera monitoring system) in which the door mirror M is further equipped with an imaging camera, the angle of view is widened in the lateral direction based on a principle imitating the aspherical mirror, which is the optical mirror described above. That is, in the image captured by the camera, a region corresponding to the aspherical portion M2 of the door mirror M is compressed in the lateral direction, and the compressed image is displayed on the display.

Patent Document 1 discloses that when displaying on a display an image obtained by enlarging or compressing a captured image of the rear side of the vehicle at a magnification that changes in the lateral direction, A horizontal scale that corrects the scale of each part of the captured image in the horizontal direction using a set horizontal enlargement/contraction magnification to prevent driving area demarcation lines such as white lines from appearing crooked in the image displayed on the display. Correction processing and further partial vertical scale correction using a vertical scaling factor set so that the image of the driving area dividing line, which is displayed crooked in the image with the horizontal scale corrected, extends in a straight line. It is disclosed that vertical scale correction processing is performed.

JP2013-85142A

By the way, when considered from the perspective of a vehicle occupant, particularly a driver (hereinafter referred to as a driver, etc.), the driver and the like view a display image displayed on an in-vehicle monitor or the like. The original image of this display image is taken by a camera installed in a door mirror or the like. It is preferable that this display image be displayed in a state with little discomfort.

Further, since the door mirror is used to grasp the situation outside the vehicle (hereinafter referred to as the outside world), it is also desirable that the displayed image is capable of appropriately understanding the situation of the outside world.

In view of the above, an object of the present disclosure is to provide a display image that causes less discomfort and allows an appropriate understanding of the external situation.

A visibility support image generation device that generates a visibility support image of a vehicle includes a camera that captures an image from the vehicle, and a processing unit, and the processing unit converts the captured image captured by the camera. A visibility support image is generated, and the image conversion is performed so that a compression rate of the captured image in the horizontal direction is higher than a compression rate of the captured image in the vertical direction around a deep vanishing point included in the captured image. In this way, the captured image is compressed. With the above-mentioned configuration, the generated visual field support image has less discomfort and allows the user to appropriately grasp the situation of the outside world.

It is possible to provide a display image that does not give a sense of discomfort and allows an appropriate understanding of the external situation.

A schematic diagram of an optical aspherical mirror, which is an existing technology. 2A and 2B are schematic diagrams illustrating the principle of image generation by the visibility support image generation device 1 of the present disclosure, in which (a) a diagram illustrating an example of a lens model; (b) a diagram illustrating a captured image (input image); and (c) a display image. (Output image). FIG. 1 is a configuration diagram showing an example of a visibility support image generation device 1 of the present disclosure. FIG. 3 is a flow diagram showing an example of image processing performed by the processing unit 11. FIG. An explanatory diagram of a perfect circular lens model. FIG. 3 is a comparison diagram between a conventional technique and a visibility support image generation device 1 according to the present disclosure when an output image I _out is generated from an input image I _in . FIG. 7 is a second comparison diagram between the conventional technology and the visibility support image generation device 1 of the present disclosure when an output image I _out is generated from an input image I _in . FIG. 7 is a third comparison diagram between the conventional technology and the visibility support image generation device 1 of the present disclosure when an output image I _out is generated from an input image I _in .

The following explanation assumes that the vehicle is a right-hand drive car, and that the image taken by the camera installed on the right side door mirror, which tends to be the driver's blind spot, is used to display the display image on the display device installed inside the car. A detailed description will be given with reference to the drawings as appropriate. However, it is not intended that the claimed subject matter be limited solely to this premise. For example, various types of objects such as moving objects other than vehicles, the position of the steering wheel of the vehicle (left steering wheel, right steering wheel, automatic driving where there is no steering wheel), the mounting position of the camera (left door mirror, right door mirror, etc.), etc. Deformations are possible.

The accompanying drawings and the following description are provided to enable those skilled in the art to fully understand the present disclosure, and are not intended to limit the claimed subject matter.

FIG. 2 is a schematic diagram showing the principle of image generation by the visibility support image generation device 1 (described later based on FIG. 3) of the present disclosure, in which (a) a diagram showing an example of a lens model, (b) a captured image (input (c) A diagram showing a display image (output image).

