JP2006270523A - Apparatus and processing program for image composition - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image composition apparatus capable of hand blur correction even in the event of large hand blur at the time of imaging. <P>SOLUTION: At the time of correcting a hand blur by composing continuously imaged images, an evaluation value EBL representing harmonic components of the image concerned is calculated by unconditionally adopting a first imaged image as a key frame. The evaluation values EBLi (i≥2) are calculated for the second frame and thereafter, and difference values D (EBLi-EBL1) from the evaluation value EBL1 of the key frame are calculated successively. Then, as to the image having the difference value D smaller than a rejection threshold D_LOW, the image concerned is rejected as a rejected image for composition because of a large hand blur amount. Thus, the hand blur can be corrected even when a large hand blur is produced at the time of imaging. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image composition apparatus and an image composition processing program suitable for use in a digital camera or a digital video camera.

  2. Description of the Related Art Conventionally, a technique for correcting camera shake by image processing is known. For example, Patent Document 1 captures a plurality of images having different exposure amounts, detects one motion vector as an entire image between one image and the other image, and detects the detected motion vector. Based on the image, determine whether the image blur is “subject shake” or “camera shake”. In the case of “camera shake”, the position of the subject, the background, etc. corresponding to the angle of view shift caused by “camera shake” An apparatus for correcting camera shake by aligning each successive image so as to cancel out the shift and superimposing the images on one image is disclosed.

JP 2003-134385 A

By the way, the amount of camera shake in continuously shot images is not constant, so when overlapping images that have been shot continuously, if there is an image with a large amount of camera shake, the quality of the combined image will be impaired, As a result, there is a problem that camera shake correction cannot be performed.
Accordingly, the present invention has been made in view of such circumstances, and an object of the present invention is to provide an image composition processing apparatus and an image composition processing program capable of correcting camera shake even when a large camera shake occurs during photographing.

  In order to achieve the above object, according to the first aspect of the present invention, an extraction unit that extracts an evaluation amount that changes in accordance with the amount of camera shake in an image from a captured image, and an evaluation that is extracted by the extraction unit. A determination unit that determines whether or not to use the image as an image to be combined for each captured image, and combines each of the images that are determined to be adopted by the determination unit among the continuously captured images. And image synthesizing means for correcting camera shake.

  In the invention according to claim 2, an extraction unit that extracts an evaluation amount EBL representing a harmonic component that changes in accordance with the amount of camera shake in the image from each of the continuously photographed images, and the extraction unit includes Each time the evaluation amount of each image photographed continuously is extracted, the difference obtained by subtracting the evaluation amount EBLi (i ≧ 2) of each second photographed image from the evaluation amount EBL1 of the first photographed image The difference value generating means for sequentially generating the value D (EBLi-EBL1) and the difference value D (EBLi-EBL1) generated by the difference value generating means are used as images to be combined for each photographed image. Determining means for determining whether or not, and image synthesizing means for correcting camera shake by synthesizing each of the images taken and determined by the determining means among the continuously photographed images.

  In the invention according to claim 3 that depends on claim 2, the extraction means includes an image reduction means for reducing a captured image, and a harmonic after noise removal from the reduced image reduced by the image reduction means. Harmonic component image generating means for detecting a component and generating a harmonic component image is provided.

  In the invention according to claim 4 subordinate to claim 2, the determination means is picked up this time if the difference value D (EBLi-EBL1) generated by the difference value generation means is larger than a predetermined threshold value. An image of the evaluation amount EBLi is adopted as an image to be synthesized, and on the other hand, if it is smaller than a predetermined threshold value, it is not adopted.

  In the invention according to claim 5 subordinate to claim 4, the determination means has a case where the difference value D (EBLi−EBL1) is larger than the first threshold and the number of images determined to be adopted is smaller than a predetermined value. In addition, an update unit that updates the evaluation amount EBLi of the image captured this time to the evaluation amount EBL1, and an image obtained by capturing the number n of images for which the difference value D (EBLi−EBL1) is smaller than the second threshold and determined to be adopted And a threshold value changing means for setting a third threshold value obtained by changing the second threshold value when the adoption ratio n / i obtained by dividing by the number i is smaller than a predetermined value.

