JP5256236B2 - Image processing apparatus and method, and image processing program - Google Patents

Description

  The present invention relates to an image processing apparatus and method, and an image processing program.

  Purple fringes may be visually recognized in an image obtained by imaging with a digital still camera or the like. When a high brightness subject (for example, a light source) is imaged and a contour (edge) having a large luminance difference (luminance gradient) is present in the captured image, purple fringes tend to occur around the image contour.

  Patent Documents 1 and 2 describe a technique that generates reduced image data obtained by reducing image data and performs purple fringe correction. Since the image data is reduced, the purple fringe detection accuracy deteriorates.

  Japanese Patent Application Laid-Open No. H10-228561 describes a technique for reducing color blur using a smoothing filter or a median filter having a size corresponding to an image size (number of pixels). Patent Document 3 describes that an edge pixel among color blur pixels is determined. However, the degree of change in color difference component between adjacent pixels is used in determining whether or not the pixel is an edge pixel, and purple fringes that are spread around the image contour cannot be detected with high accuracy.

JP 2006-121138 A JP 2006-135744 A JP 2003-102027 A

  An object of the present invention is to accurately detect pixels in which purple fringes may occur.

  An image correction apparatus according to the present invention has different sizes for determining the distance between a pixel to be processed and a luminance calculation target pixel located around the pixel to be processed according to the size of image input means for receiving input of given image data. By scanning a plurality of kernels on a processing target image represented by image data received by the image input means while maintaining a positional relationship in which one includes the other, each processing target pixel and each kernel Further, by each of the horizontal direction, the vertical direction, and the two oblique directions, a luminance difference calculation unit that calculates a luminance difference of luminance calculation target pixels that are symmetrical with respect to the processing target pixel, and the luminance difference calculation unit. Comparing means for comparing the largest luminance difference among a plurality of calculated luminance differences with a predetermined threshold value When the maximum luminance difference is determined to be larger than the threshold value by the comparing means, the processing target pixel is equal to or proceed to decide whether a pixel is generated purple fringing.

  In the image processing method according to the present invention, the image input means accepts input of given image data, and the luminance difference calculation means determines the distance between the processing target pixel and the luminance calculation target pixel located around the processing target pixel according to the size. By scanning a plurality of kernels having different sizes from each other on a processing target image represented by image data received by the image input means while maintaining a positional relationship in which one includes the other, For each pixel to be processed and for each of the kernels, a luminance difference between luminance calculation target pixels at symmetrical positions with respect to the pixel to be processed is calculated for each of a horizontal direction, a vertical direction, and two oblique directions. The largest brightness difference among the plurality of brightness differences calculated by the brightness difference calculating means is compared with a predetermined threshold, and the comparison When the maximum luminance difference is determined to be larger than the threshold value by the step, the processing target pixel is equal to or proceed to decide whether a pixel is generated purple fringing.

  The present invention also provides a program for causing a computer to function as the image processing apparatus.

  According to the present invention, a plurality of kernels having different sizes scan the processing target image while maintaining a positional relationship in which one includes the other. The kernel determines the distance between the processing target pixel and the luminance calculation target pixel located around the processing target pixel, and the distance differs depending on the size of the kernel. Since a plurality of kernels having different sizes are scanned on the processing target image while maintaining the positional relationship including one of the other, the processing target pixel and the surroundings of the processing target pixel are centered on the processing target pixel. Two or more distances with respect to the luminance calculation target pixel located at are determined. In addition, since a plurality of kernels are scanned while maintaining a positional relationship in which one of them includes the other, surrounding pixels surrounding the processing target pixel doubly or more around the processing target pixel are candidates for luminance calculation target pixels. become.

  For each of the horizontal direction, the vertical direction, and the two diagonal directions, the luminance calculation target pixel at the symmetrical position with respect to the processing target pixel is used, and the luminance difference is calculated. Since the luminance difference is calculated for each of the horizontal direction, the vertical direction, and the two diagonal directions, four luminance differences are calculated for one kernel. As described above, since the brightness difference is calculated for each kernel, the number of kernels × 4 brightness differences is calculated.

