JP2009010711A - Device and method for filter processing, and imaging device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To control ringing due to filter processing. <P>SOLUTION: In the filter processing, a local area L around a target pixel Pa is set within an image with changing the position of the target pixel Pa, and a pixel value Xa of the target pixel Pa is converted according to a specified operation result based on pixel values within the set local area L. Before the operation, the pixel value Xb of each surrounding pixel Pb rather than the target pixel Pa within the local area L is subjected to limit processing. The surrounding pixel Pb is sequentially selected from the local area L, and then the difference ¾Xa-Xb¾ with respect to the pixel value Xa of the target pixel Pa is calculated. The pixel value Xb is changed so that the difference ¾Xa-Xb¾ is configured to be a limit value Xr (or less than the limit value Xr) for the surrounding pixel Pb, the difference ¾Xa-Xb¾ of which is greater than the limit value Xr. The limit value Xr is set so as to be linked to the surrounding pixel Pb and monotonically decrease with separating from the target pixel Pa. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a filter processing device and method for performing filter processing on an image as preprocessing prior to face detection processing and the like, and an imaging device including the filter processing device.

  2. Description of the Related Art In recent years, a photographing device such as a digital camera has automatically detected and detected a person's face so that an image in which the face portion is in focus and exposure can be easily and reliably photographed in a photographing scene including the person. Some have a face detection function that performs focus adjustment and exposure adjustment on the face area. This face detection is performed by sequentially cutting out partial images from an image and determining whether or not the cut out partial images are faces based on previously learned data (see, for example, Patent Document 1).

  In the above-described photographing apparatus, when the photographing condition is bad and the brightness and contrast of the face portion are not good, there is a problem that the face detection rate is lowered. For example, at the time of backlight photographing, as shown in FIG. 14A, the face area is dark and the contrast is lowered, so that the detection rate is extremely lowered.

In order to solve such a problem, it has been proposed to perform a filtering process called a normalization process on the image so that the contrast of the image is adjusted to a predetermined level suitable for the face detection process. As this normalization processing, a local area of a predetermined size (for example, 7 × 7 pixels) centered on the target pixel is set in the image, and the pixel value distribution in the local area is changed while changing the position of the target pixel. A method of converting (gradation conversion) the pixel value (luminance value) of the target pixel is known in order to make the degree of the above constant at a certain level. When this normalization process is performed on the image shown in FIG. 14A, the face area is enlarged in brightness and contrast and the facial features are sharpened as shown in FIG. The rate is improved.
JP 2007-25766 Gazette

  However, there is a problem that ringing occurs in the contour portion of the subject as shown in FIG. In the region where the ringing occurs (ringing region), the pixel value changes greatly, and overshoot / undershoot occurs. For this reason, when the size of the face is small like the person on the right side, the ratio of the ringing area to the face area is large, and the ringing has a large influence, so that the face detection rate is particularly lowered.

  This problem due to ringing also occurs when edge detection processing (feature extraction processing) is performed as filter processing. Needless to say, this problem also occurs when a specific part of the subject other than the face is detected.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a filter processing device and method capable of suppressing the occurrence of ringing, and an imaging device including the filter processing device.

  In order to achieve the above object, the filter processing apparatus of the present invention sequentially sets a local region around the target pixel in the image while changing the position of the target pixel, and sets each pixel in the set local region. In the filter processing device that converts the pixel value of the target pixel based on the result of the predetermined calculation process based on the value, prior to the calculation process, the pixel value of each peripheral pixel other than the target pixel in the local region There is provided a limit processing means for changing so that a difference from the pixel value becomes equal to or less than a limit value set in association with each peripheral pixel.