  In the invention according to claim 6, the evaluation amount that changes in accordance with the amount of camera shake in the image is extracted from the captured image, and the evaluation amount extracted by the extraction processing is used to capture the image. A determination process for determining whether or not to adopt each image as an image to be combined, and an image composition process for correcting camera shake by combining the images that have been determined to be adopted by the determination process among the continuously captured images. Is executed by a computer.

  In the invention according to claim 7, the extraction processing for extracting the evaluation amount EBL representing the harmonic component that changes in accordance with the amount of camera shake in the image from each of the continuously photographed images, and the extraction processing Each time the evaluation amount of each image photographed continuously is extracted, the difference obtained by subtracting the evaluation amount EBLi (i ≧ 2) of each second photographed image from the evaluation amount EBL1 of the first photographed image A difference value generation process for sequentially generating a value D (EBLi-EBL1) and a difference value D (EBLi-EBL1) generated by the difference value generation process are used as images to be combined for each captured image. A computer that executes determination processing for determining whether or not, and image combination processing for correcting camera shake by combining the images that have been determined to be adopted by the determination processing among the continuously shot images.

  In the invention according to claim 8, which is dependent on claim 7, the extraction processing includes image reduction processing for reducing a captured image, and harmonics after noise removal from the reduced image reduced by the image reduction processing. And harmonic component image generation processing for detecting a component and generating a harmonic component image.

  In the invention according to claim 9, which is dependent on claim 7, the determination process is performed when the difference value D (EBLi-EBL1) generated by the difference value generation process is larger than a predetermined threshold value. An image of the evaluation amount EBLi is adopted as an image to be synthesized, and on the other hand, if it is smaller than a predetermined threshold value, it is not adopted.

  In the invention according to claim 10, which is dependent on claim 9, the determination process is performed when the difference value D (EBLi-EBL1) is larger than the first threshold and the number of images determined to be adopted is smaller than a predetermined value. In addition, an update process for updating the evaluation amount EBLi of the image captured this time to the evaluation amount EBL1, and an image in which the difference value D (EBLi−EBL1) is smaller than the second threshold and the number n of the images determined to be adopted is captured. And a threshold value changing process for setting a third threshold value obtained by changing the second threshold value when the adoption ratio n / i obtained by dividing by the number i is smaller than a predetermined value.

  According to the first and sixth aspects of the present invention, the evaluation amount that changes in accordance with the amount of camera shake in the image is extracted from the photographed image, and for each image photographed using the extracted evaluation amount. It is determined whether or not to adopt as an image to be synthesized. Then, among the continuously photographed images, the images determined to be adopted are synthesized to correct camera shake. Therefore, since an image with a large amount of camera shake is excluded without being adopted as an image to be synthesized, a camera shake can be corrected even if a large amount of camera shake occurs during shooting.

  According to the second and seventh aspects of the invention, each time an evaluation amount EBL representing a harmonic component that changes in accordance with the amount of camera shake in an image is extracted from each image that is continuously captured, A difference value D (EBLi−EBL1) is sequentially generated by subtracting the evaluation amount EBLi (i ≧ 2) of each image photographed after the second from the evaluation amount EBL1 of the image photographed at the same time. Then, using the generated difference value D (EBLi-EBL1), it is determined whether or not to adopt as an image to be combined for each captured image, and each of the images determined to be adopted among the continuously captured images. Composes images and corrects camera shake. Therefore, since an image with a large amount of camera shake is excluded without being adopted as an image to be synthesized, a camera shake can be corrected even if a large amount of camera shake occurs during shooting.

  In the following, an outline of the present invention will be described first, and then an embodiment of the present invention will be described with reference to the drawings.

A. SUMMARY OF THE INVENTION The present invention utilizes the property that harmonic components in the direction of camera shake in an image change according to the amount of camera shake. That is, as will be described later, an evaluation amount EBL for evaluating the amount of camera shake is obtained from the captured image. The evaluation amount EBL representing the harmonic component in the image represents an image with less camera shake as the value increases. Therefore, in the case of performing camera shake correction by combining continuously shot images, it is determined using the evaluation amount EBL whether or not to use as an image to be combined, and an image with a large amount of camera shake and a low evaluation amount EBL is determined. By excluding it from the synthesis target, it is possible to correct camera shake even if a large camera shake occurs during shooting.