  Among the plurality of calculated luminance differences, the largest luminance difference is compared with a predetermined threshold value, and when it is determined that the maximum luminance difference is larger than the threshold value, the processing target pixel generates a purple fringe. Proceed to determine if it is a pixel. This is because purple fringes occur in image contours with a large luminance difference, and purple fringes do not occur in image portions with a small luminance difference, or even if they occur, they are hardly visible.

  According to the present invention, by using a plurality of kernels having different sizes, it is possible to accurately detect whether there is an image portion having a large luminance difference around the processing target pixel, that is, an image contour having a large luminance difference. Pixels in which purple fringes may have occurred are subjected to subsequent processing with almost no omission. The accuracy of purple fringe correction can be increased.

  In one embodiment, whether or not the processing target pixel is a pixel that has generated purple fringes is determined based on whether or not the color of the processing target pixel is purple. This is because purple fringe has a purple color as one of its characteristics.

  Preferably, when it is determined that the color of the processing target pixel is purple, a correction coefficient calculation unit that calculates a correction coefficient that changes the color or saturation of the processing target pixel, and the correction calculated by the correction coefficient calculation unit Color correction means for correcting the color of the pixel to be processed using a coefficient is provided. Since the color of the processing coping pixel in which purple fringe is generated is not purple or purple is not noticeable, an image in which purple fringe is suppressed can be obtained.

  In one embodiment, color correction avoiding means for avoiding color correction using the correction coefficient by the color correcting means when the correction coefficient increases the signal amount of the blue component signal of the processing target pixel. Prepare. Purple fringe appears purple because the level of the blue (B) component signal is higher than that of the green (G) component. When the correction coefficient increases the signal amount of the blue component signal of the pixel to be processed, that is, the signal amount of the blue (B) component signal is set to a signal amount far from the signal amount of the green (G) component. As a result, a sense of incongruity such as a color difference may occur between the pixel to be processed after color correction and the surrounding pixels. When the correction coefficient increases the signal amount of the blue component signal of the pixel to be processed, the color correction using the correction coefficient by the color correction unit is avoided, thereby suppressing uncomfortable color correction. be able to.

  In another embodiment, the flaw noise judging means for judging whether or not the image represented by the image data received from the image input means contains flaw noise caused by flaws in the image sensor, and the flaw noise When it is determined by the determination that flaw noise due to flaws in the image sensor is included, flaw noise removing means for removing the flaw noise is further provided. Since the image data from which the flaw noise has been removed is used for the subsequent processing, it is possible to accurately detect pixels in which purple fringes may have occurred.

  In still another embodiment, a noise amount detection unit for detecting a noise amount of noise included in an image represented by the image data received from the image input unit, and a noise amount equal to or greater than a predetermined threshold is detected by the noise amount detection unit. A noise removing unit for removing the noise when detected is further provided. Since the image data from which the noise has been removed is used in the subsequent processing, it is possible to accurately detect pixels in which purple fringes may have occurred.

An image in which purple fringes are generated is shown. The graph of RGB signal amount is shown. The graph of RGB signal amount is shown. The graph of RGB signal amount is shown. It is a block diagram which shows the electric constitution of the digital still camera of 1st Example. It is a flowchart which shows the operation | movement procedure of the purple fringe correction apparatus of 1st Example. The positional relationship between the processed image and the three kernels is shown. The positional relationship between the kernel and the luminance calculation target pixel is shown. (A) shows the positional relationship between an image having a low luminance part and a high luminance part and three kernels, and (B), (C) and (D) show images of regions defined by the three kernels. (A) shows the positional relationship between an image having a low luminance part and a high luminance part and three kernels, and (B), (C) and (D) show images of regions defined by the three kernels. It is a block diagram which shows the electric constitution of the digital still camera of 2nd Example. It is a flowchart which shows the operation | movement procedure of the purple fringe correction apparatus of 2nd Example.