In the visibility support image generation device 1 of the present disclosure, the compression rate in the horizontal direction of the captured image is higher than the compression rate in the vertical direction of the captured image, centering on the deep vanishing point included in the captured image. A geometric lens model that can compress the captured image is used. In this embodiment, a vertically elongated elliptical lens model E shown in FIG. 2(a) is used as the geometric lens model.

If the length of the long axis (vertical axis) of this vertically elongated elliptical lens model E is b and the length of the short axis (horizontal axis) is c, then b>c. As will be described later, when this vertically elongated elliptical lens model E is used, the compression rate of the image changes linearly from the horizontal direction to the vertical direction. In other words, if the deflection angle is 0 degrees in the horizontal direction and the deflection angle is 90 degrees in the vertical direction, the compression ratio decreases linearly as the deflection angle increases from 0 degrees to 90 degrees. Also, as will be described later, the compression ratio increases as the distance from the center of the ellipse increases.

FIG. 2(b) is a diagram showing the input image I _in , which corresponds to a captured image captured by a camera attached to the right door mirror of the own vehicle 100 in this embodiment.

In the input image I _in , the body of the own vehicle 100 is reflected on the left side thereof. Also, although this is just an example for explanation, objects OBJ1 to OBJ4 are reflected in the input image I _in . In this example, each of the objects OBJ1 to OBJ4 is a vehicle different from the own vehicle 100 that is traveling in the lane next to the own vehicle 100.

FIG. 2(c) is a diagram showing an output image I _out , which in this embodiment corresponds to a display image displayed toward the driver holding the right steering wheel by a monitor installed inside the host vehicle 100. do.

As can be seen from FIG. 2(c), the body of the host vehicle 100 and objects OBJ1 to OBJ4 are also reflected in the output image I _out . Here, when comparing the input image I _in (FIG. 2(b)) and the output image I _out (FIG. 2(c)), the objects OBJ1 to OBJ4 reflected in the output image I _out are compressed in the horizontal direction. The vertical compression ratio is higher than the vertical compression ratio. In other words, the angle of view widens in the horizontal direction. Note that the fact that such an output image I _out is a display image that does not give a sense of discomfort and allows an appropriate grasp of the external situation will be described later.

FIG. 3 is a configuration diagram showing an example of the visibility support image generation device 1 of the present disclosure. The visibility support image generation device 1 of the present disclosure includes a processing section 11 and a camera 12. Note that components other than these may be included. As illustrated, the visibility support image generation device 1 may further include a nonvolatile memory 14 and the like.

The processing unit 11 is a component that performs information processing in the visibility support image generation device 1. The processing unit 11 performs image processing, processes commands and signals input from other components within the device and from outside the device, and conversely sends commands and signals to other components within the device and outside the device. You may send it.

The camera 12 is a means for capturing an image from the vehicle to obtain the above-mentioned input image I _in . In this embodiment, the camera 12 is a camera attached to a door mirror for capturing an image of the rear side of the vehicle, but the camera 12 is not limited thereto. For example, it may be a camera that captures images of the front or rear of the vehicle.

The display device 13 is a device that can display the display image generated by the visibility support image generation device 1. Typically, the display device 13 is a monitor or the like provided within the own vehicle 100, but is not limited thereto. The driver and the like will see the display image displayed on this display device 13. In FIG. 3, the display device 13 is separate from the visibility support image generation device 1. However, the display device 13 may be included in the visibility support image generation device 1.

The nonvolatile memory 14 may store programs used for image processing performed by the processing unit 11, various parameter information, conversion tables based on a perfect circular lens model (described later), and the like.

Note that the components included in the visibility support image generation device 1 may be further integrated, or conversely may be further divided into a plurality of subcomponents.

FIG. 4 is a flow diagram showing an example of image processing performed by the processing unit 11.

FIG. 4 shows an example of a process for creating a conversion table used in an image conversion process to generate an output image I _out based on an input image I _in . As a precondition, in this example, the vertically elongated elliptical lens model E shown in FIG. 2(a) is used. Furthermore, as will be described later based on FIG. 5, in this example, lens distortion is also removed.

