JP5896873B2 - Blur image detection method - Google Patents

Description

  The present invention relates to a blurred image detection method for detecting a blurred image from a plurality of images obtained by continuously photographing a plurality of adjacent areas in a subject.

  2. Description of the Related Art Conventionally, in order to detect the presence or absence of cracks or steps on a concrete surface (for example, a tunnel wall surface), the concrete surface is continuously photographed with a camera (see, for example, Patent Documents 1 and 2).

  FIG. 4 is a perspective view showing an example of a state in which a concrete surface is continuously photographed. Reference numeral 100 denotes a tunnel wall surface as a subject. Reference numeral 101 denotes a camera for photographing the tunnel wall surface 100. Reference numeral 102 denotes a carriage (vehicle) on which the camera 101 is mounted and moved. In this apparatus, the tunnel wall surface 100 can be continuously photographed by moving the carriage 102 at a constant speed and releasing the shutter of the camera 101 at a constant interval.

JP 2011-117788 A JP 2012-098170 A

  However, there are cases where so-called out-of-focus images (that is, images that are totally out of focus and have low edge strength) are mixed in many images captured as described above. . Since such a defocused image cannot be used for detecting cracks or the like, it is necessary to determine whether or not it is a defocused image, but there is a problem that the work becomes complicated if it is determined by human eyes.

  An object of the present invention is to provide a blurred image detection method capable of solving the above-described problems.

The invention according to claim 1 is exemplified in FIGS. 1 and 2 (a) and 2 (b), and is a plurality of images obtained by continuously photographing a plurality of adjacent areas in a subject (FIG. 2 ( a)
In the blurred image detection method for detecting a defocused image from the codes Q _i-2 , Q _i−1 , Q _i , Q _{i + 1} , and Q _{i + 2} in FIG.
Two images (Q _i−1 , Q _i ) continuously captured so as to partially overlap are referred to as “first captured image” and “second captured image”, and the first captured image (Q _i−1). ) In the second captured image (Q _i ) (refer to the reference symbol R _{i-1 in} FIG. 2B) is the “first overlap portion”, and the second captured image (Q _i ) Overlapping part with one captured image (Q _i-1 ) (FIG. 2 (b)
) (Referring to the symbol R _i ) of “second overlapping portion”,
The first captured image (Q _i-1) is first captured image determination step of determining visually whether out-of-focus image (see S ₀ in Fig. 1),
A first overlapping portion specifying step (S ₁₁ ) for specifying the first overlapping portion (R _i-1 ) in the first captured image (Q _i-1 );
First pre-processing pixel value acquisition step (P ₁ (j)) for acquiring a pixel value of each pixel (hereinafter referred to as “first pre-processing pixel value”) (P ₁ (j)) in the first overlapping portion (R _i-1 ) S ₁₂ ),
A first filtering step (S ₁₃ ) for filtering the first overlapping portion (R _i-1 );
A first pixel value (P _1m (j)) for acquiring the pixel value of each pixel in the first overlap portion (R _i-1 ) after the filter processing (hereinafter referred to as “first post-processing pixel value”) is obtained. A post-processing pixel value acquisition step (S ₁₄ );
For each pixel, the absolute value of the difference value between the first pre-processing pixel value (P ₁ (j)) and the first post-processing pixel value (P _1m (j)) (hereinafter referred to as “first difference value”). ) (| P _1m (j) −P ₁ (j) |) to obtain a first difference value acquisition step (S ₁₅ ),
A first difference value integrating step (S ₁₆ ) for integrating the first difference value (| P _1m (j) −P ₁ (j) |) of each pixel with respect to the first overlapping portion (R _i-1 );
A second overlapping portion specifying step (S ₂₁ ) for specifying the second overlapping portion (R _i ) in the second captured image (Q _i );
Second pre-processing pixel value acquisition step (S ₂₂ ) for acquiring a pixel value of each pixel (hereinafter referred to as “second pre-processing pixel value”) (P ₂ (j)) in the second overlapping portion (R _i ). )When,
A second filtering step (S ₂₃ ) for filtering the second overlapping portion (R _i );
After the second process of obtaining the pixel value of each pixel in the second overlapping portion (R _i ) after the filter process (hereinafter referred to as “second processed pixel value”) (P _2m (j)) A pixel value acquisition step (S ₂₄ );
For each pixel, the absolute value of the difference value between the pre-second pixel value (P ₂ (j)) and the post-second pixel value (P _2m (j)) (hereinafter referred to as “second differential value”). ) (| P _2m (j) −P ₂ (j) |) second difference value acquisition step (S ₂₅ ),
A second difference value integrating step (S ₂₆ ) for integrating the second difference value of each pixel with respect to the second overlapping portion (R _i );
The value obtained in the first difference value integration step (S ₁₆ ) (hereinafter referred to as “first integration value”) ([i−1]) and the second difference value integration step (S ₂₆ ). A value based on the first integrated value ([i-1]) and the second integrated value ([i]) from the value (hereinafter referred to as “second integrated value”) ([i]) ( Hereinafter, a feature amount calculating step (S ₃ ) for calculating “a feature amount”),
A blur image determination step (S ₄ ) for determining whether or not the second captured image (Q _i ) is a blur image based on the feature amount calculated in the feature amount calculation step (S ₃ ); It is characterized by having.