FIG. 1 shows an image in which purple fringes are generated. 2, 3 and 4 are graphs showing signal amounts (intensities) of RGB components for each pixel position corresponding to each of the _three line positions L _{1 to} L ₃ shown in the image 80 shown in FIG. is there.

  Purple fringing is a phenomenon in which the periphery of an image contour blurs purple when a high-luminance subject (for example, a light source) is imaged. Purple fringes are likely to occur in contour portions having a large luminance difference in the image. The purple fringe becomes more visible as the luminance difference in the contour portion increases.

  In the image 80 shown in FIG. 1, the white high-luminance portion HA and the black low-luminance portion LA are in contact. The boundary between the high luminance portion HA and the low luminance portion LA is an image contour (edge) E. Purple fringes are generated around the image contour E. A region (range) where purple fringes are generated in the image 80 is shown as a purple fringe generation region PF.

  In the image 80, the brightness of the high brightness portion HA decreases as it goes upward. The luminance of the low luminance part LA is constant. That is, the luminance difference between the high luminance portion HA and the low luminance portion LA in the vicinity of the lower end portion in the image 80 is very large, and the luminance difference between the high luminance portion HA and the low luminance portion LA is smaller in the vicinity of the upper end portion.

Referring to FIGS. 2 to 4, purple fringe is a signal of blue (B) component or blue (B) component and red (R) component when the RGB component is viewed as green (G) component. By having a signal amount greater than the amount, it represents purple. 2 to 4, double-ended arrows D _{1 to} D ₃ indicate the difference between the signal amount of the blue (B) component and the signal amount of the green (G) component at the pixel position in the vicinity of the image contour E. As the luminance difference between the high luminance portion HA and the low luminance portion LA is larger, the difference between the signal amount of the blue (B) component and the signal amount of the green (G) component at the pixel position near the image contour E is larger (D ₁ <D ₂ <D ₃ ). In proportion to this signal amount difference, the purple fringe generation region PF also increases and becomes conspicuous.

  The digital still camera according to the embodiment of the present invention can perform image processing for reducing purple fringes on an image having the purple fringe generation area PD described above. Hereinafter, a digital still camera having a purple fringe reduction function will be described.

  FIG. 5 is a block diagram schematically showing the electrical configuration of the digital still camera.

  The overall operation of the digital still camera is controlled by the CPU 2.

  The CPU 2 is connected via the data bus 1 to the imaging unit 10, purple fringe correction device 20, external memory interface (external memory I / F) 5, memory interface (memory I / F) 7, and compression / decompression processing circuit 9. It is connected to the.

  The imaging unit 10 includes a CCD 12, and an optical unit 11 including an imaging lens, a diaphragm, an infrared cut filter, an optical low pass filter (OLPF), and the like is provided in front of the CCD 12.

  The digital still camera includes a flash 4 for flash imaging and a charging circuit 3 for supplying power to the flash 4. The CPU 2 issues a light emission instruction for the flash 4 and a charging instruction for the charging circuit 3.

  When the power of the digital still camera is turned on and the photographing mode is set, a light beam representing the subject image is incident on the optical unit 11. The light beam is incident on the light receiving surface of the CCD 12 via the optical unit 11. The CCD 12 has a number of photodiodes (photoelectric conversion elements) arranged two-dimensionally on the light receiving surface, and a red array arranged on the light receiving surface with a predetermined arrangement structure (Bayer arrangement, G stripe arrangement, etc.). Color filters of (R), green (G), and blue (B) are provided. A subject image formed by the optical unit 11 is electronically captured by the CCD 12. The CCD 12 is driven by an image sensor driving circuit 17 that outputs a timing signal and the like in response to a command from the CPU 2.

  An analog signal representing the subject image output from the CCD 12 is input to the analog signal processing device 13.