  The filter processing device of the present invention is set by a local region setting unit that sequentially sets a local region around the target pixel in the image while changing the position of the target pixel, and the local region setting change unit. A peripheral pixel selecting unit that sequentially selects each peripheral pixel other than the target pixel from within the local region, and a difference for calculating a difference between the pixel value of the peripheral pixel selected by the peripheral pixel selecting unit and the pixel value of the target pixel A calculating unit; a determining unit that determines whether or not the difference calculated by the difference calculating unit is greater than a limit value set in association with each peripheral pixel; and the determination unit determines that the difference is greater than a limit value A peripheral pixel value changing unit that changes a pixel value of the peripheral pixel so that the difference is less than or equal to a limit value, and an image of the peripheral pixel changed by the peripheral pixel value changing unit. A pixel-of-interest conversion means for performing predetermined arithmetic processing based on each pixel value in the local region including the value and converting the pixel value of the target pixel based on the result of the arithmetic processing. .

  The determination unit preferably sets the limit value so as to monotonously decrease as the distance from the target pixel increases.

  Moreover, it is preferable that the surrounding pixel value changing unit changes the pixel value of the surrounding pixel so that the difference matches the limit value when the determination unit determines that the difference is larger than the limit value.

  The surrounding pixel value changing unit preferably changes the pixel value of the surrounding pixel so that the difference is zero when the determining unit determines that the difference is larger than the limit value.

  Further, the determination means sets first and second limit values for each peripheral pixel, and the first and second limit values are based on the relationship between the pixel value of each peripheral pixel and the pixel value of the target pixel. It is preferable to alternatively use the limit value as the limit value.

  Further, the determination means sets each first limit value to a value larger than the corresponding second limit value, and averages the difference between the pixel value of each peripheral pixel and the pixel value of the target pixel Is equal to or greater than a predetermined threshold value, the first limit value is used. If the average difference between the pixel value of each peripheral pixel and the pixel value of the target pixel is less than the threshold value, the second limit value is used. More preferably, the value is used.

  Further, it is preferable that the pixel-of-interest conversion unit performs a local normalization process or an edge detection process.

  In addition, the filter processing method of the present invention sequentially sets a local region around the target pixel in the image while changing the position of the target pixel, and performs a predetermined calculation based on each pixel value in the set local region. In the filter processing method for converting the pixel value of the target pixel based on the processing result, prior to the calculation process, the pixel values of the surrounding pixels other than the target pixel in the local region are compared with the pixel values of the target pixel. Is changed to be equal to or less than a limit value set in association with each peripheral pixel.

  In addition, a photographing apparatus according to the present invention includes the above-described filter processing device, and performs a filtering process on the image obtained by the image sensor with the filter processing device, and then detects a specific portion of the subject from the image. In the apparatus, it is determined whether or not the image is in a backlight state by multi-division metering, the limit process is performed when the image is in a backlight state, and the limit process is performed when the image is not in a backlight state. Control means for controlling the filter processing apparatus so as not to perform the control is provided.

  The specific portion is preferably a human face.

  In the present invention, prior to the calculation processing of the filter processing, the difference between the pixel value of each peripheral pixel other than the target pixel in the local region and the pixel value of the target pixel is equal to or less than the limit value set in association with each peripheral pixel. Therefore, it is possible to suppress the occurrence of ringing caused by performing the filtering process on a region (such as a contour portion of a subject at the time of backlighting) where the difference between the pixel values of the target pixel and the surrounding pixels is large. In addition, by setting the limit value so as to monotonously decrease as the distance from the target pixel increases, it is possible to reduce the influence of peripheral pixels that are distant from the target pixel, and to more effectively suppress the occurrence of ringing. Also, first and second limit values are set for each peripheral pixel, and the first and second limit values are set as the limit values based on the relationship between the pixel value of each peripheral pixel and the pixel value of the target pixel. By using it alternatively, the occurrence of ringing can be more effectively suppressed.

  Further, it is determined whether or not the image is in a backlight state by multi-division metering, and the limit process is performed when the image is in a backlight state, and the limit process is not performed when the image is not in a backlight state. By configuring the photographing device that controls the filter processing device, the efficiency of the filter processing can be improved.

  In FIG. 1, the digital camera 10 includes an operation unit 11, a CPU 12, a focus lens 13, an aperture 14, an image sensor 15, a lens driver 16, a lens motor 17, an aperture driver 18, an aperture motor 19, an image processing circuit 20, and a built-in memory 21. AE / AF calculation circuit 22, face detection circuit 23, display circuit 24, memory card slot 25, monitor 26, and the like.

