JP4659354B2 - Image processing method and apparatus - Google Patents

Description

  The present invention relates to an image processing method and apparatus, and more particularly to an image processing method and apparatus for correcting output image data to obtain output image data.

  As still image input devices, digital cameras, digital videos, PDAs, mobile phones, etc. are diversified, but JPEG is the mainstream as the image storage file format. When observing or printing a captured still image, it has become common to use the captured image by correcting it. However, the image of interest is defined in the image and the image correction is centered on that part. This is called high-level correction.

  When performing exposure amount correction when printing a negative image on photographic paper, the most noticeable part when viewing as a main part in an image is generally a part corresponding to a person. Extracting and setting the weighting factor for each area so that the weight of the area corresponding to the extracted person is larger than the weight of the other areas so as to improve the image quality of the area corresponding to the main part in the image A technique for correcting the exposure amount is proposed (for example, see Patent Document 1).

Japanese Patent Application Laid-Open No. 9-304846

However, there are the following three problems in using the above-described conventional method.
(1) Since the influence of the main part in the entire image is not considered in the weight setting in the main part, for example, even when the area occupied by the main part in the entire image is very small, the main part is mainly As a result, the quality of the entire image may be deteriorated.
(2) Since the prior art treats the detection probability of the target image as 100% certainty as the premise, there is a risk of performing correction with an increased weight for an object that is not the target of attention if the detection is wrong. There's a problem.
(3) The target detected as the target image may not necessarily require weighting for this area. For example, there is a case where the target image is not limited to the main part, such as detection of a blue sky region in the image as the target image.

  The present invention aims to solve the above-mentioned problems, and the invention according to claim 1 is a step of extracting a target image included in input image data, and a standard correction calculated based on a feature amount of the target image. A correction step of correcting input image data using values to generate output image data, analyzing a region of the target image, and ratio of the region of the target image to the entire region of the input image data A step of acquiring feature information including an area ratio, a step of calculating a success rate of detecting a target image from chromaticity information or DCT information included in the feature information, and a calculation using the area ratio and the success rate. A correction step of correcting the standard correction value obtained to obtain an actual correction value, and a step of calculating an average luminance in the target image from luminance information in the region of the target image The step corrects the input image data using the actual correction value, and the correction step increases the correction ratio of the standard correction value as the area ratio and the success rate are larger. Based on the average brightness and the average brightness of the input image data, brightness correction is performed on the input image data.

  According to a second aspect of the present invention, in the image processing method according to the first aspect, the target image is a blue sky.

  According to a third aspect of the present invention, in the image processing method according to the first aspect, the target image is human skin.

  According to a fourth aspect of the present invention, there is provided an output unit that corrects input image data using extraction means for extracting a target image included in input image data and a standard correction value calculated based on a feature amount of the target image. Analyzing means for analyzing featured image area and obtaining feature information including area ratio, which is ratio of noticed image area to total area of input image data, in image processing apparatus including correction means for generating data And correction means for calculating the success rate of detecting the target image from the chromaticity information or DCT information included in the feature information, and correcting the standard correction value calculated using the area ratio and the success rate, and actually correcting Correction means for acquiring a value, and means for calculating the average luminance in the target image from the luminance information in the region of the target image, the correction means corrects the input image data using the actual correction value, and the correction means The correction ratio of the standard correction value is increased as the area ratio and the success rate are larger, and the correction means adds the correction ratio to the input image data based on the area ratio, the average luminance in the target image, and the average luminance of the input image data. On the other hand, luminance correction is performed.

  According to the present invention, the step of analyzing the region of the target image and obtaining feature information including the area ratio, which is the ratio of the region of the target image region in the input image data to the entire image region, and calculating using the area ratio A correction step of correcting the standard correction value obtained to obtain an actual correction value, and the correction step corrects the input image data using the actual correction value, so that the target image in the image is detected and corrected. In this case, the correction effect can be enhanced by executing an optimal correction with respect to the entire image centering on the target image.

(First embodiment)
Hereinafter, an image processing method according to the present invention will be described with reference to the drawings.
The present invention is performed in order to realize a higher quality image correction based on the detection result obtained by the image detection technique. FIG. 1 shows a flowchart for explaining the entire processing when the present invention is implemented.

In S101, information about printing is acquired together with image acquisition. In addition, initial settings including the presence or absence of various operations are performed from the acquired data. In this step, it is determined which type of attention image is to be processed, and the information is set for the following processing. In the following step, the type of attention image set by referring to the information is set. Each process is performed. In S102, although it may or may not be performed depending on the set type of the target image, the target image in the image is detected, and the feature amount for the detected target image is calculated. Further, it is analyzed whether or not the input image data has good image quality. Here, a good image generally has the following characteristics, and an image satisfying this condition is also defined as a good image in this specification.
(1) The white balance of the light source is good (2) The contrast is good (3) The gradation of the necessary part is secured

  The input image is analyzed to determine whether the image satisfies the above conditions. If the image is not satisfied, various parameters for correction items necessary for image correction in subsequent steps are calculated and output.