  Specifically, first, the input image P obtained by imaging is reduced to a predetermined image size. Camera shake is recognized as obvious “blurring” in the image when it fluctuates by 1% or more with respect to the angle of view. If the number of pixels is 640 dots in the horizontal direction, it is 6 dots or more. To put it the other way around, there are many cases where you don't care much about camera shake below this.

Therefore, for example, about 640 dots × 480 dots is sufficient as the image size to be reduced. After the input image P is reduced, the reduced image Ps is filtered by a Gaussian filter Hg to generate an image Pg from which spike noise and the like are removed, as represented by the following equation [1].
Pg = Hg * Ps [1]
According to the above formula [1], spike noise and the like are removed so as not to affect the evaluation of the amount of camera shake. For removing spike noise, a median filter may be used in addition to the Gaussian filter.

Next, as represented by the following equation [2], the image Pg from which spike noise or the like has been removed is filtered by a Laplacian filter (secondary differential filter) Hl to obtain a harmonic component Pgl of the image Pg.
Pgl = Hl * Pg [2]
Then, in the harmonic component Pgl of the image Pg, as represented by the following expressions [3a] and [3b], the maximum value (absolute value) of the harmonic component Pgl in the row direction (x-axis direction) and the column direction (y-axis direction) Value) FH (y) and FV (x) are obtained.
FH (y) = max (abs (Pgl (x, y))) [3a]
FV (x) = max (abs (Pgl (x, y))) [3b]
Note that FH (y) is the maximum value (absolute value) of each y column in the x-axis direction in the harmonic component Pgl of the image Pg, while FV (x) is the maximum value (absolute value) of each x-row in the y-axis direction. Value).

Next, as represented by the following equations [4a] and [4b], the sums FHS and FVS of FH (y) and FV (x) are obtained.
FHS = ΣFH (y) [4a]
FVS = ΣFV (x) ... [4b]
Therefore, FHS is the harmonic component accumulated value in the x-axis direction (lateral direction), and FVS is the harmonic component accumulated value in the y-axis direction (vertical direction). The evaluation amount EBL obtained by adding the harmonic component accumulated value FHS and the harmonic component accumulated value FVS is used for evaluating the amount of camera shake, as represented by the following equation [5].
EBL = FHS + FVS [5]
When camera shake occurs, harmonic components in the direction of camera shake are reduced. Therefore, as the evaluation amount EBL is larger, the captured image (input image P) can be evaluated as an image with less camera shake.

In the embodiment described later, as described above, the evaluation amount EBL is generated by adding the harmonic component accumulated value FHS and the harmonic component accumulated value FVS. However, the present invention is not limited to this, and the following equation [6] As described above, the evaluation value EBL may be calculated by accumulating the absolute value of the harmonic component Pgl of the image Pg described above.
EBL = ΣΣabs (Pgl (x, y)) [6]

  Next, based on the evaluation amount EBL obtained in this way, it is determined whether or not to adopt as an image to be synthesized. Of the continuously photographed images, the first photographed image is unconditionally adopted as a key frame, and the evaluation amount EBL1 is calculated based on the relationship of the above-described equations [1] to [5]. Similarly, the evaluation amount EBLi (i ≧ 2) is obtained for the second and subsequent frames, and a difference value D (EBLi−EBL1) from the evaluation amount EBL1 of the key frame is sequentially calculated. If the difference value D is a negative value smaller than a threshold value to be described later, the image of the i-th frame has a large amount of camera shake, and is determined to be inappropriate as an image to be combined. In this case, the image of the i-th frame is not adopted as an image to be combined and is excluded.

  On the other hand, if the difference value D is a positive value larger than a threshold value, which will be described later, it is considered that a large amount of camera shake is included in the image of the first key frame. Therefore, it is preferable to replace the key frame with the next frame. However, if many images have already been combined, it is preferable to refrain from replacing key frames at this point. In actual operation, it is often possible to shoot a stable image with few camera shakes for several frames between 400 and 500 milliseconds from the start of shooting. Therefore, an image with less camera shake may be taken after that. However, every time an image with less camera shake is taken, if the key frame is replaced, it takes time to synthesize the image, so the key frame is not replaced. There may be situations where it is preferable to continue the process.

  Note that the period for which the adoption is determined may be defined by the elapsed time from the start of shooting in addition to the number of images to be combined (the number of frames). In addition, when the number or ratio of non-accepted frames is large, there may be a case where the evaluation amount EBL1 of the key frame is very good. Therefore, it is preferable to variably set the threshold used for the adoption determination, for example, gradually. .