  The operation unit 11 includes a release button for the user to give a shooting instruction to the digital camera 10, a cross key and a determination button for performing various setting operations, a face detection execution button for executing face detection, and the like. An operation signal corresponding to the operation is transmitted to the CPU 12. Based on the operation signal input from the operation unit 11, the CPU 12 executes a program stored in the memory in the CPU 12 and controls each unit of the digital camera 10.

  The focus lens 13 forms a subject image on the light receiving surface of the image sensor 15 through the diaphragm 14. The lens driver 16 drives the lens motor 17 based on a command from the CPU 12, moves the focus lens 13 along the optical axis direction, and performs focus adjustment. The diaphragm 14 increases or decreases the amount of light incident on the image sensor 15 according to the open / close state. The aperture driver 18 drives the aperture motor 19 based on a command from the CPU 12 to change the open / close state of the aperture 14 and adjust the amount of incident light.

  The image sensor 15 is a CCD type or CMOS type image sensor that photoelectrically converts a subject image incident through the focus lens 13 and the diaphragm 14 and converts the subject image into an electrical image signal, and performs imaging based on a command from the CPU 12. I do. The image sensor 15 has an electronic shutter function for electronically adjusting the exposure time (charge accumulation time). The image processing circuit 20 performs image processing on the image signal output from the image sensor 15 to generate image data. Specifically, the image processing circuit 20 performs analog / digital conversion on the image signal output from the image sensor 15 to generate a digital signal, and then performs color interpolation, YC (luminance / color difference) conversion, gamma correction, and the like. Various signal processing is performed, and the generated image data is recorded in the built-in memory 21. As the built-in memory 21, for example, a non-volatile semiconductor memory such as a flash memory is used.

  Based on the image data recorded in the built-in memory 21, the AE / AF arithmetic circuit 22 is AE data (subject brightness) for automatic exposure adjustment (AE) and AF data (automatic focus adjustment (AF) data ( The contrast value is calculated for a predetermined area in the image, and the calculated value is input to the CPU 12. When executing AE, the CPU 12 controls the aperture driver 18 and the image sensor 15 based on AE data input from the AE / AF arithmetic circuit 22 and adjusts the exposure amount determined by the aperture value and the shutter speed. Further, when executing AF, the CPU 12 controls the lens driver 16 based on the AF data input from the AE / AF arithmetic circuit 22 and moves the focus lens 13 to a position where the AF data is maximized. Adjust (focus adjustment).

  As will be described in detail later, the face detection circuit 23 detects the position and size of the face area of the subject based on the image data recorded in the built-in memory 21. When the face detection function of the digital camera 10 is enabled, the AE / AF calculation circuit 22 calculates AE data and AF data from the face area detected by the face detection circuit 23, and the CPU 12 AE / AF control is executed based on the calculation result. When a plurality of face areas are detected from one image, the AE control is performed on all face areas, but the AF control is performed based on various conditions (the size and position of the face area). A face area is selected, and the selected face area is selected as a target.

  The display circuit 24 generates screen data to be displayed on the monitor 26 based on the image data stored in the built-in memory 21 or the image data stored in the memory card installed in the memory card slot 25. The monitor 26 is an image display device including a liquid crystal display and displays the screen data generated by the display circuit 24. The memory card slot 25 writes, reads, or erases image data with respect to the inserted memory card.

  In FIG. 2, the face detection circuit 23 includes a luminance signal extraction unit 30, a multi-resolution conversion unit 31, a filter processing unit 32, a face detection processing unit 34, and a duplicate detection determination unit 35. A limit processing unit 33 is provided. The luminance signal extraction unit 30 extracts only the luminance signal (Y signal) from the image data stored in the built-in memory 21, and generates a gray-scale image S0. Here, the number of gradations of luminance is 8 bits (pixel value range is 0 to 255).