  Note that the numbers in parentheses are for the sake of convenience indicating the corresponding elements in the drawings, and therefore the present description is not limited to the descriptions on the drawings.

  According to the first aspect of the present invention, whether or not an image other than the first photographed image (that is, the second photographed image or another photographed image) is an out-of-focus image is automatically determined without depending on human eyes. Therefore, the determination work can be simplified.

FIG. 1 is a flowchart showing a specific example of a blurred image detection method according to the present invention. FIG. 2A is a perspective view showing an example of shooting, and FIG. 2B is a diagram for explaining the positional relationship between two images (first and second shot images). It is a perspective view. FIG. 3 is a diagram for explaining the outline of the difference value acquisition step. FIG. 4 is a perspective view showing an example of a state in which a concrete surface is continuously photographed.

  Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 3.

The blurred image detection method according to the present invention includes a plurality of images (symbols Q _i-2 , Q _i-1 , Q _{i in} FIG. 2A) acquired by continuously photographing a plurality of adjacent areas in a subject. , Q _{i + 1} , Q _{i + 2} )).

  Here, the subject may be a concrete wall surface of a tunnel. In addition, the term “continuous shooting” in this specification does not mean that shooting is performed continuously in time as in so-called continuous shooting (that is, the shutter is continuously released), but the shooting area is continuous in position. This means that “one area” and “other areas adjacent to the one area” in the subject are each photographed, regardless of whether or not they are photographed continuously in time. It is efficient to perform this continuous shooting while moving a vehicle (not shown) equipped with a photographing means such as a digital camera (see reference numeral 1 in FIG. 2 (a)). No photography is excluded from the scope of the present invention. In the case of continuous shooting using the vehicle, it is preferable to release the shutter at a substantially constant interval with the vehicle traveling speed being substantially constant. However, the vehicle traveling speed and the timing at which the shutter is released are constant. It does not exclude from the scope of the present invention the case where the change is made without doing. Further, the “out-of-focus image” in the present specification refers to an image (image with low edge strength) that is not focused and is entirely blurred.

By the way, the determination of a blurred image according to the present invention is performed using two images (for example, Q _i−1 , Q _i ) that are continuously photographed so as to partially overlap. The images Q _i-1 and Q _i are referred to as “first photographed image” and “second photographed image”, and the first photographed image Q _i-1 overlaps with the second photographed image Q _i ( The reference R _{i-1 in} FIG. 2B is referred to as a “first overlapping portion”, and the second captured image Q _i overlaps with the first captured image Q _i-1 (see FIG. 2B).
(Refer to the reference symbol _Ri ).

<First photographed image determination step>
Blurred image detecting method according to the present invention, the first captured image Q _i-1 is the human whether out-of-focus image comprises a first captured image determination process S ₀ is determined by visual observation.

<Duplicate part identification process>
The blurred image detection method according to the present invention includes:
· In the first captured image _{Q i-1} and the first overlap portion specifying step _{S 11} for identifying the first overlapping portion _{R i-1,}
A second overlap portion specifying step _{S 21} for identifying the second overlapping portion _{R i} in - the second captured image _{Q i,}
It has.

