CN101726855B - Correction method of fisheye image distortion on basis of cubic projection - Google Patents

Abstract

The invention belongs to the field of advanced manufacturing technology, relating to a correction method of fisheye image distortion on the basis of cubic projection. The correction method comprises the following steps: 1. using a fisheye lens to obtain an imaging model, and calibrating the fisheye lens according to the imaging model to obtain a calibration parameter; 2. according to the collected fisheye image, determining a viewpoint, taking the viewpoint as the original point of a coordinate system, and building a cubic perspective projection model to obtain the cubic perspective projection of a spatial point; 3. building a mapping relation between the cubic perspective projection of the spatial point and the fisheye image; and 4. using bilinearity interpolation to realize the distortion correction of the fisheye image. The invention utilizes a cubic projection model to correct distorted fisheye images and can effectively overcome the geometric distortion existing in the original fisheye images, and the corrected image conforms to the intuitional feeling of people and has strong sense of reality.

Description

Based on cubic projection to the fish eye images distortion correction method

Technical field

The invention belongs to the advanced manufacturing technology field, relate to how much antidotes of a kind of fish eye lens fish eye lens picture distortion.

Background technology

The panorama picture of fisheye lens stereo visual system is at the micro robot vision guided navigation and closely have a wide range of applications in large field object identification and the location. nearly 180 ° of fish eye lens camera field of view corner connection, can realize closely or the stereoscopic vision perception of the big visual field of super close distance, but also introduced very serious anamorphose simultaneously. the main cause of anamorphose is that fish-eye real imaging surface is not the plane, but be similar to spherical curved surface. therefore, panorama picture of fisheye lens is not desirable perspective projection, can not demarcate with conventional camera marking method.This will use have big visual field (field of view) if wide-angle or flake video camera. but use but the image that uses wide-angle or flake video camera to take has very serious deformation. we want to utilize these to have the perspective projection information of gross distortion image, so just these deformation patterns need be proofreaied and correct the perspective projection image that be accustomed to for people.

At present, the method for fish eye images distortion correction has a lot, wherein based on sphere model and the most frequently used based on the fish eye lens rectification antidote of cylinder model.

Summary of the invention

The object of the present invention is to provide how much antidotes of a kind of fish eye lens picture distortion.This method at first obtains imaging model with fish eye lens, and fish eye lens is demarcated according to imaging model, obtain calibrating parameters, utilizing the fisheye projection curve then, use cubic projection, determine mapping relations according to imaging plane, fish eye images is carried out distortion correction, the image splicing after will proofreading and correct at last.

Of the present invention based on cubic projection to the fish eye images distortion correction method, comprise the steps:

Step 1: obtain imaging model with fish eye lens, and according to imaging model fish eye lens is demarcated and to be obtained calibrating parameters;

Step 2: according to the fish eye images that collects, determine viewpoint, and be coordinate origin, set up cube perspective projection model, obtain the cube perspective projection of spatial point with the viewpoint;

Step 3: set up the cube perspective projection of spatial point and the mapping relations between the fish eye images;

Step 4: use bilinear interpolation value method, realize distortion correction to fish eye images.

In the above-mentioned step 4, establish spatial point p (x ₀, y ₀, z ₀) through cubic projection model Nonlinear Mapping be on the fish eye images some m (x, y), the floating-point coordinate of this point is (i+u, j+v), (i ∈ Z+, j ∈ Z+, u ∈ [0,1), v ∈ [0,1)), i wherein, j is the integer part of floating-point coordinate, u, v is the fractional part of floating-point coordinate, it is value [0,1) interval floating number is with f (x, image coordinate (x after y) expression is proofreaied and correct, y) pixel value of locating, then m (x, pixel value y) is according to f (i+u, j+v)=(1-u) (1-v) f (i, j)+(1-u) vf (i, j+1)+and u (1-v) f (i+1, j)+(i+1 j+1) tries to achieve uvf.

The present invention's fish eye images of correcting distortion based on the cubic projection model can effectively overcome the geometry deformation that original fish eye images exists, and makes the image after proofreading and correct meet people's visual sense, and the sense of reality is stronger.

Description of drawings

Fig. 1 has identified the view coordinate of cubic projection.

The cubical perspective projection model of Fig. 2.