The multi-resolution unit 31 normalizes the rectangular side having a short side of 416 pixels by converting the resolution (image size) of the image S0 generated by the luminance signal extraction unit 30, and then normalizes the image. By further converting the resolution based on S0 ′, a multi-resolution image group S1 composed of a plurality of images having different resolutions is generated. Specifically, as shown in FIG. 3, by the image S0 'as a reference image S1_1, to reduce the short sides and long sides by ^{two -1/3} times, the short side respectively, 330 pixels, 262 pixels, A plurality of images S1_2, S1_3, S1_4, S1_5,... With 208 pixels, 165 pixels,... Are generated and set as a multi-resolution image group S1. As described above, the reason why the multi-resolution image group S1 is generated is to detect faces of various sizes using a face sample image having a certain size.

  The filter processing unit 32 performs normalization processing on each image of the multi-resolution image group S1 as pre-processing for improving the accuracy of face detection so as to suppress variations in brightness and contrast. Specifically, the processing is performed according to the procedure shown in FIG. First, one image is selected from the multi-resolution image group S1 (step S1). Next, as shown in FIG. 5, one pixel in the selected image is set as the target pixel Pa, and a local region L having a predetermined size (7 × 7 pixels) is set around the target pixel Pa (step S2). ).

  Next, in the set local region L, limit processing is performed on the peripheral pixels Pb other than the target pixel Pa by the limit processing unit 33 (step S3). Specifically, as illustrated in FIG. 6, the limit processing unit 33 first selects one peripheral pixel Pb (step S <b> 3-1), and calculates a difference | Xa−Xb | of the pixel value from the target pixel Pa. Calculate (step S3-2). Here, Xa is the pixel value of the target pixel Pa, and Xb is the pixel value of the peripheral pixel Pb. Next, it is determined whether or not the difference | Xa−Xb | is greater than the limit value Xr (Xr> 0) (step S3-3). As shown in FIG. 7, the limit value Xr is associated with the position of each peripheral pixel Pb, and is set so as to decrease monotonously as the distance from the position of the target pixel Pa increases.

  If it is determined in step S3-3 that the difference | Xa-Xb | is greater than the limit value Xr, the difference sign direction (the magnitude relationship between the target pixel value Xa and the peripheral pixel value Xb) is kept the same. The peripheral pixel value Xb is converted (steps S3-4 to S3-6). Specifically, first, it is determined whether or not the target pixel value Xa is larger than the peripheral pixel value Xb (step S3-4). If it is determined in step S3-4 that the target pixel value Xa is larger than the peripheral pixel value Xb, the peripheral pixel value Xb is expressed by the formula “Xb” so that the difference between the target pixel value Xa and the target pixel value Xa is the limit value Xr. It changes based on '= Xa + Xr' (step S3-5). When it is determined in step S3-4 that the target pixel value Xa is equal to or smaller than the peripheral pixel value Xb, the peripheral pixel value Xb is expressed by the formula “in order to set the difference between the target pixel value Xa and the target pixel value Xa as the limit value Xr. Change based on “Xb ′ = Xa−Xr” (step S3-6). Here, Xb ′ is a pixel value after the change of the peripheral pixel Pb.

  On the other hand, if it is determined in step S3-3 that the difference | Xa−Xb | is equal to or smaller than the limit value Xr, the surrounding pixel value Xb is not changed. Then, it is determined whether or not the peripheral pixel Pb selected in step S3-1 is the last pixel in the local region L (step S3-7). If it is determined in step S3-7 that the peripheral pixel Pb is not the final pixel, the process returns to step S3-1, and the next pixel in the same local region L is selected as the peripheral pixel Pb. On the other hand, if it is determined in step S3-7 that the peripheral pixel Pb is the final pixel, the process proceeds to step S4 in the flowchart of FIG. As described above, since the limit processing is performed more severely in the local region L with respect to the local pixel Pb having a greater distance from the target pixel Pa in each local region L, the pixel of the peripheral pixel Pb that is separated from the target pixel Pa in the subsequent normalization processing. “Ringing” caused by a strong correlation between the difference between the value Xb and the target pixel value Xa is suppressed.