B. Configuration Next, a configuration of a digital camera according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2. FIG. 1 is a block diagram showing the overall configuration of a digital camera, and FIG. 2 is a block diagram showing the configuration of an image sensor.
As shown in FIG. 1, the digital camera according to the present invention includes an operation unit 11, a CPU 12, a program code storage unit 13, an optical lens unit 14, an image sensor 15, an image memory 16, a display unit 17, an image processing unit 18, and a PC. An interface 19, an external memory input / output control unit 20, and a memory card 21 are provided. The operation unit 11 includes various operators such as a power switch and a shutter button, generates an operation event corresponding to the operated operator, and supplies the operation event to the CPU 12.

  The CPU 12 executes a control program stored in a program code storage unit 13 including a ROM, a flash memory, and the like, and generates a control signal for controlling each unit in response to an operation event input from the operation unit 11. The optical lens unit 14 includes a photographing lens and a lens driving mechanism, and focuses the photographing lens by a control signal supplied from the CPU 12 to form a subject image on the image sensor 15.

  As shown in FIG. 2, the image sensor 15 includes an image sensor (photo sensor) 101 in which primary color filters (R, G, B) are arranged in an array called Bayer color scheme and performs photoelectric conversion, and each charge of the image sensor 101. Of the vertical register 102, the horizontal register 103 that reads out the photoelectric signal held in the vertical register 103 and transfers it to the amplifier 105, and the reading of the vertical register 102 / horizontal register 103 in accordance with the read control signal supplied from the CPU 12. A timing control unit 104 that controls the write timing and an A / D converter 106 that A / D converts the output signal of the gain-adjusted amplifier 105 and outputs image data.

  The image sensor 15 configured as described above is controlled by the CPU 12, and generates image data with a low preview resolution at a predetermined frame period (for example, about 30 frames per second) when the shutter button of the operation unit 11 is not pressed under the shooting mode. To the image memory 16. On the other hand, when the shutter button of the operation unit 11 is pressed in the shooting mode, image data with high resolution is generated and supplied to the image memory 16. Further, the image sensor 15 sets the imaging sensitivity (ISO sensitivity) indicated by the CPU 12 according to the brightness of the subject.

  The image memory 16 temporarily stores low-resolution preview image data and high-resolution image data output from the image sensor 15, or temporarily stores image data processed by the image processing unit 18. The image data temporarily stored in the image memory 16 is read under the control of the CPU 12 and supplied to the display unit 17. The display unit 17 includes, for example, a color liquid crystal panel and a drive driver, and displays the image data supplied from the image memory 16 on the screen. The image processing unit 18 performs image compression, brightness correction, white balance adjustment, and the like on the image data temporarily stored in the image memory 16 under the control of the CPU 12. The PC interface unit 19 operates as an input / output driver when connected to a computer (not shown) via USB. As a result, a memory card 21 described later can be handled as an external storage device. The external memory input / output control unit exchanges image data with the memory card 21.

C. Operation Next, with reference to FIG. 3 to FIG. 5, an operation of the imaging process executed in the embodiment having the above-described configuration will be described. 3 is a flowchart showing the schematic operation of the shooting process, FIG. 4 is a flowchart showing the operation of the continuous shooting / image synthesis process executed in the shooting process, and FIG. 5 is executed in the continuous shooting / image synthesis process. It is a flowchart which shows the operation | movement of an image adoption determination process.

(1) Operation of Shooting Process When the digital camera having the above configuration is set to the shooting mode, the CPU 12 executes the shooting process shown in FIG. 3 and proceeds to Step SA1 to perform preview setting. The preview setting is a setting for reading out, in real time, low resolution image data (preview image data) of 30 frames per second written in the image memory 16 from the image sensor 15 and displaying the preview image on the display unit 17. Point to.

  Next, in step SA2, it is determined whether or not the shutter button has been pressed. If the shutter button is not pressed, the determination result is “NO”, the process proceeds to step SA3, and a read control signal for generating preview image data is supplied to the timing control unit 104 (see FIG. 2) of the image sensor 15. Subsequently, in step SA4, the preview image data output from the image sensor 15 and written in the display area of the image memory 16 is read in real time, and the preview image is displayed on the display unit 17. Thereafter, the above steps SA1 to SA4 are repeated until the shutter button is pressed to display a preview image.