Fig. 3 fish eye images polar coordinate system.

Fig. 4 bilinear interpolation is calculated schematic diagram.

Embodiment

Step 1: obtain imaging model with fish eye lens, and according to imaging model fish eye lens is demarcated and to be obtained calibrating parameters.This step is demarcated according to fish-eye concrete parameter according to the scaling method of routine.

Step 2: according to the fish eye images that collects, determining viewpoint, is initial point with the viewpoint, sets up cube perspective projection model.

Referring to Fig. 1, set up coordinate system earlier, owing to adopt cubic projection, and there are 6 faces a cube surface, so need carry out segmentation to representing cubical function.

If cube center (being the viewpoint of the fish eye images gathered) is the view coordinate initial point, a is the cubical length of side.

Then the equation in coordinates of 6 faces of cube is respectively:

ABFE： $\{x=\frac{a}{2},(-\frac{a}{2}\≤y,z\≤\frac{a}{2})\}$

DCGH： $\{x=-\frac{a}{2},(-\frac{a}{2}\≤y,z\≤\frac{a}{2})\}$

ADHE： $\{y=-\frac{a}{2},(-\frac{a}{2}\≤x,z\≤\frac{a}{2})\}$

BCGF： $\{y=\frac{a}{2},(-\frac{a}{2}\≤x,z\≤\frac{a}{2})\}$

HEFG： $\{z=\frac{a}{2},(-\frac{a}{2}\≤x,y\≤\frac{a}{2})\}$

ABCD： $\{z=-\frac{a}{2},(-\frac{a}{2}\≤x,y\≤\frac{a}{2})\}$

Space any point p (x ₀, y ₀, z ₀) with viewpoint (0,0,0) line equation be:

$\frac{x}{x_0}=\frac{y}{y_0}=\frac{\mathrm{<figure-callout\; id="0"\; label="z\; z"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">z</figure-callout>}}{z_{}}$

Calculate corresponding points on each face of a p and cube, the coordinate of corresponding points is expressed as respectively on each face of cube:

ABFE： $(\frac{a}{2},\frac{{\mathrm{<figure-callout\; id="0"\; label="ay"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ay</figure-callout>}}_0}{{2x}_0},\frac{{\mathrm{az}}_0}{{2x}_0})$

DCGH： $(-\frac{a}{2},-\frac{{\mathrm{<figure-callout\; id="0"\; label="ay"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ay</figure-callout>}}_0}{{2x}_0},\frac{{\mathrm{az}}_0}{{2x}_0})$

ADHE： $(-\frac{{\mathrm{<figure-callout\; id="0"\; label="ax"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ax</figure-callout>}}_0}{{2y}_0},-\frac{a}{2},-\frac{{\mathrm{az}}_0}{{2y}_0})$

BCGF： $(\frac{{\mathrm{ax}}_0}{{2y}_0},\frac{a}{2},\frac{{\mathrm{az}}_0}{{2y}_0})$

HEFG： $(\frac{{\mathrm{ax}}_0}{{2z}_0},\frac{{\mathrm{<figure-callout\; id="0"\; label="ay"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ay</figure-callout>}}_0}{{2z}_0},\frac{a}{2})$

ABCD： $(-\frac{{\mathrm{<figure-callout\; id="0"\; label="ax"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ax</figure-callout>}}_0}{{2z}_0},-\frac{{\mathrm{<figure-callout\; id="0"\; label="ay"\; filenames="H2009102285813E00011.png"\; state="\{\{state\}\}">ay</figure-callout>}}_0}{{2z}_0},-\frac{a}{2})$

So just obtained cubical perspective projection model, and with spatial point

Behind known cubical projected image, it is as above shown in Figure 2 that the cube perspective projection is converted to plane projection.For on the cube a bit, such as the p on the ABCD face ₀Pixel coordinate be (X, Y, Z), then the p on its corresponding projection plane (x, pixel coordinate y) then is: $\begin{array}{c}x=X\\ y=Y\end{array},$ Similar calculating all can be done in remaining plane.

Step 3: set up the cube perspective projection of spatial point and the mapping relations between the fish eye images.

According to the plane projection of space cube perspective projection model cubic projection image and the mapping relations of fish eye images;

Can obtain the fish eye lens correction model by cube perspective projection model, fish eye images is carried out distortion correction.