Since the first overlapping portion R _i-1 and the second overlapping portion R _i are portions where the same area of the subject is imaged, of course, the first captured image Q _i-1 and the second captured image Q _As long as _i is the same size, the first overlap portion R _i-1 and the second overlap portion R _i have the same shape and the same size. The identification of the first overlapping portion R _i-1 and the second overlapping portion R _i is as follows:
The operator performs the operation manually (that is, operation of the mouse or the like) while viewing the captured images while the first captured image Q _i-1 and the second captured image Q _i are displayed on the monitor of the personal computer. Or
・ Even if it is done automatically with a computer,
both are fine. In addition, as the latter method (that is, a method in which the personal computer itself automatically identifies the overlapping portions R _i-1 and R _i ),
A method of identifying the overlapping portions R _i-1 and R _i from the data on the traveling speed and the data on the timing of releasing the shutter when the traveling speed of the vehicle is substantially constant;
A method may be mentioned in which the first photographed image Q _i-1 and the second photographed image Q _i are compared by a matching technique to identify the positions and shapes of the overlapping portions R _i-1 , R _i . Further, the identification of the first overlap portion R _i-1 and the second overlap portion R _i is as follows:
The first overlapping portion R _i-1 and the second overlapping portion R _i may be cut out from the captured images Q _i-1 and Q _i as “images”, or
The position and shape of the first overlapping portion R _i-1 in the first captured image Q _i-1 is specified by a numerical value (for example, pixel address) without being cut out as an “image”, and the second captured image In Q _i , the position and shape of the second overlapping portion R _i may be specified by a numerical value (for example, a pixel address).
both are fine.

<Pixel value acquisition process before processing>
The blurred image detection method according to the present invention includes:
A first pre-processing pixel value acquisition step S ₁₂ for acquiring a pixel value (hereinafter referred to as “first pre-processing pixel value”) P ₁ (j) of each pixel in the first overlap portion R _i−1 ;
A second pre-processing pixel value acquisition step S ₂₂ for acquiring a pixel value (hereinafter referred to as “second pre-processing pixel value”) P ₂ (j) of each pixel in the second overlapping portion R _i ;
It has.

<Filtering process>
Furthermore, the blurred image detection method according to the present invention includes:
- said first overlap portion _{R i-1} and the first filtering step _{S 13} that filters,
- the second overlapping portion _{R i} and the second filtering step _{S 23} that filters,
It has. Here, as the filter processing of each of the overlapping portions R _i−1 and R _i , for example, median filter processing (that is, the sizes of the peripheral pixel values are sequentially arranged, and the median value (Median) is set as the pixel value of the target pixel. Filter processing, which is a process for removing a region (pixel value) that is significantly different from the surrounding pixel values. When such a filtering process is performed, noise and edges (for example, fine irregularities on the concrete wall surface that is the subject, construction traces, crack traces, free lime edges, and the like are detected from the overlapping portions R _i-1 and R _i. That is, a portion where the change of the pixel value is steep) is removed and suppressed.

<Post-treatment pixel value acquisition process>
The blurred image detection method according to the present invention includes:
A first post-processing pixel that obtains a pixel value (hereinafter, referred to as “first post-processing pixel value”) P _1m (j) of each pixel in the first overlap portion R _i−1 after the filtering process. a value acquisition step _{S 14,}
The second post-processing pixel value acquisition for acquiring the pixel value (hereinafter referred to as “second post-processing pixel value”) P _2m (j) of each pixel in the second overlapping portion R _i after the filtering process. a step _{S 24,}
It has.

<Difference value acquisition process>
Furthermore, the blurred image detection method according to the present invention includes:
For each pixel, the absolute value of the difference value between the first pre-processing pixel value P ₁ (j) and the first post-processing pixel value P _1m (j) (hereinafter referred to as “first difference value”) “| A first difference value acquisition step S ₁₅ for acquiring P _1m (j) −P ₁ (j) | ″;
For each pixel, an absolute value of a difference value between the second pre-processing pixel value P ₂ (j) and the second post-processing pixel value P _2m (j) (hereinafter referred to as “second difference value”) “| A second difference value acquisition step S ₂₅ for acquiring P _2m (j) −P ₂ (j) | ″;
It has. FIG. 3 is a diagram for explaining the outline of the difference value acquisition step, where the horizontal axis represents pixels (pixel positions) and the vertical axis represents pixel values. In the drawing, a broken line (solid line) indicated by a symbol D indicates a pre-processing pixel value P (j) of each pixel, and a broken line (broken line) indicated by a symbol E indicates a post-processing pixel value P _m (j). A broken line (solid line) indicated by F indicates a difference value | P _m (j) −P (j) |. In this manner, by the first difference value acquiring step S _15, in the first overlapping portion R _i-1, change of the pixel steep portion (pixel having a noise or edge) is extracted (detected), the the second difference value acquiring step S _25, in the second overlapping portion R _i, change of the pixel steep portion (pixel having a noise or edge) is extracted (detected).