  In step S4, a variance Vlocal of pixel values (including Xa) in the local region L is calculated. Next, it is determined whether or not the calculated variance Vlocal is greater than or equal to the threshold value C1 (step S5). If it is determined in step S5 that the variance Vlocal is greater than or equal to the threshold C1, the pixel value Xa of interest is converted based on the following equation (1) as the first gradation conversion processing (step S6). . On the other hand, when it is determined in step S5 that the variance Vlocal is less than the threshold value C2, the pixel value Xa of interest is converted based on the following equation (2) as the second gradation conversion process (step S5). S7).

  Here, Xa ′ is a pixel value after conversion of the target pixel Pa, mlocal is an average of pixel values (including Xa) in the local region L, and SDlocal is a standard deviation of pixel values (including Xa) in the local region L. (Square root of variance Vlocal), C2 is a reference value, and SDc is a predetermined constant. That is, the above equation (1) performs conversion so that the degree of dispersion approaches a constant level (C2 × C2) higher than the predetermined level for the local region L where the distribution Vlocal is equal to or higher than the predetermined level C1. The above equation (2) performs conversion so that the degree of dispersion is suppressed to a level lower than a certain level (C2 × C2) for the local region L where the dispersion Vlocal is less than the predetermined level C1. When calculating the variance Vlocal and the average mlocal in steps S4, S6, and S7, the pixel value Xb 'after the change is used for the peripheral pixel Pb subjected to the limit process in step S3.

  Steps S <b> 5 to S <b> 7 are local normalization processing that performs processing for each unit of the local region L. When this processing ends, next, whether or not the target pixel Pa set in step S <b> 2 is the last pixel in the image. Is determined (step S8). If it is determined in step S8 that the target pixel Pa is not the final pixel, the process returns to step S2, the next pixel on the same image is set as the target pixel Pa, and at the same time, the new target pixel Pa is centered. A local region L is set. Note that when a new local region L is set, the pixel values in the local region L are the pixel values Xa and Xb before conversion. On the other hand, if it is determined in step S8 that the target pixel Pa is the final pixel, it is determined whether or not the image selected in step S1 is the final image of the multi-resolution image group S1 (step S9). . If it is determined in step S9 that the image is not the final image, the process returns to step S1 to select the next image. On the other hand, if it is determined in step S9 that the image is the final image, the filtering process for the multi-resolution image group S1 is terminated. That is, as shown in FIG. 8, by performing the above filter processing on each image of the multi-resolution image group S1, the multi-resolution image group S1 ′ subjected to appropriate normalization processing while suppressing occurrence of ringing is obtained. can get.

  Returning to FIG. 2, the face detection processing unit 34 shows the images S1′_1, S1′_2, S1′_3,... Of the multi-resolution image group S1 ′ generated by the filter processing unit 32 as shown in FIG. As described above, the partial image W having a predetermined size (32 × 32 pixels) is sequentially cut out while changing the cutout position pixel by pixel, and it is determined whether the cut out partial image W is a face image. Detect the upper face area. Whether or not the partial image W is a face image is determined by using a plurality of face sample image groups that are known to be faces and a plurality of non-face sample groups that are known not to be faces. This is done using a discriminator that has learned the feature of being. This discriminator is disclosed in detail in Japanese Patent Application Laid-Open No. 2007-25766.

  In the face detection processing by the face detection processing unit 34, the same face may be detected repeatedly between different images of the multi-resolution image group S1 '. This is because the size of the face that can be detected with respect to the size of the partial image W has a certain width. Based on the position information of the face area detected by the face detection processing unit 34, the overlap detection determination unit 35 selects one of the same face areas detected on each image of the multi-resolution image group S1 ′. A process of grouping as a face area is performed, and the position and size of the true face area on the image S0 are specified. This face detection information is input to the CPU 12.