In step S01, the processing unit 11 captures the input image I _in . The captured image may be held in a memory (not shown) or the like.

In step S02, the processing unit 11 determines whether the next pixel to be processed remains. If there is a next pixel (yes in the figure), the process advances to step S03. If the next pixel does not remain (no in the figure), this indicates that the processing has been completed for all of the input image I _in , and the conversion table creation process ends.

In step S03, the processing unit 11 selects the next pixel to be processed. For convenience, the coordinates of the selected pixel are referred to as coordinates P.

In step S04, the processing unit 11 calculates the distance Do between the central coordinate O and the coordinate P. Here, the central coordinate O means a deep vanishing point included within the input image I _in . For example, to illustrate using FIG. 2(b), a plurality of white lines W are drawn on the road surface on which a car is driving. There is a point on the extension line of these plurality of white lines W where the extension lines intersect at one point. This intersection is the deep vanishing point. In FIG. 2(b), there is a deep vanishing point (central coordinate O) inside the image, on the upper left side.

In step S05, the processing unit 11 converts the distance Do using a conversion table of a perfect circular lens model, and determines the converted distance Do'. This step S05 will be described in further detail below with reference to FIG. 5 as well.

FIG. 5 shows an explanatory diagram of a perfect circular lens model. A lens 50 shown in FIG. 5 is a typical perfectly circular lens used in pinhole cameras and the like. Note that FIG. 5 shows the perfect circular lens viewed not from the front but from the side.

The value of the image position on the evaluation surface 51 of the optical system expressed as a distance from the optical axis is called an image height, and there are two types of image height: an ideal image height 53 and a real image height 52. The ideal image height 53 is an ideal image height. However, the image height of a normal optical system is not the ideal image height 53 because it is affected by lens distortion and the like. On the other hand, the real image height 52 is an image height that indicates the position where the image is actually formed on the evaluation surface. Note that the distance Do between the central coordinate O and the coordinate P calculated in step S04 corresponds to the real image height 52.

Then, the real image height 52 at each coordinate and the ideal image height 53 at each coordinate are combined to form one conversion table. In the example shown in FIG. 5, the lens 50 is a perfect circle lens. Therefore, a conversion table that combines the real image height 52 shown in FIG. 5 and the ideal image height 53 of a perfect circular lens becomes a conversion table for a perfect circular lens model. By using the conversion table for the perfect circular lens model, it is possible to convert the real image height 52 affected by lens distortion to the ideal image height 53 (of the perfect circular lens model) from which lens distortion has been removed.

The conversion table for the perfect circular lens model as described above may be stored, for example, in the nonvolatile memory 14 or the like. Then, in step S05, the processing unit 11 converts the distance Do, which is the real image height 52, into a converted distance Do' corresponding to the ideal image height 53 of the perfect circular lens model, using the conversion table of the perfect circular lens model. Determine.

In step S06, the processing unit 11 calculates the image height change rate a based on the perfect circular lens model. Note that a=(Do'/Do)-1.

The image height change rate a is the value obtained by dividing the distance Do', which is the ideal image height 53 of the perfect circular lens model, by the distance Do, which is the real image height 52, and subtracting 1 from the distance Do'. For example, when the distance Do', which is the ideal image height 53, is 120, and the distance Do, which is the real image height 52, is 100, the image height change rate a=(120/100)-1=0.2. This means that the ideal image height 53 (of the perfect circular lens model) changes (increases in this case) by 20% with respect to the real image height 52.

Here, the visibility support image generation device 1 according to the present disclosure uses the vertically elongated elliptical lens model E (FIG. 2(a)), as already described. Therefore, it is necessary to include an element of a vertically long ellipse in the image height change rate a calculated in step S06. Therefore, in the subsequent step S07 and subsequent steps, the following processing is performed.

In step S07, the processing unit 11 calculates the coordinates of an intersection point P1 between a straight line extending from the coordinate O to the coordinate P and the vertical elliptic function E defined by the major axis b and the minor axis c. In the next step S08, the processing unit 11 calculates the distance D1 between the coordinate O and the coordinate P1.