<Differential value integration process>
The blurred image detection method according to the present invention includes:
A first difference value integration step S ₁₆ for integrating the first difference value | P _1m (j) −P ₁ (j) | of each pixel with respect to the first overlap portion R _i−1 ;
A second difference value integration step S ₂₆ for integrating the second difference value | P _2m (j) −P ₂ (j) | of each pixel with respect to the second overlapping portion R _i ;
It has. Incidentally, the first difference value by the first difference value integration process _{S 16 | P 1m (j)} -P 1 (j) | of the first overlapping portion _{R i-1} integrated values for (edge amount) [ i-1] is referred to as a “first integrated value”, and the second difference value | P _2m (j) −P ₂ (j) | is used as the second overlap value in the second difference value integrating step S ₂₆ . The value (total edge amount) [i] integrated for the part R _i is referred to as a “second integrated value”. The first integrated value [i-1] and the second integrated value [i] are represented by the following formulas.

<Feature amount calculation process>
Furthermore, the blurred image detection method according to the present invention includes the first integrated value [i−1] and the second integrated value [i] from the first integrated value [i−1] and the second integrated value [i]. i] value based on (hereinafter referred to as "feature quantity") and a feature calculation step S ₃ calculated. As this feature amount,
The difference between the first integrated value [i−1] and the second integrated value [i], that is, Ai = [i] − [i−1]
A value obtained by dividing the first integrated value [i−1] by the second integrated value [i], that is, Bi = [i] ÷ [i−1]
When [i−1]> [i], Ci = − ([i−1] ÷ [i]) + 1.0, and when [i−1] ≦ [i], Ci = + ([i ] ÷ [i−1]) − 1.0
And so on.

<Bokeh image determination process>
Moreover, the blurred image detecting method according to the present invention determines blurred image determining whether the second captured image Q _i based on the feature amount calculated by the feature value calculation step S ₃ is blurred image which comprises the step _{S 4.} For example, when it is visually determined in the first captured image determination step S ₀ that the first captured image Q _i-1 is a blurred image, the first pre-process pixel value P ₁ (j) and the first The post-processing pixel value P _1m (j) is substantially equal, and the first difference value | P _1m (j) −P ₁ (j) | and the first integrated value [i−1] are values close to 0. Become. For example, if Ai = [i] − [i−1] ≈0 is calculated in the feature amount calculation step S ₃ , the second difference value | P _2m (j) −P ₂ (j) | Since the second integrated value [i] is approximately 0, it is determined that the second captured image Qi is a defocused image.

In addition,
· First overlap portion specifying step _{S 11,} the Ya first processing previous pixel value obtaining step _{S 12,} the first filtering step _{S 13} or after the first processing pixel value obtaining step _{S 14,} the first difference value acquiring step _{S 15} must be carried out before the first difference value integration process S _16,
· Second overlap portion specifying step _{S 21,} the Ya second processing previous pixel value obtaining step _{S 22} and the second filtering step _{S 23} and the second processing after the pixel value obtaining step _{S 24} and the second difference value acquiring step _{S 25} must be carried out before the second difference value integration process S _26,
- the first processed pixel value obtaining step S ₁₄ should be carried out after the first filtering step S _13,
- second processing after the pixel value obtaining step S ₂₄ must be carried out after the second filtering step S _23,
- first difference value acquiring step S ₁₅ must be carried out after the first pre-processing pixel value obtaining step S ₁₂ or the first processed pixel value obtaining step S _14,
- second difference value acquiring step S ₂₅ must be carried out after the second pre-processing pixel value obtaining step S ₂₂ and the second processing after the pixel value obtaining step S _24,
- the first difference value integration process S ₁₆ must be carried out after the first difference value acquiring step S _15,
- second difference value integration process S ₂₆ must be carried out after the second difference value acquiring step S _25,
The first captured image determination step S ₀ needs to be performed before the blurred image determination step S ₄ , but the execution order of the other steps may be changed as appropriate.