  Next, the operation of the digital camera 10 configured as described above will be described. When the digital camera 10 is set to the still image shooting mode, the image pickup device 15 performs an image pickup operation at a predetermined cycle, and the image processed by the image processing circuit 20 is sequentially displayed on the monitor 26 as a through image. indicate. The user can perform framing of the subject using the through image displayed on the monitor 26. At this time, when the user presses a face detection execution button included in the operation unit 11, the digital camera 10 automatically detects a person's face from the image and automatically detects the detected face area as described above. So-called “face recognition AF & AE” is executed to set the focus and exposure. When the user presses the shutter button, an image in which focus and exposure are optimized with respect to the face area is acquired as data, and is recorded in the internal memory 21 or a memory card attached to the memory card slot 25.

  When the digital camera 10 performs face detection in response to pressing of the face detection execution button, in order to increase the accuracy of face detection, normalization processing is performed on the image as preprocessing. For each peripheral pixel Xb in L, since the limit process is performed to change the pixel value Xb so that the difference from the target pixel value Xa is within the limit value Xr, ringing that occurs around the face area is suppressed, The detection rate of face detection is improved. In particular, the detection rate is dramatically improved for a small face area that is greatly affected by ringing. This ringing suppression method is effective not only for the above-described backlight but also for the influence of spotlights and partial reflection of reflected light.

  In the above embodiment, the limit value Xr used for the limit process is changed according to the position of the peripheral pixel Pb as shown in FIG. 7, but the present invention is not limited to this and is limited. The value Xr may be a fixed value that does not depend on the position of the peripheral pixel Pb. In this case as well, the effect of suppressing ringing can be obtained. FIG. 10 shows an example in which the limit value Xr is fixed to “64”.

  In the above embodiment, one limit value Xr is set for one peripheral pixel Pb. However, the present invention is not limited to this, and a plurality of limits are set for one peripheral pixel Pb. Limit processing may be performed by setting values and switching these limit values based on various conditions. For example, the first limit value Xr_1 is set as shown in FIG. 11A, the second limit value Xr_2 is set as shown in FIG. 11B, and these are set as the target pixel value Xa and the peripheral pixel value. It is used by switching from the relationship with Xb. Note that the second limit value Xr_2 is set smaller than the first limit value Xr_1 at the position of the same peripheral pixel Pb.

  FIG. 12 shows a specific switching procedure between the first and second limit values Xr_1 and Xr_2. This procedure is performed by the limit processing unit 33 between step S3-2 and step S3-3 in FIG. That is, after step S3-2, the limit processing unit 33 calculates the average Ave (| Xa-Xb |) of the difference | Xa-Xb | in the local region L (step S10). Next, it is determined whether or not the calculated average Ave (| Xa−Xb |) is equal to or greater than a predetermined threshold C3 (step S11). In step S11, when the average Ave (| Xa−Xb |) is equal to or greater than the threshold value C3, the first limit value Xr_1 is selected and used as the limit value Xr in subsequent steps. On the other hand, if the average Ave (| Xa−Xb |) is less than the threshold value C3 in step S11, the second limit value Xr_2 is selected and used as the limit value Xr in subsequent steps. Thereby, it is possible to more effectively suppress the influence of the peripheral pixel Pb having a large difference | Xa−Xb |.

  In the above embodiment, when the difference | Xa−Xb | is larger than the limit value Xr, the pixel value Xb of the surrounding pixel Pb is changed so that the difference from the target pixel value Xa becomes the limit value Xr (see FIG. However, the present invention is not limited to this, and the pixels of the peripheral pixel Pb are set so that the difference from the target pixel value Xa is less than the limit value Xr. The value Xb may be changed. For example, when the difference | Xa−Xb | is larger than the limit value Xr, the pixel value Xb of the surrounding pixel Pb is replaced with the value of the target pixel value Xa so that the difference becomes zero (that is, “Xb ′ = Xa ") is also preferable.