Note that the lengths of the major axis b and the minor axis c can be determined as appropriate. In this example, since a vertically elongated elliptical lens model E is used, the major axis b extends in the vertical direction (Y-axis direction), and the minor axis c extends in the horizontal direction (X-axis direction). c<b.

Here, two specific examples will be shown for easier understanding.

First, a first specific example will be shown. Assume that the coordinate P exists on the X axis with the coordinate O as the origin. In other words, it is assumed that the coordinates P=(m, 0). m is any positive real number. At this time, the coordinates of the intersection P1 between the straight line extending from the coordinate O to the coordinate P and the vertically elongated elliptic function E are (c/2, 0), so D1=c/2.

A second specific example will be shown. Assume that the coordinate P exists on the Y axis when the coordinate O is the origin. In other words, it is assumed that the coordinates P=(0, n). n is any positive real number. At this time, the coordinates of the intersection P1 between the straight line extending from the coordinate O to the coordinate P and the vertically elongated elliptic function E are (0, b/2), so D1=b/2.

Here, when comparing the above two specific examples, since c<b, c/2<b/2. That is, the value of the distance D1 is larger in the second specific example than in the first specific example.

In the following step S09, the processing unit 11 calculates the compression coefficient A _p =(a/D1)+1 for the coordinate P. As described above, a is the image height change rate, and D1 is the distance from the coordinate O to the coordinate P1. Since D1 is used as a reciprocal number here, the magnitude relationship of the values is reversed between the above two specific examples. That is, the value of the compression coefficient A _p is larger in A _p =(2a/c)+1 in the first specific example than in A _p =(2a/b)+1 in the second specific example. This compression coefficient A _p corresponds to the compression ratio.

In other words, the first concrete example with horizontal coordinates P = (m, 0) has a higher compression coefficient A than the second concrete example with vertical coordinates P = (0, n). The value of _p is large. This indicates that the compression ratio in the horizontal direction is higher than the compression ratio in the vertical direction around the coordinate O (deep vanishing point).

In subsequent step S10, the processing unit 11 writes the compression coefficient A _p for the coordinate P into the conversion table of the vertically long elliptical lens model E as the ideal image height 53 of the vertically long elliptical lens model. That is, the real image height 52 and the ideal image height 53 of the vertically long elliptical lens model E are combined and recorded. As can be seen from the above description, the conversion table for the vertically elongated elliptical lens model E is a modified version of the conversion table for the perfect circular lens model described above.

The process then returns to step S02. In other words, the combination of the real image height 52 and the ideal image height 53 of the vertically elongated elliptical lens model E for all pixels (coordinates P) included in the input image I _in captured by the processing unit 11 in step S01 described above. is recorded and used as a conversion table for the vertically elongated elliptical lens model E.

In the manner described above, a conversion table for the vertically elongated elliptical lens model E can be created by transforming the conversion table for the perfect circular lens model.

Then, by applying the conversion table of the vertically long elliptical lens model E to the input image I _in , an output image I _out can be generated. More specifically, the line segment from coordinate O to coordinate P in the input image I _in is compressed by A _p times (distance Do is changed to distance Do/A _p ). In this compression, the compression ratio in the horizontal direction centered on the coordinate O (deep vanishing point) is higher than the compression ratio in the vertical direction.

Note that the lengths of the long axis b and short axis c of the vertically elongated elliptical lens model E can be changed as appropriate. That is, the visibility support image generation device 1 of the present disclosure can separately and independently adjust the compression ratio in the horizontal direction and the compression ratio in the vertical direction centering on the coordinate O (deep vanishing point).

FIG. 6 is a comparison diagram between the conventional technology and the visibility support image generation device 1 of the present disclosure when generating the output image I _out from the input image I _in . FIG. 6(a) shows the input image _Iin . FIG. 6(b) shows an output image _Iout generated using the conventional technique for this input image _Iin . FIG. 6C shows an output image I _out generated using the visibility support image generation device 1 of the present disclosure for the same input image I _in .