  When the shutter button is pressed, the determination result in step SA2 is “YES”, and the flow advances to step SA5 to execute continuous shooting / image composition processing. In the continuous shooting / image synthesis processing, as will be described later, the continuously shot images are synthesized to generate an output image Pout that is corrected for camera shake. Next, in step SA6, the output image Pout created in step SA5 is reduced to a shooting confirmation image having an image size that can be confirmed on the display screen of the display unit 17, and the reduced shooting confirmation image is displayed on the display unit 17. . Thereafter, the process proceeds to step SA7, where the photographed confirmation image is subjected to image compression in the JPEG format or the like, and in the subsequent step SA8, the compressed image data is stored in the memory card 21 via the external memory input / output control unit 20. The process returns to step SA1.

(2) Operation of Continuous Shooting / Image Composition Processing Next, the operation of continuous shooting / image composition processing will be described with reference to FIG. When the continuous shooting / image composition processing is executed via the above-described step SA5 (see FIG. 3), the CPU 12 proceeds to step SB1 shown in FIG. 4 and clears the memory area for writing the output image in the image memory 16. Perform initialization. Note that the size of the output image written in the image memory 16 is set larger than the input image (captured image). This is because when overlapping input images in which positional deviation occurs due to camera shake, overlapping portions of the continuously shot input images can be cut out.

  When the image memory 16 is initialized, the process proceeds to step SB2, and an initial value “1” is set to the number of imaging frames i. Subsequently, in step SB3, the timing control unit 104 (see FIG. 2) of the image sensor 15 is supplied with a read control signal for generating high-resolution image data to capture an image. In step SB4, image data obtained by imaging is temporarily stored in a work memory area in the image memory 16. Next, in step SB5, it is evaluated whether or not there is a large camera shake in the image data of the number of captured frames i captured this time. Based on the evaluation result, whether or not the image data of the number of captured frames i captured this time is employed. judge. It should be noted that the evaluation as to whether or not there is a large camera shake in the image data is performed in an image adoption determination process (see FIG. 5) described later.

  If it is determined in step SB5 that it is not adopted, the determination result is “NO”, the process returns to step SB3, and image capturing is performed again. On the other hand, if it is determined to be adopted, the determination result in step SB5 is “YES”, and the flow proceeds to step SB6. In step SB6, it is determined whether or not the number i of captured frames is greater than “1”, that is, whether or not the image data is the first captured image. If it is the first imaged image data, the number of imaging frames i is the initial value “1”, so the determination result is “NO”, and the process proceeds to Step SB8.

  On the other hand, if the number of imaging frames i is the second or later, the determination result here is “YES”, the process proceeds to step SB7, and a conversion parameter is calculated. The conversion parameter represents coordinate conversion for making the subject of the image Pi photographed this time overlap the subject of the first photographed image P1. Specifically, for example, based on the well-known epipolar geometry, the coordinate system of the image Pi captured this time is converted into the coordinate system of the image P1 captured first.

  Next, in step SB8, the coordinate system of the image Pi captured this time is converted into the coordinate system of the image P1 captured first in accordance with the conversion parameter calculated in step SB7, and the image Pi is added to the image P1 and synthesized. . Next, in step SB9, the number of captured frames i is incremented and stepped, and in the subsequent step SB10, it is determined whether or not the number of stepped captured frames i has reached the prescribed number of continuous shooting. If the number of imaging frames i has not reached the prescribed number for continuous shooting, the determination result in step SB10 is “NO”, and the process returns to the above-described step SB3 to continue continuous shooting.

  On the other hand, when the number of imaging frames i reaches the prescribed number for continuous shooting, the determination result in step SB10 is “YES”, the process proceeds to step SB11, and the brightness that optimizes the brightness of the image added and synthesized in step SB8. The image processing unit 18 is instructed to execute the gain correction and white balance adjustment to generate the output image Pout, and the process is finished.