  The analog signal processing device 13 includes a correlated double sampling circuit, a signal amplifier, and the like. An analog signal representing the subject image output from the CCD 12 is input to the analog signal processing device 13, where correlated double sampling, signal amplification, and the like are performed. An analog video signal (analog RGB signal) output from the analog signal processing device 13 is input to an analog / digital conversion circuit (ADC) 14 and converted into digital image data (digital RGB data) after predetermined signal processing. . Further, the RGB data is converted into image data (YUV data) including luminance data (Y data) and color difference data (Cr, Cb data) as necessary.

  The digital signal data output from the ADC 14 is subjected to white balance adjustment, gamma correction, and synchronization processing in the digital signal processing unit 15 (interpolation of the spatial shift of the color signal associated with the color filter array of the single CCD). Predetermined digital signal processing, such as a process for simultaneously converting

  Digital image data is temporarily recorded in the RAM 8A under the control of the memory I / F 7. A ROM 8B is also connected to the memory I / F 7. The ROM 8B stores a control program executed by the CPU 2, various data necessary for control, photographer setting information, and various setting information related to the operation of the digital still camera.

  A subject image represented by digital image data read from the RAM 8A is displayed on a display screen of a display device (not shown).

  When a shutter release button (not shown) is pressed in the first stage, the lens of the optical unit 11 is driven by the motor drive circuit 16 to perform focusing. Luminance data is obtained in the digital signal processor 15 based on the image data read from the RAM 8A. Data representing the integrated value of the luminance data is given to the CPU 2 to calculate the exposure amount. The aperture of the diaphragm of the optical unit 11 is controlled by the motor drive circuit 16 so that the calculated exposure amount is obtained, and the shutter speed of the CCD 12 is controlled by the image sensor drive circuit 17.

  When the shutter release button is pressed in the second stage, the digital image data is stored in the RAM 8A. Predetermined digital signal processing is performed on the image data read from the RAM 8A in the digital signal processing device 15, and thereafter the data is compressed in the compression / decompression processing circuit 9. The compressed image data is recorded on the memory card 6 under the control of the external memory I / F 5.

  When the reproduction mode is set, the compressed image data recorded on the memory card 6 is read. The read compressed image data is expanded by the compression / expansion processing circuit 9 and then given to the display device to display a reproduced image.

  The purple fringe correction device 20 includes an image input device (external device connection interface (USB port, etc.)) 21, a luminance difference calculation circuit 22, a maximum luminance difference selection circuit 23, a hue determination circuit 24, a correction determination circuit 25, and a purple fringe correction circuit. 26 and a process end determination circuit 27. A purple fringe correction device 20 is used for image data input to the digital still camera from the image input device 21, image data obtained by imaging with the CCD 12, and image data read from the memory card 6. Image processing for reducing the above-described purple fringe is performed.

  FIG. 6 is a flowchart showing a procedure of purple fringe reduction processing performed in the purple fringe correction apparatus 20.

  Image data to be processed is input (step 51). As described above, the image data to be processed may be image data obtained by being imaged by the CCD 12, image data read from the memory card 6, or through the image input device 21. It may be image data captured by a digital still camera. The input image data is temporarily stored in the RAM 8A.

  The luminance difference calculation circuit 22 uses three kernels KA, KB, and KC having different sizes, and the processing target pixels constituting the image represented by the image data have four directions (vertical direction, horizontal direction, diagonal right direction). And the process of calculating the luminance difference ΔY in the left diagonal direction is performed in parallel (simultaneously) (steps 52, 53, 54).

  FIG. 7 shows the relationship between the image 40 to be processed and the three kernels KA, KB, and KC having different sizes.

  The kernels KA, KB, and KC all specify the distance between the central pixel and eight pixels (luminance difference calculation target pixels) that are located in the vicinity of the central pixel and that calculate the luminance difference ΔY. It has a square shape. The distance between the center pixel and the luminance difference calculation target pixel is defined by the size of the kernel. The size of the kernel KA is the smallest, and the size of the kernel KB is larger than the size of the kernel KA. The size of the kernel KC is larger than the size of the kernel KB. A pixel at a position farther from the center pixel than the luminance difference calculation target pixel determined by the kernel KA is set as a luminance difference calculation target pixel by the kernel KB. A pixel at a position farther from the center pixel than the luminance difference calculation target pixel determined by the kernel KB is set as a luminance difference calculation target pixel by the kernel KC.