On the other hand, if another captured image that is continuous with the second captured image Q _i is “third captured image Q _{i + 1} ”, the relationship between the second captured image Q _i and the third captured image Q _{i + 1} is also described above. Steps S ₁₁ , S ₁₂ ,..., S ₂₁ , S ₂₂ ,..., S ₃ , S ₄ may be performed to determine the out of focus of the third captured image Q _{i + 1} . Further, the third captured image _{Q i + 1} and the fourth captured image _{Q i + 2} of the above-mentioned also in relation step _{S 11, S 12, ...,} S 21, S 22, ..., said implement S 3, _{S 4} second The out-of-focus determination of the four captured images Q _{i + 2} may be performed.

According to the present invention, it is possible to determine whether an image other than the first photographed image (that is, the second photographed image Q _i , the third photographed image Q _{i + 1} ,...) Is a blurred image without depending on human eyes. Therefore, the determination work can be simplified.

[I-1] First integrated value [i] Second integrated value P ₁ (j) First pre-processing pixel value P _1m (j) First post-processing pixel value | P _1m (j) −P ₁ (j) | First difference value P ₂ (j) second pre-processing pixel value P _2m (j) second post-processing pixel value | P _2m (j) −P ₂ (j) | second difference value Q _i−1 first Captured image Q _i Second captured image R _i-1 First overlap portion R _i Second overlap portion S ₀ First captured image determination step S ₁₁ First overlap portion specifying step S ₁₂ First pre-processing pixel value acquisition step S ₁₃ first filtering step S ₁₄ first processed pixel value obtaining step S ₁₅ the first difference value acquiring step S ₁₆ the first difference value integration process S ₂₁ second overlap portion specifying step S ₂₂ second processing previous pixel value acquisition step S ₂₃ second filtering step S ₂₄ second post-processing pixel value obtaining step S ₂₅ second difference value acquiring step S ₂₆ second difference value product Step S ₃ feature quantity calculation step S ₄ blurred image determination step

Claims (1)

In a blurred image detection method for detecting a blurred image from a plurality of images obtained by continuously shooting a plurality of adjacent areas in a subject,
Two images continuously photographed so as to partially overlap each other are referred to as a “first photographed image” and a “second photographed image”, and an overlapping portion of the first photographed image with the second photographed image is designated as “first photographed image”. In the case where the overlapping portion with the first captured image in the second captured image is the “second overlapping portion”,
A first captured image determination step for visually determining whether the first captured image is a defocused image;
A first overlapping portion specifying step for specifying the first overlapping portion in the first captured image;
A first pre-processing pixel value acquisition step of acquiring a pixel value of each pixel in the first overlapping portion (hereinafter referred to as “first pre-processing pixel value”);
A first filtering process for filtering the first overlapping portion;
A first post-processing pixel value acquisition step of acquiring a pixel value of each pixel in the first overlapping portion after the filter processing (hereinafter referred to as “first post-processing pixel value”);
A first difference value acquisition step of acquiring an absolute value of a difference value between the first pre-processing pixel value and the first post-processing pixel value (hereinafter referred to as “first difference value”) for each pixel;
A first difference value integrating step of integrating the first difference value of each pixel with respect to the first overlapping portion;
A second overlapping portion specifying step of specifying the second overlapping portion in the second captured image;
A second pre-processing pixel value acquisition step of acquiring a pixel value of each pixel in the second overlapping portion (hereinafter referred to as “second pre-processing pixel value”);
A second filter processing step of filtering the second overlapping portion;
A second post-processing pixel value acquisition step of acquiring a pixel value of each pixel in the second overlapping portion after the filter processing (hereinafter referred to as “second post-processing pixel value”);
A second difference value acquisition step of acquiring an absolute value (hereinafter referred to as “second difference value”) of a difference value between the pixel value before the second process and the pixel value after the second process for each pixel;
A second difference value integrating step of integrating the second difference value of each pixel with respect to the second overlapping portion;
The value obtained in the first difference value integrating step (hereinafter referred to as “first integrated value”) and the value obtained in the second difference value integrating step (hereinafter referred to as “second integrated value”) A feature amount calculating step for calculating a value based on the first integrated value and the second integrated value (hereinafter, referred to as “feature amount”);
A blurred image determination step of determining whether or not the second captured image is a blurred image based on the feature amount calculated in the feature amount calculation step;
A blurred image detection method comprising:

Images