(2) Operation of Image Adoption Determination Process Next, the operation of the image adoption determination process will be described with reference to FIG. When this process is executed in step SB5 (see FIG. 4) described above, the CPU 12 proceeds to step SC1 shown in FIG. 5 to initialize the parameters used for image adoption determination and capture the above-described continuous shooting. A reduced image Ps obtained by reducing the image of the number of imaging frames i to a predetermined image size is created. Note that the parameter initialization referred to here means resetting a rejection threshold D_LOW and a reference update threshold D_HIGH, which will be described later, to initial values.

  Next, in step SC2, an image Pg is created by removing spike noise and the like from the reduced image Ps using a Gaussian filter or a median filter. Subsequently, in step SC3, the image Pg is filtered by a Laplacian filter (secondary differential filter) Hl to obtain a harmonic component image Pgl.

  In step SC4, as described in the outline of the invention described above, in the harmonic component image Pgl, the maximum value (absolute value) of the harmonic component Pgl in the row direction (x-axis direction) and the column direction (y-axis direction). After obtaining FH (y) and FV (x), the sum of FH (y) and FV (x) is used to calculate the harmonic component accumulated value FHS in the x-axis direction (lateral direction) and the y-axis direction (longitudinal direction). ) To obtain a harmonic component accumulated value FVS. In step SC4, the harmonic component accumulated value FHS and the harmonic component accumulated value FVS are added to calculate the camera shake evaluation amount EBL. The camera shake evaluation amount EBL obtained in this way represents an image with less camera shake as the value increases.

  Subsequently, in step SC5, it is determined whether or not the number i of captured frames is greater than “1”, that is, whether or not the image data is the first captured image. If it is the first captured image data, the determination result is “NO”, the process proceeds to step SC6, and the camera shake evaluation amount EBL calculated in step SC4 is set to the camera shake reference value EBL1. Thereafter, the process proceeds to step SC7, where the composite frame number n is set to the initial value “1”, and in the subsequent step SC8, “initial adoption” is determined. If it is determined as “initial adoption”, the determination result in step SB5 (see FIG. 4) of the above-described continuous shooting / image composition processing is “YES”.

  On the other hand, if the number of captured frames i is the second and subsequent captured image data, the determination result in step SC5 is “YES”, and the flow proceeds to step SC9. In step SC9, the difference value D is calculated by subtracting the camera shake reference value EBL1 obtained from the image data initially captured from the camera shake evaluation amount EBL of the image data captured this time. Next, in step SC10, it is determined whether or not the difference value D is smaller than the rejection threshold D_LOW, that is, whether or not the camera shake of the image data captured this time is larger than the reference. Hereinafter, the operation will be described separately for the case where the hand shake of the image data captured this time is smaller than the reference and the case where it is larger.

<When the camera shake of the image data captured this time is smaller than the standard>
In this case, the determination result in step SC10 is “NO”, and the flow proceeds to step SC11. In step SC11, it is determined whether or not the difference value D is greater than the reference update threshold value D_HIGH, that is, whether or not the camera shake of the image data captured this time is significantly smaller than the reference. If the camera shake of the image data captured this time is not significantly smaller than the reference, the determination result is “NO”, and the flow proceeds to step SC12. In step SC12, the number of combined frames n is incremented and stepped, and in subsequent step SC13, “combined adoption” is determined. If it is determined that “synthetic adoption” is selected, the determination result in step SB5 (see FIG. 4) of the above-described continuous shooting / image composition processing is “YES”.

  On the other hand, if the camera shake of the image data captured this time is significantly smaller than the reference, the determination result in step SC11 is “YES”, and the flow proceeds to step SC14. In step SC14, it is determined whether or not the current composite frame number n is larger than a predetermined value REJ_LOCK, that is, whether or not a key frame that is a reference for camera shake is replaced. If the current composite frame number n is larger than the predetermined value REJ_LOCK and many images have already been composited, the determination result is “YES”. In this case, the key frame is not replaced and the process proceeds to step SC12. After incrementing the composite frame number n, it is determined that “composite adoption” is made at step SC13.

  On the other hand, if the current composite frame number n is smaller than the predetermined value REJ_LOCK and many images are not composited, the determination result in step SC14 is “NO”. In this case, the key frame is replaced. Proceeding to step SC6 described above, the camera shake evaluation amount EBL calculated in step SC4 is reset as the camera shake reference value EBL1. Then, the process proceeds to step SC7, and after the composite frame number n is reset to the initial value “1”, it is determined as “initial adoption” in the subsequent step SC8. If it is determined as “initial adoption”, the determination result in step SB5 (see FIG. 4) of the above-described continuous shooting / image composition processing is “YES”.