  The three kernels KA, KB, and KC have a positional relationship in which the largest kernel KC includes the second largest kernel KB, and the second largest kernel KB includes the smallest kernel KC. Have. The three kernels KA, KB, and KC are scanned on the target image 40 while maintaining this positional relationship.

  FIG. 8 shows the positional relationship among the kernel KA, the central pixel (processing target pixel) P1, and the eight luminance difference calculation target pixels P2 to P9 located around the central pixel P1.

  Referring to FIG. 8, when the center pixel P1 is determined, the four directions of the center pixel P1 in the left-right direction (H direction), the up-down direction (V direction), the right diagonal direction (NE direction), and the left diagonal direction (NW). In each of these, eight luminance difference calculation target pixels determined according to the size of the kernel KA are determined. When the kernel KA has a size of 4 vertical pixels × 4 horizontal pixels, eight pixels P2 to P9, which are separated from the central pixel P1 by 4 pixels on the left, right, top, bottom, right diagonal and left diagonal, are luminance difference calculation target pixels. become.

  In the calculation process of the luminance difference ΔY in four directions (vertical direction, horizontal direction, diagonal right direction and diagonal left direction), the horizontal direction (H direction), vertical direction (V direction), diagonal right direction (NE direction) and diagonal left For each direction (NW), a difference is taken between two pixels located symmetrically with respect to the center pixel P1. For example, in the left-right direction (H direction), the luminance difference between the pixel P2 at the position separated by 4 pixels in the left direction from the center pixel P1 and the luminance of the pixel P3 separated by 4 pixels in the right direction from the center pixel P1. Is calculated. Similarly, the luminance difference of the pixel P4 at a position separated by four pixels upward from the center pixel P1 and the luminance of the pixel P5 separated by four pixels downward from the center pixel P2 is calculated (up and down direction). The difference between the luminance of the pixel P6 at a position separated by four pixels in the diagonally upper left direction from the central pixel P1 and the luminance of the pixel P7 separated by four pixels in the diagonally downward right direction from the central pixel P2 is calculated (diagonal in the left diagonal direction). ), The difference between the luminance of the pixel P8 at a position separated by four pixels in the diagonally upper right direction from the center pixel P1 and the luminance of the pixel P9 separated by four pixels in the diagonally lower left direction from the central pixel P2 is calculated ( Diagonally to the right). Four luminance differences corresponding to each of the four directions are calculated for one kernel.

  Referring to FIG. 7, as described above, the sizes of kernels KA, KB, and KB are the smallest in the size of kernel KA, kernel KB is larger than the size of kernel KA, and the size of kernel KC is the size of kernel KB. Bigger than. The kernel KC includes the kernel KB, and the kernel KB further has a positional relationship including the kernel KA. Therefore, as described above, as the eight luminance calculation target pixels determined by the kernel KB, pixels that are located farther from the central pixel P1 than the luminance calculation target pixels (FIG. 8) determined by the kernel KA are used. As the eight luminance calculation target pixels determined by the kernel KC, pixels located at positions farther from the central pixel P1 than the luminance calculation target pixels determined by the kernel KB are used.

  The positions of the central pixels of the kernels KA, KB, and KC are shifted by one pixel in the row direction and the column direction, and the three kernels KA, KB, and KC are scanned on the target image 40. Each pixel constituting the processed image 40 sequentially becomes the central pixel P1 (that is, the processing target pixel). A luminance difference calculation process is performed for each pixel to be processed. A total of 12 luminance differences ΔY are calculated for one processing target pixel. The luminance difference calculation process is performed on all the pixels of the processed image 40.

  Returning to FIG. 6, when twelve luminance differences ΔY are calculated for the pixel to be processed, the maximum luminance difference selection circuit 23 then calculates the maximum one of the twelve luminance differences ΔY calculated (maximum luminance difference ΔYmax). ) Is selected (step 55). It is determined whether or not the selected maximum luminance difference ΔYmax is greater than a predetermined threshold (step 56).