  In the above embodiment, when performing the face detection process, the limit process (step S3 in FIG. 4) is performed regardless of the exposure state of the image, but the present invention is not limited to this. Limit processing may be switched on / off based on the exposure state of the image. For example, as shown in FIG. 13, the CPU 12 first performs multi-division photometry (photometry method that divides into a plurality of areas and calculates an exposure value for each area) based on the image S0 (step S20). By comparing the luminance difference between the central portion and the peripheral portion of the image S0, it is determined whether or not the backlight is in the backlit state (step S21). In step S21, when the luminance at the central portion is lower than the luminance at the peripheral portion by a predetermined value or more in the backlight state, the limit processing by the limit processing unit 33 is set to ON (step S22). On the other hand, in step S21, when there is no luminance difference between the central portion and the peripheral portion and the backlight is not in the backlit state, the limit processing by the limit processing unit 33 is set to OFF (step S23). When it is set to OFF in step S23, the limit process of step S3 is skipped in the flowchart of FIG. As a result, the limit process is executed only when the exposure state of the image is poor and ringing is likely to occur, so that the efficiency of the filter process can be improved.

  In the above embodiment, the normalization process is described as an example of the filter process performed by the filter processing unit 32. However, the present invention is not limited to this, and the edge detection process is performed as the filter process. It is also applicable to cases. As this edge detection process, for example, a filter process using a Laplacian filter is known. In this Laplacian filter processing, a local region centered on the pixel of interest is set, each pixel in the local region is multiplied by a filter coefficient, and all pixel values multiplied by the coefficient are added as a new pixel value of interest. Arithmetic processing. Ringing is a problem common to filter processing for converting a target pixel value by a calculation based on a pixel value in a local region centered on the target pixel, and ringing can be suppressed by applying the present invention.

  In the above-described embodiment, the image processing circuit 20, the AE / AF arithmetic circuit 22, and the face detection circuit 23 are shown as functional circuits as hardware circuits. However, the present invention is not limited to this. Instead, it may be configured by a program describing each processing procedure described above and a computer that executes processing based on the program.

  In the above embodiment, specific numerical values are shown as the size of each image of the multi-resolution image group S1, the size of the local region L, and the size of the partial image W. However, these are merely examples and can be changed as appropriate. is there.

  In the above-described embodiment, the face detection process using a human face as a detection target is illustrated. However, the present invention is not limited to a face, and can be applied to a specific part detection process for detecting a specific part of a subject.

  Finally, the correspondence between the constituent elements in the claims and the embodiments will be described. The local region setting means corresponds to the function of the filter processing unit 32 represented by steps S2 and S8 in FIG. The peripheral pixel selection means corresponds to the function of the limit processing unit 33 represented in steps S3-1 and S3-7 in FIG. The difference calculation means corresponds to the function of the limit processing unit 33 represented in step S3-2 in FIG. The determination unit corresponds to the function of the limit processing unit 33 represented in step S3-3 in FIG. The peripheral pixel value changing unit corresponds to the function of the limit processing unit 33 expressed in steps S3-4 to S3-6 in FIG. The pixel-of-interest conversion unit corresponds to the function of the filter processing unit 32 represented in steps S4 to S7 in FIG. And a control means respond | corresponds to the function of CPU12 represented with the flowchart of FIG.

It is a block diagram which shows the structure of a digital camera. It is a block diagram which shows the structure of a face detection circuit. It is a figure which shows the production | generation process of a multi-resolution image group. It is a flowchart which shows the procedure of a normalization process. It is a figure which shows the example of the local area | region set centering on an attention pixel. It is a flowchart which shows the procedure of a limit process. It is a figure which illustrates the limit value used for a limit process. It is a figure which shows the normalization process notionally. It is a figure which shows a face detection process notionally. It is a figure which shows the example which fixed the limit value irrespective of the position of the surrounding pixel. It is a figure which illustrates two limit values used switching about the same surrounding pixel, (A) shows the 1st limit value and (B) the 2nd limit value. It is a flowchart which shows the switching procedure of the 1st and 2nd limit value. It is a flowchart which shows the example which switches a limit process based on the exposure state of an image. It is a figure for demonstrating the problem of a prior art.