If p is (x ₀, y ₀, z ₀) be any cube perspective projection point of space, the coordinate of its corresponding points on each face of cube is shown in step 1 two, because fish eye lens, this point is mapped to 1 m on the fish eye images plane, the rectangular coordinate of m is (x, y), the intersection point of main shaft z and fish eye images, promptly the rectangular coordinate of principal point is designated as (x _p, y _p), with principal point (x _p, y _p) (generally all recognize the center that is positioned at figure) set up polar coordinate system for the origin of coordinates of polar coordinate system, sees Fig. 3, then have $r=\sqrt{{(x-x_p)}^2+{(y-y_p)}^2},$

$\tan \mathrm{\&theta;}=\frac{y-y_p}{x-x_p}$

P (x ₀, y ₀, z ₀) according to step 1,2., 3 calculate the coordinate figure m make new advances, and (r θ) has just finished the one by one mapping of picture planar point to cube point, and then the gray value of this point is obtained a gray value comparatively accurately by the method for bilinear interpolation.

Step 4: use bilinear interpolation value method, realize distortion correction to fish eye images.

Use contrary projection promptly by the some p (x on the space projection cube according to step 3 ₀, y ₀, z ₀) correspondence find m on the imaging plane (x, coordinate y), existing to its rectification.

Use bilinear interpolation, (X, pixel value Y) will be by 4 pixel coordinate point interpolations acquisitions on the corresponding fish eye images can to make image object pixel coordinate point P after proofreading and correct.

Spatial point p (x ₀, y ₀, z ₀) be that (x y), obtains m (x by contrary projection to fish-eye image picture point m through cubic projection model Nonlinear Mapping, y) floating-point coordinate: (i+u, j+v), (i ∈ Z+, j ∈ Z+, u ∈ [0,1), v ∈ [0,1)) (and be x=i+u, y=j+v, i wherein, j is the integer part of floating-point coordinate, u, v is the fractional part of floating-point coordinate, is value [0,1) interval floating number), with f (x, image coordinate (x, the pixel value of y) locating after y) expression is proofreaied and correct, m (x then, pixel value f y) (x, y)=(i+u can be (i by coordinate on the fish eye images j+v) to f, j), (i+1, j), (i, j+1), (i+1, j+1) value of pairing 4 pixels decision (see figure 4), because the gradation of image value is followed all linear distributions of vertical both direction in level, can obtain the transformation for mula as subinverse projection coordinate:

f(i+u，j+v)＝(1-u)(1-v)f(i，j)+(1-u)vf(i，j+1)+u(1-v)f(i+1，j)+uvf(i+1，j+1)

Use bilinear interpolation value method, promptly according to above-mentioned formula, (X, pixel value Y) will be obtained by 4 pixel interpolation on the corresponding fish eye images image object pixel P after the correction.

Claims (1)

1. One kind based on cubic projection to the fish eye images distortion correction method, comprise the steps:

   Step 1: obtain imaging model with fish eye lens, and according to imaging model fish eye lens is demarcated and to be obtained calibrating parameters;

   Step 2: according to the fish eye images that collects, determine viewpoint, and be coordinate origin, set up cube perspective projection model, obtain the cube perspective projection of spatial point with the viewpoint;

   Step 3: set up the cube perspective projection of spatial point and the mapping relations between the fish eye images;

   Step 4: use bilinear interpolation value method, realize the distortion correction to fish eye images, method is as follows: establish spatial point p (x ₀, y ₀, z ₀) through cubic projection model Nonlinear Mapping be on the fish eye images some m (x, y), the floating-point coordinate of this point be (i+u, j+v), (i ∈ Z+, j ∈ Z+, u ∈ [0,1), v ∈ [0,1)), wherein i, j are the integer part of floating-point coordinate, and u, v are the fractional part of floating-point coordinate, are values [0,1) interval floating number is with f (x, image coordinate (x after y) expression is proofreaied and correct, y) pixel value of locating, then m (x, pixel value y) is tried to achieve according to following formula:

   f(i+u，j+v)＝(1-u)(1-v)f(i，j)+(1-u)vf(i，j+1)+u(1-v)f(i+1，j)+uvf(i+1，j+1)。

Images