In the input image I in shown _in FIG. 6A, the body of the own vehicle 100 is reflected on the left side thereof. Further, although this is an example for explanation, objects OBJ1 to OBJ4 are reflected in the input image I _in . In this example, objects OBJ1 to OBJ4 are vehicles different from the own vehicle 100 that are traveling in the lane next to the own vehicle.

The output image _{Iout generated} by the conventional technique shown in FIG. It is generated by further scaling in the direction. In other words, scaling is performed in two stages: horizontally and vertically. This also applies to FIG. 7(b) and FIG. 8(b), which will be described later.

Looking at the output image I _out shown in FIG. 6(b), the shapes of the vehicles OBJ1 and OBJ2 are significantly different from the input image I _in . These have been greatly crushed in the horizontal direction, making it difficult to recognize their original shape. Furthermore, the sense of distance in the lateral direction has also changed significantly, and the inter-vehicle distance from OBJ1 to OBJ2 has become extremely short. When looking at this from the perspective of a driver or the like, although it is possible to perceive that some kind of object exists there, it may be difficult to discern what that object actually is. Further, the inter-vehicle distance between OBJ1 and OBJ2 is originally sufficiently wide (FIG. 6(a)). However, in FIG. 6(b) they appear to be about to collide with each other.

On the other hand, in the case of the output image _Iout generated by the visibility support image generation device 1 of the present disclosure shown in FIG. 6(c), the change in the three-dimensional shape is alleviated. Further, the change in the sense of distance in the lateral direction is also gradual. Therefore, from the viewpoint of a driver or the like, it is easy to recognize the shapes of objects OBJ1 to OBJ4 moving in the adjacent lane. Further, there is no confusion due to sudden changes in the sense of distance.

FIG. 7 is a second comparison diagram between the conventional technology and the visibility support image generation device 1 of the present disclosure when generating the output image I _out from the input image I _in . FIG. 7(a) shows the input image _Iin . FIG. 7B shows an output image I _out generated using the conventional technique for this input image I _in . FIG. 7C shows an output image I _out generated using the visibility support image generation device 1 of the present disclosure for the same input image I _in .

In FIG. 7, straight lines L1 and L2 perpendicular to the traveling direction of the own vehicle 100 are added as auxiliary lines. A straight line L1 is a line near the own vehicle 100, and a straight line L2 is a line far away from the own vehicle 100.

In the input image I _in shown in FIG. 7(a), the two straight lines L1 and L2 are parallel.

Looking at the output image _Iout generated by the conventional technique shown in FIG. 7(b), the two straight lines L1 and L2 have greatly different slopes. This difference in inclination is perceived as image distortion or an unnatural feeling from the viewpoint of a driver or the like. Furthermore, the plurality of vehicles reflected in the image are traveling in a direction perpendicular to the two straight lines L1 and L2. In other words, if the output image I _out is a video, the vehicle reflected in the video should actually be traveling straight, but as it moves toward the right side of the image, it will curve. It looks like that. This also makes drivers and the like feel uncomfortable.

On the other hand, in the case of the output image _Iout generated by the visibility support image generation device 1 of the present disclosure shown in FIG. 7(c), the difference in slope between the two straight lines L1 and L2 is gentler than before become. Therefore, there is little sense of discomfort for drivers and the like. Further, when the output image I _out is a moving image, the progress of the vehicle reflected in the moving image looks more natural, so there is less discomfort.

FIG. 8 is a third comparison diagram between the conventional technology and the visibility support image generation device 1 of the present disclosure when generating the output image I _out from the input image I _in . FIG. 8(a) shows the input image _Iin . FIG. 8(b) shows an output image _Iout generated using the conventional technique for this input image _Iin . FIG. 8C shows an output image I _out generated using the visibility support image generation device 1 of the present disclosure for the same input image I _in .

In FIG. 8, an explanatory arrow L3 is drawn near the front bumper of the vehicle OBJ2. This arrow L3 indicates a straight line perpendicular to the traveling direction of the own vehicle 100, which is provided by an object reflected in the image.

Looking at the output image _Iout generated by the conventional technique shown in FIG. 8(b), it can be seen that the arrow L3 is curved. In other words, from the viewpoint of the driver or the like, the linearly formed members (such as the front bumper) that constitute the object OBJ2 appear curved. This also makes drivers and the like feel uncomfortable.