<When camera shake of image data taken this time is larger than the standard>
In this case, the determination result in step SC10 is “YES”, and the flow proceeds to step SC15. In step SC15, it is determined whether or not the adoption ratio n / i obtained by dividing the composite frame number n by the imaging frame number i is greater than a predetermined value TYP_RATE. If the adoption ratio n / i is greater than the predetermined value TYP_RATE, the determination result is “YES”, the process proceeds to step SC18, and the adoption of the image data captured this time is “rejected”. When it is determined as “rejected”, the determination result in step SB5 (see FIG. 4) of the above-described continuous shooting / image composition processing is “NO”.

  On the other hand, if the employment ratio n / i is smaller than the predetermined value TYP_RATE, the determination result in step SC15 is “NO”, and the flow proceeds to step SC16. In step SC16, the rejection threshold D_LOW is relaxed. That is, if the adoption ratio n / i is smaller than the predetermined value TYP_RATE, the rejection threshold D_LOW may be too high, so the rejection threshold D_LOW is reduced to ease the determination criteria.

  In step SC17, based on the new rejection threshold value D_LOW whose criteria are relaxed, it is determined whether or not the camera shake of the image data captured this time is larger than the criterion. If the camera shake is smaller than the reference, the determination result is “NO”, the process proceeds to step SC12, the combined frame number n is incremented and stepped, and in the subsequent step SC13, “composite adoption” is determined. On the other hand, if the camera shake is larger than the reference, the determination result in step SC17 is “YES”, and the process proceeds to step SC18 to “reject” the adoption of the image data captured this time. The relaxed rejection threshold D_LOW is stored and used for the next and subsequent imaging. Further, in this embodiment, the adoption determination is immediately performed in SC17 for the relaxed D_LOW, but it is long-term just to save the relaxed D_LOW without using the determination in SC17 and use it for the next shooting. If it sees, it can raise a substantially equivalent effect. Furthermore, if this rejection threshold value D_LOW is managed and used together with photographing information such as an aperture, a shutter speed, and a photographer, more accurate determination can be made.

  As described above, in the present embodiment, when continuously captured images are combined to perform camera shake correction, the first captured image is unconditionally adopted as a key frame, and harmonic components of the image are used. An evaluation amount EBL representing is calculated. Similarly, the evaluation amount EBLi (i ≧ 2) is obtained for the second and subsequent frames, and a difference value D (EBLi−EBL1) from the evaluation amount EBL1 of the key frame is sequentially calculated. An image having a difference value D smaller than the rejection threshold D_LOW has a large amount of camera shake, and is rejected as an image to be combined. As a result, even if a large camera shake occurs during shooting, it is possible to correct the camera shake.

In this embodiment, the harmonic component of the photographed image is calculated to estimate the amount of camera shake in the image, and an image with large camera shake is excluded from the synthesis target. Further, it is possible to adopt an aspect in which an image with a large amount of camera shake is excluded from the synthesis target by using the conversion parameter calculated in step SB7 (see FIG. 4) of the continuous shooting / image synthesis process described above.
That is, the positional deviation between the two images is obtained from the conversion parameter representing the positional relationship between the previously captured image and the currently captured image. If the positional deviation is large, there is a high possibility that camera shake will be included. Therefore, if the positional deviation obtained from the image captured this time exceeds a predetermined value, the image is excluded from the objects of synthesis. This makes it possible to correct camera shake even when large camera shake occurs during shooting.

It is a block diagram which shows the whole structure of one Embodiment of this invention. 2 is a block diagram showing a configuration of an image sensor 15. FIG. It is a flowchart which shows operation | movement of an imaging | photography process. It is a flowchart which shows the operation | movement of continuous shooting and image composition processing. It is a flowchart which shows the operation | movement of an image adoption determination process.