  As described above, the purple fringe occurs around the image contour having a large luminance difference. Therefore, if the selected maximum luminance difference ΔY is low, it is unlikely that purple fringes have occurred in the processing target pixel. If the maximum luminance difference ΔY is larger than the predetermined threshold value, the process proceeds to the subsequent processing (YES in step 56). If the maximum luminance difference ΔY is equal to or smaller than the predetermined threshold value, the subsequent processing for the processing target pixel is skipped, The process proceeds to the process for the next pixel (NO in step 56).

  FIG. 9A shows the relationship between an image 42 including an image contour having a luminance difference and three kernels KA, KB, and KC. In FIG. 9A, a hatched portion (range) indicates a low luminance image portion, and a non-hatched portion indicates a high luminance image portion. FIGS. 9B, 9C, and 9D respectively show image portions in a range surrounded by the kernels KA, KB, and KC in FIG. 9A.

  With reference to FIGS. 9A, 9B, and 9C, if all of the pixels surrounded by the kernel KA are pixels of the low-luminance image portion LA1, the calculation is performed for all of the four directions described above. All the luminance differences are small, and therefore the maximum luminance difference is also small. The same applies to the kernel KB. On the other hand, with reference to FIGS. 9A and 9D, in the range defined by the kernel KC, the pixel of the high luminance image portion HA1 is located at one end and the low luminance image portion is located at the other end. The pixel of LA1 is located. A large luminance difference appears among the four luminance differences calculated using the kernel KC (in the case of FIG. 9D, the luminance difference between the pixels P2 and P3 positioned in the horizontal direction becomes a large value). In the kernel KA and the kernel KB, even if the subsequent processing is skipped, the pixel is processed by the kernel KC having a large size.

  FIG. 10A also shows an image 42 including an image contour having a luminance difference, and three kernels KA, KB, and KC. The positions of the kernels KA, KB, and KC are different from those in FIG. FIGS. 10B, 10C, and 10D show image portions in a range defined by the kernels KA, KB, and KC in FIG.

  Referring to FIGS. 10A and 10B, the high-luminance image portion HA2 is located on one side and the low-luminance image portion LA2 is located on the other side in the range defined by the kernel KA. A large luminance difference is calculated in the luminance difference calculation process using the kernel A (in the case of FIG. 10A, the luminance difference between the pixels P2 and P3 located in the horizontal direction becomes a large value). On the other hand, referring to FIGS. 10A, 10C, and 10D, the range defined by the kernel KB and the range defined by the kernel KC are both high-luminance image portion HA1 and low-luminance image. Although the portion LA3 is included, the high luminance image portion is located on one side and the low luminance image portion is not located on the other side in any of the four directions described above. In the luminance calculation processing using the kernels KB and KC, a large luminance difference is not calculated. By using the kernel KB and the kernel KC, even if the subsequent processing is skipped, the pixel is processed by the kernel KA having a small size.

  Returning to FIG. 6, when the maximum luminance difference ΔYmax is larger than the predetermined threshold value (YES in step 56), it is determined whether purple fringing has occurred in the processing target pixel. In this embodiment, the hue determination circuit 24 calculates the hue of the pixel to be processed and determines whether it is purple or a hue close to purple (step 57). If it is determined that the color is not purple, the subsequent process is skipped and the process proceeds to the next pixel (NO in step 57). This is because purple fringe has one of the characteristics of appearing in purple.

  When it is determined that the hue is purple, a correction coefficient for reducing purple fringe, for example, a correction coefficient for the blue (B) component, or the blue (B) component and the red (R) component is calculated by the correction determination circuit 25. (YES in step 57, step 58). The calculated correction coefficient may be a coefficient for matching the color of the pixel to be processed with the color of the pixel located at a predetermined number of pixels (for example, 8 pixels) away from the pixel to be processed. It may be a low coefficient. In any case, a correction coefficient so that the color of the pixel to be processed is different from purple, a correction coefficient that makes the purple of the pixel to be processed inconspicuous, and the like are calculated.