Explanation of symbols

10 Digital camera (photographing device)
12 CPU (control means)
DESCRIPTION OF SYMBOLS 15 Image pick-up element 20 Image processing circuit 21 Built-in memory 22 AE / AF arithmetic circuit 23 Face detection circuit 30 Luminance signal extraction part 31 Multi-resolution part 32 Filter processing part (filter processing apparatus)
33 Limit processing section (limit processing means)
34 Face detection processing unit 35 Duplicate detection determination unit

Claims (11)

While changing the position of the pixel of interest, a local area centered on the pixel of interest is sequentially set in the image, and based on the result of a predetermined calculation process based on each pixel value in the set local area, In a filter processing device that converts pixel values,
Prior to the calculation process, the pixel value of each peripheral pixel other than the target pixel in the local region is changed so that the difference from the pixel value of the target pixel is equal to or less than the limit value set in association with each peripheral pixel. A filter processing apparatus comprising a limit processing means. Local area setting means for sequentially setting a local area around the target pixel in the image while changing the position of the target pixel;
Peripheral pixel selection means for sequentially selecting each peripheral pixel other than the target pixel from within the local area set by the local area setting change means;
Difference calculating means for calculating a difference between the pixel value of the peripheral pixel selected by the peripheral pixel selecting means and the pixel value of the target pixel;
Determining means for determining whether or not the difference calculated by the difference calculating means is larger than a limit value set in association with each peripheral pixel;
A peripheral pixel value changing unit that changes the pixel value of the peripheral pixel so that the difference is equal to or less than the limit value when the determination unit determines that the difference is greater than the limit value;
Including the pixel value of the peripheral pixel changed by the peripheral pixel value changing means, a predetermined calculation process is performed based on each pixel value in the local region, and the pixel value of the target pixel is converted based on the result of the calculation process Attention pixel conversion means;
A filter processing apparatus comprising:   The filter processing apparatus according to claim 2, wherein the determination unit sets the limit value so as to monotonously decrease as the distance from the target pixel increases.   The surrounding pixel value changing unit changes the pixel value of the surrounding pixel so that the difference matches the limit value when the determination unit determines that the difference is larger than the limit value. The filter processing device according to 2 or 3.   The peripheral pixel value changing unit changes the pixel value of the peripheral pixel so that the difference becomes zero when the determination unit determines that the difference is larger than a limit value. 4. The filter processing device according to 3.   The determination means sets first and second limit values for each peripheral pixel, and the first and second limit values are based on the relationship between the pixel value of each peripheral pixel and the pixel value of the target pixel. The filter processing apparatus according to claim 2, wherein a value is alternatively used as the limit value.   The determination unit sets each first limit value to a value larger than the corresponding second limit value, and an average of a difference between a pixel value of each peripheral pixel and a pixel value of a target pixel is predetermined. The first limit value is used when the threshold value is equal to or greater than the threshold value, and when the average difference between the pixel values of the surrounding pixels and the pixel value of the target pixel is less than the threshold value, the second limit value is set. The filter processing apparatus according to claim 6, wherein the filter processing apparatus is used.   The filter processing apparatus according to claim 2, wherein the pixel-of-interest conversion unit performs a local normalization process or an edge detection process. While changing the position of the pixel of interest, a local area centered on the pixel of interest is sequentially set in the image, and based on the result of a predetermined calculation process based on each pixel value in the set local area, In a filter processing method for converting a pixel value,
Prior to the calculation process, the pixel value of each peripheral pixel other than the target pixel in the local region is changed so that the difference from the pixel value of the target pixel is equal to or less than the limit value set in association with each peripheral pixel. A filtering method characterized by the above. An imaging device comprising the filter processing device according to claim 1, wherein the image processing device detects a specific portion of the subject from the image after performing filter processing on the image obtained by the image sensor with the filter processing device.
It is determined whether or not the image is in a backlight state by multi-division metering, and the limit process is performed when the image is in a backlight state, and the limit process is not performed when the image is not in a backlight state. An imaging apparatus comprising: control means for controlling the filter processing apparatus.   The photographing apparatus according to claim 10, wherein the specific portion is a human face.

Images