Explanation of symbols

11 Operation unit 12 CPU
13 Program Code Storage Unit 14 Optical Lens Unit 15 Image Sensor 16 Image Memory 17 Display Unit 18 Image Processing Unit 19 PC I / F
20 External memory input / output controller 21 Memory card

Claims (10)

An extraction means for extracting an evaluation amount that changes in accordance with the amount of camera shake in the image from the captured image;
A determination unit that determines whether to use the evaluation amount extracted by the extraction unit as an image to be combined for each captured image;
An image synthesizing apparatus comprising: an image synthesizing unit that synthesizes each of the images that are determined to be adopted by the determining unit among the continuously photographed images and corrects camera shake. Extraction means for extracting an evaluation amount EBL representing a harmonic component that changes in accordance with the amount of camera shake in the image, from each of the images that are continuously captured;
Each time the extraction unit extracts an evaluation amount of each image that is continuously photographed, an evaluation amount EBLi (i ≧ 2) of each image photographed after the second from the evaluation amount EBL1 of the first photographed image. Difference value generating means for sequentially generating a difference value D (EBLi-EBL1) obtained by subtracting
A determination unit that determines whether or not to employ a difference value D (EBLi-EBL1) generated by the difference value generation unit as an image to be combined for each captured image;
An image synthesizing apparatus comprising: an image synthesizing unit that synthesizes each of the images that are determined to be adopted by the determining unit among the continuously photographed images and corrects camera shake. The extraction means includes
Image reduction means for reducing the captured image;
The apparatus according to claim 2, further comprising: harmonic component image generating means for detecting a harmonic component and generating a harmonic component image after removing noise from the reduced image reduced by the image reducing means. Image composition device. If the difference value D (EBLi-EBL1) generated by the difference value generation means is larger than a predetermined threshold, the determination means adopts the image of the evaluation amount EBLi captured this time as an image to be combined, The image synthesizing apparatus according to claim 2, wherein if it is smaller than a predetermined threshold, the image is not adopted. The determination means includes
Updating means for updating the evaluation amount EBLi of the currently captured image to the evaluation amount EBL1 when the difference value D (EBLi−EBL1) is larger than the first threshold and the number of images determined to be adopted is smaller than a predetermined value; ,
When the difference value D (EBLi−EBL1) is smaller than the second threshold value and the adoption ratio n / i obtained by dividing the number n of adopted images by the number i of captured images is smaller than a predetermined value, The image synthesizing apparatus according to claim 5, further comprising: a threshold value changing unit that sets a third threshold value obtained by changing the second threshold value. An extraction process for extracting an evaluation amount that changes in accordance with the amount of camera shake in the image from the captured image;
A determination process for determining whether to use the evaluation amount extracted by the extraction process as an image to be combined for each captured image;
An image composition processing program for causing a computer to execute image composition processing for correcting camera shake by compositing images determined to be adopted by the determination processing among images photographed continuously. An extraction process for extracting an evaluation amount EBL representing a harmonic component that changes in accordance with the amount of camera shake in the image, from each of the continuously photographed images;
Each time the extraction process extracts an evaluation amount of each image that is continuously photographed, an evaluation amount EBLi (i ≧ 2) of each image photographed after the second from the evaluation amount EBL1 of the first photographed image. Difference value generation processing for sequentially generating a difference value D (EBLi-EBL1) obtained by subtracting
A determination process for determining whether or not to employ the difference value D (EBLi-EBL1) generated by the difference value generation process as an image to be combined for each captured image;
An image composition processing program for causing a computer to execute image composition processing for correcting camera shake by compositing images determined to be adopted by the determination processing among images photographed continuously. The extraction process is:
Image reduction processing for reducing the captured image;
9. A harmonic component image generation process for detecting a harmonic component and generating a harmonic component image after removing noise from the reduced image reduced by the image reduction process. Image composition processing program. If the difference value D (EBLi-EBL1) generated by the difference value generation process is larger than a predetermined threshold, the determination process adopts the image of the evaluation amount EBLi captured this time as an image to be synthesized, The image composition processing program according to claim 8, wherein the image composition processing program is not adopted if it is smaller than a predetermined threshold value. The determination process includes
An update process for updating the evaluation amount EBLi of the currently captured image to the evaluation amount EBL1 when the difference value D (EBLi−EBL1) is larger than the first threshold and the number of images determined to be adopted is smaller than a predetermined value; ,
When the difference value D (EBLi−EBL1) is smaller than the second threshold value and the adoption ratio n / i obtained by dividing the number n of adopted images by the number i of captured images is smaller than a predetermined value, The image composition processing program according to claim 11, further comprising: a threshold value changing process for setting a third threshold value obtained by changing the second threshold value.

Images