  The correction determination circuit 25 further determines whether or not the calculated correction coefficient is one that reduces the signal amount (intensity) of the blue (B) component (one that corrects the signal amount in the negative direction) (step 59). That is, as described above, the purple fringe has a characteristic that the signal amount of the blue (B) component is far away in the (plus) direction than the signal amount of the green (G) component. In order to achieve this, it is necessary to reduce the signal amount of the blue (B) component (correct in the minus direction). When the calculated blue (B) component correction coefficient increases the signal amount of the blue (B) component of the pixel to be processed, color correction using the correction coefficient may cause a sense of incongruity in the image contour. There is a risk of being corrected to color.

  If the calculated correction coefficient does not reduce (increase) the signal amount (intensity) of the blue (B) component, the subsequent processing is skipped and the processing proceeds to the next pixel (NO in step 59). ).

  If the calculated correction coefficient reduces the signal amount (intensity) of the blue (B) component, the purple fringe correction circuit 26 uses the calculated correction coefficient to correct the color of the processing target pixel. That is, purple fringe correction processing is performed on the processing target pixel (step 60).

  The above-described process is repeated for all pixels of the target image (NO in step 61). When the processing for all the pixels is completed, the processing end determination circuit 27 determines the end of the processing, and the corrected image data is output from the RAM 8A and recorded in an external device or the like connected to the memory card 6 and the image input device 21. (Step 62).

  As described above, by using the three kernels KA, KB, and KC having different sizes, pixels in the vicinity of the image contour having a large luminance difference can be targeted for subsequent processing with almost no omission. It is possible to reduce (reduce) purple fringes by accurately detecting pixels in which purple fringes may occur.

  In the above-described embodiment, the purple fringe reduction processing is executed by the hardware device (purple fringe correction device 20). However, the CPU 2 or the digital signal processing device 15 may execute the purple fringe reduction processing. In this case, a program for causing the CPU 2 or the digital signal processing device 15 to perform purple fringe reduction processing is stored in the ROM 8B. The same applies to other embodiments described later.

  FIG. 11 is a block diagram showing the electrical configuration of a digital still camera according to another embodiment. 5 differs from that shown in FIG. 5 in that the purple fringe correction device 20A further includes a CCD flaw detection circuit 31, a CCD flaw correction circuit 32, a noise amount detection circuit 33, and a noise reduction processing circuit. FIG. 12 is a flowchart showing a procedure of purple fringe reduction processing by the digital still camera of another embodiment shown in FIG. The same processes as those in the flowchart shown in FIG.

  The CCD flaw detection circuit 31 detects image features that appear due to CCD flaws on the image represented by the image data to be processed (step 71). For example, if the CCD 12 is scratched, a white point may occur in the image represented by the image data output from the CCD 12. When this white point is detected, CCD flaw noise correction processing such as matching the white point with the color of the surrounding pixels is performed by the CCD flaw correction circuit 32 (step 72).

  Further, the noise amount detection circuit 33 detects the amount of noise included in the image data. For example, if the difference between the pixel value of the pixel to be processed and the median value in the vicinity thereof is greater than or equal to a predetermined threshold, it is determined that noise exists, and if it is smaller than the predetermined threshold, it is determined that there is no noise. A noise amount is obtained according to the magnitude exceeding the threshold (step 73).

  If the amount of noise is less than the predetermined threshold, the process proceeds to the purple fringe reduction process described above (NO in step 74). When it is determined that the noise amount is equal to or greater than the predetermined threshold value, noise reduction processing using a median filter or the like is performed by the noise reduction processing circuit 34, and the image data subjected to noise reduction processing is subjected to purple fringe reduction processing (in step 74). YES, step 75).

  Since image data in which noise due to scratches on the CCD 12 and other noises is reduced is subjected to purple fringe reduction processing, it is possible to increase the detection accuracy of pixels in which purple fringes may occur.