On the other hand, in the case of the output image I _out generated by the visibility support image generation device 1 of the present disclosure shown in FIG. However, it appears to drivers etc. as a shape that is close to a straight line. In other words, there is less discomfort for drivers and the like.

As exemplified above, the output image I _out generated by the visibility support image generation device 1 of the present disclosure is an image with less discomfort.

In addition, the output image I _out generated by the visibility support image generation device 1 of the present disclosure has a compression ratio in the horizontal direction of the captured image centered on the deep vanishing point included in the captured image. The captured image is compressed so that the compression ratio is higher than that in the vertical direction (see FIGS. 1 to 5). Therefore, the advantage of the optical aspherical mirror of expanding the angle of view in the lateral direction can still be enjoyed.

That is, the visibility support image generation device 1 of the present disclosure can generate an output image I _out that causes less discomfort and allows an appropriate grasp of the external situation.

In addition, some supplementary matters are explained below.

The flowchart explained based on FIG. 4 creates a conversion table for the vertically long elliptical lens model E based on the conversion table for the perfect circular lens model, and generates an output image I _out using the conversion table for the vertically long elliptical lens model E. It was an example. If the conversion table for this vertically elongated elliptical lens model E is stored in the nonvolatile memory 14 or the like, there is no need to generate a new conversion table each time the input image I _in is input. That is, it becomes possible to generate the output image I _out for the input image I _in by referring to the conversion table of the vertically long elliptical lens model E that has already been stored. Conversely, each time the input image I _in is input, the compression coefficient A _p may be dynamically calculated as described above, and then the output image I _out may be generated.

Furthermore, in the embodiment described above, the processing unit 11 performs the processing related to steps S01 to S10. A program related to this processing may be stored in the nonvolatile memory 14 or the like, and the processing unit 11 may read this program and perform image processing. On the other hand, the above-described processing may be performed by hardware processing instead of software processing. For example, the processing may be performed by a dedicated circuit or the like.

Next, a supplementary explanation regarding the compression ratio will be given. Two specific examples are shown for step S09 in FIG. That is, the compression ratio (compression coefficient A _p ) was compared in two directions: the horizontal direction (first specific example) and the vertical direction (second specific example). However, there is also a compression rate in the diagonal direction that is in between. As described above, the conversion table for the vertically elongated elliptical lens model E is a modification of the conversion table for the perfect circular lens model described above. Since a perfect circle is converted into a vertically elongated ellipse, the compression ratio (compression coefficient A _p ) changes linearly from the horizontal direction to the vertical direction (see FIG. 2(a)). In other words, when the yaw angle is 0 degrees in the horizontal direction and the yaw angle is 90 degrees in the vertical direction, the compression ratio (compression coefficient A _p ) decreases linearly as the yaw angle increases.

Further, as the perfect circular lens model, a lens model in which the compression rate increases as the distance from the center increases can be used. The vertically elongated elliptical lens model E created based on such a perfect circular lens model also has a compression ratio that increases as the distance from the center (deep vanishing point) increases.

In the above configuration, in the image conversion, the compression ratio changes linearly as the declination changes from the horizontal direction of the captured image to the vertical direction of the captured image, with the deep vanishing point included in the captured image as the center. It may be something that does. If it is a linear change, the compression ratio will gradually change depending on the declination angle, so the generated visibility support image will also be natural.

In the above configuration, the image conversion may be such that the compression rate increases as the distance from the deep vanishing point increases. With this configuration, it is possible to naturally widen the angle of view while maintaining the amount of image information by applying low compression to the peripheral area of the deep vanishing point.

In the above configuration, the image conversion may be compression using a vertically elongated elliptical lens model. If it is a vertically elongated elliptical model, the shape is similar to a normally used perfect circular lens model, and there is less discomfort in the displayed image after image conversion. Further, whereas the radius r was the only parameter that defined a perfect circle, in the case of a vertically elongated ellipse, there can be two parameters, the major axis b and the minor axis c. By appropriately adjusting these two parameters, it is possible to flexibly vary the compression ratio of the image in the vertical and horizontal directions while reducing the unnatural feeling of the displayed image that conventionally occurs.