20,20A Purple fringe correction device
21 Image input device
22 Luminance difference calculation circuit
23 Maximum brightness difference selection circuit
24 Hue judgment circuit
25 Correction judgment circuit
26 Purple fringe correction circuit
31 CCD scratch detection circuit
32 CCD scratch correction circuit
33 Noise detection circuit
34 Noise reduction processing circuit

Claims (8)

Image input means for accepting input of given image data;
A plurality of kernels having different sizes that define the distance between the processing target pixel and the luminance calculation target pixel located around the processing target pixel according to the size, while maintaining the positional relationship including one of the other. By scanning the processing target image represented by the image data received by the input means, the processing target pixel for each of the processing target pixel and for each of the kernels in the horizontal direction, the vertical direction, and the two diagonal directions. A luminance difference calculating means for calculating a luminance difference of a luminance calculation target pixel at a symmetrical position with respect to the center, and comparing the largest luminance difference among a plurality of luminance differences calculated by the luminance difference calculating means with a predetermined threshold value A comparison means,
When the comparison means determines that the maximum luminance difference is greater than the threshold value, the process proceeds to a determination as to whether the processing target pixel is a pixel that has generated purple fringes.
Image processing device. Whether or not the processing target pixel is a pixel that has generated purple fringes is determined by whether or not the color of the processing target pixel is purple.
The image processing apparatus according to claim 1. When it is determined that the color of the processing target pixel is purple, a correction coefficient calculation unit that calculates a correction coefficient that changes the color or saturation of the processing target pixel, and a correction coefficient calculated by the correction coefficient calculation unit Using color correction means for correcting the color of the pixel to be processed using
The image processing apparatus according to claim 2. A color correction avoiding means for avoiding color correction using the correction coefficient by the color correcting means when the correction coefficient increases the signal amount of the blue component signal of the processing target pixel;
The image processing apparatus according to claim 3. Scratch noise determining means for determining whether or not the image represented by the image data received from the image input means includes scratch noise due to scratches in the image sensor; and A scratch noise removing means for removing the scratch noise when it is determined that the scratch noise is caused by
The image processing apparatus according to claim 1. A noise amount detecting means for detecting a noise amount of noise included in an image represented by image data received from the image input means, and a noise amount equal to or greater than a predetermined threshold value detected by the noise amount detecting means; A noise removing means for removing the noise so that at least the amount of noise is less than the threshold value;
The image processing apparatus according to claim 1. Image input means accepts input of given image data,
The luminance difference calculating means determines the distance between the processing target pixel and the luminance calculation target pixel located around the processing target pixel according to the size, and includes a plurality of kernels having different sizes, one of which includes the other By scanning the processing target image represented by the image data received by the image input means while maintaining the horizontal direction, the vertical direction, and the two diagonal directions for each processing target pixel and for each kernel. , The luminance difference of the luminance calculation target pixel at the symmetrical position with respect to the processing target pixel is calculated,
A comparing means for comparing the largest brightness difference among the plurality of brightness differences calculated by the brightness difference calculating means with a predetermined threshold;
When the comparison means determines that the maximum luminance difference is greater than the threshold value, the process proceeds to a determination as to whether the processing target pixel is a pixel that has generated purple fringes.
Image processing method. Computer,
Image input means for accepting input of given image data;
A plurality of kernels having different sizes that define the distance between the processing target pixel and the luminance calculation target pixel located around the processing target pixel according to the size, while maintaining the positional relationship including one of the other. By scanning the processing target image represented by the image data received by the input means, the processing target pixel for each of the processing target pixel and for each of the kernels in the horizontal direction, the vertical direction, and the two diagonal directions. A luminance difference calculating means for calculating a luminance difference of a luminance calculation target pixel at a symmetrical position with respect to the center, and comparing the largest luminance difference among a plurality of luminance differences calculated by the luminance difference calculating means with a predetermined threshold value Function as a means of comparison,
If the maximum brightness difference is determined to be greater than the threshold, the computer is controlled to proceed to determine whether the processing target pixel is a pixel that has generated purple fringes;
program.