In the above configuration, the visibility support image generation device 1 may further include a display unit that displays the visibility support image. By displaying the visibility support image on such a display unit, the driver and the like can view a natural visual support image with a wider field of view in the lateral direction.

The present disclosure also relates to an image conversion program for generating a visibility assistance image of a vehicle. The image conversion program causes a processing unit included in the device to convert the real image height of the image and the geometric lens model based on data indicating the correspondence between the real image height of the image and the ideal image height of the perfect circular lens model. and calculating a compression ratio of each pixel included in the image based on the correspondence between the real image height of the image and the ideal image height of the geometric lens model. and a step of compressing the image based on the compression rate of each pixel, and the geometric shape is such that the length in the horizontal direction from the center of the geometric shape is equal to or smaller than the geometric shape. The length may be shorter than the length in the vertical direction from the center of the shape. With the above-mentioned configuration, it is possible to provide a visibility support image that does not give a sense of discomfort to the driver or the like and allows an appropriate grasp of the external situation from an input image captured by a camera installed on a door mirror or the like of the vehicle.

In the above configuration, the geometric shape may be such that the distance from the center of the geometric shape changes linearly as the deviation angle of the geometric shape changes from the horizontal direction to the vertical direction. If it is a linear change, the compression rate will gradually change depending on the declination angle, so the image generated using the program will also look natural to drivers and the like.

In the above configuration, the geometrical shape may be a vertically elongated ellipse. If it is a vertically elongated elliptical model, the shape is similar to a normally used perfect circular lens model, and there will be less discomfort in the visual field support image after image conversion. Further, whereas the radius r was the only parameter that defined a perfect circle, in the case of a vertically elongated ellipse, there can be two parameters, the major axis b and the minor axis c. By appropriately adjusting these two parameters, it is possible to flexibly vary the compression ratio of the image in the vertical and horizontal directions while reducing the unnatural feeling of the displayed image that conventionally occurs.

Although various embodiments have been described above with reference to the drawings, it goes without saying that the present invention is not limited to such examples. It is clear that those skilled in the art can come up with various changes or modifications within the scope of the claims, and these naturally fall within the technical scope of the present invention. Understood. Further, each component in the above embodiments may be arbitrarily combined without departing from the spirit of the invention.

1 Visibility support image generation device 11 Processing unit 12 Camera 13 Display device 14 Non-volatile memory 50 Lens 51 Evaluation surface 52 Real image height 53 Ideal image height 100 Own vehicle I _In input image I _Out output image E Vertical elliptical lens model M Door mirror M1 Mirror Part M2 Aspherical part OBJ1 to OBJ4 Object

Claims (4)

A visibility support image generation device that generates a visibility support image of a vehicle, the device comprising:
A camera that captures images from the vehicle;
Equipped with a processing section,
The processing unit converts the captured image captured by the camera to generate the visibility support image,
The image conversion is performed by converting the captured image so that the compression rate in the horizontal direction of the captured image is higher than the compression rate in the vertical direction of the captured image, with a deep vanishing point included in the captured image as the center. It compresses
The image conversion is compression using a vertically elongated elliptical lens model,
Vision support image generation device. The visibility support image generation device according to claim 1 ,
In the image conversion, the compression rate increases as the distance from the deep vanishing point increases.
Vision support image generation device. The visibility support image generation device according to claim 1 or 2 ,
further comprising a display unit that displays the visibility support image;
Vision support image generation device. An image conversion program for generating a visibility support image of a vehicle, the program comprising:
In the processing section of the device,
calculating the correspondence between the real image height and the ideal image height of the geometric lens model based on data indicating the correspondence between the real image height of the perfect circle lens and the ideal image height of the perfect circle lens model; ,
calculating a compression ratio of each pixel included in the image based on the correspondence between the real image height and the ideal image height of the geometric lens model;
compressing the image based on the compression rate of each pixel;
run the
The geometric shape has a length in the horizontal direction from the center of the geometric shape that is shorter than a length in the vertical direction from the center of the geometric shape,
The compression is compression using a vertically elongated elliptical lens model,
Image conversion program.

Images