  In S103, an image correction value that is a standard correction value is determined based on the various parameters obtained in S102. In S104, image correction is performed using a correction parameter determined based on the relationship between the input image data and the detected target image, that is, the actual correction value obtained by correcting the image correction value determined in S103 based on the feature information. Do.

  In S101, attention image detection is performed among initial settings performed in addition to setting information acquisition. Based on the relationship with the acquired setting information, detection is performed for each type of attention image as shown in FIG. Set whether or not to target the operation. That is, the flow of FIG. 2 described next is processing included in S101.

  In step S201, it is determined whether or not to perform processing by detecting human skin as an attention image in the acquired image. When not performing, it progresses to the selection with respect to the next attention image of S203. When processing is performed, a flag for human skin detection processing permission is set in S202, and then the process proceeds to selection for the next target image in S203.

  In step S203, it is determined whether or not to perform processing by detecting a blue sky as a noticed image in the acquired image. When not performing, it progresses to the selection with respect to the next attention image of S205. When performing, in S204, a blue sky detection process permission flag is set, and then the process proceeds to selection for the next target image in S205.

  In step S205, it is determined whether or not a night scene is detected in the target image in the acquired image and processing is performed. If not, the determination target setting process is terminated. Also, when performing, in S206, a night scene determination detection process permission flag is set, and then the determination target setting process is terminated.

Next, the attention image detection / analysis process performed in step S102 will be described.
FIG. 3 shows a flow when human skin is detected as an attention image. In S301, it is confirmed whether or not the human skin detection process permission flag is set in S202, that is, whether or not it is ON. If it is OFF, the process proceeds to the next process without performing the human skin detection process. If it is ON, the process proceeds to S302.

  In S302, the target image detection process according to the background art described above is executed. When the detection process ends, the process proceeds to S303. In S303, it is determined whether or not a final candidate has been detected as a result of human skin detection. If not detected, the process proceeds to S304, an undetected flag is set, this process is terminated, and the process proceeds to the next process. When the final candidate is detected, in step S305, the feature amount of the image area in which the final candidate is detected, that is, as feature information, (1) area information, (2) luminance information, (3) chromaticity information, (4) DCT information is calculated and stored.

  Here, the DCT information is an average of DCT transform values of the 8 × 8 pixel block included in the region where the target image is detected. That is, in data compression of a Jpeg file, discrete cosine transformation (DCT transformation) is applied to data in units of 8 × 8 pixel blocks, and DCT transformation values obtained by this transformation are averaged to calculate DCT information. .

FIG. 4 is a flow when a blue sky is detected as an attention image. In S401, it is confirmed whether or not the blue sky detection process permission flag is set in S204. If it is OFF, this process is terminated and the process proceeds to the next process. If it is ON, the process proceeds to S402. In S402, the above detection process is executed, and when the detection process is completed, the process proceeds to S403.
In S403, it is determined whether or not a final candidate has been detected as a blue sky detection processing result. If not detected, the process proceeds to S404, an undetected flag is set, the process is terminated, and the process proceeds to the next process. When the final candidate is detected, in step S405, (1) area information, (2) luminance information, (3) chromaticity information, and (4) DCT information are calculated as the feature amounts of the image region in which the final candidate is detected. Remember.

  FIG. 5 is a flow for determining whether or not the image of interest is a night scene. In S501, it is confirmed whether or not a detection process permission flag is set in S206, and if it is OFF, the detection process is terminated in the night scene and the process proceeds to the next process. If it is ON, the process proceeds to S502.

  In S502, night scene determination processing is executed. When the determination process ends, the process proceeds to S503. In S503, it is determined whether or not a final candidate has been detected as a night scene determination process result. When it is not detected, the process proceeds to S504, and an undetected flag is set and the process ends. If it is determined that the scene is a night scene, in step S505, (1) area information, (2) luminance information, (3) chromaticity information, and (4) DCT information are used as the feature amounts of the image area in which the final candidate is detected. Is calculated and stored.

Details of the subsequent processing when the blue sky region is set as the attention image will be described.
FIG. 6 is an image sample taken as a landscape image. A blue sky region exists in this image. The result of performing the blue sky region detection shown in FIG. 4 on the above image sample is the image of FIG. In 8 × 8 pixel block units, blocks that match a preset chromaticity and luminance range are determined, candidate areas are formed based on a set of matched adjacent blocks, and DCT characteristics in the areas are calculated. It is determined whether or not the definition characteristics of the blue sky are met.

  The (1) area information, (2) luminance information, (3) chromaticity information, and (4) DCT information calculated during the determination process are stored as feature amounts for correcting the standard correction value.

  FIG. 8 is a statistical result of determining whether or not an image detected based on the result of the detection of the blue sky region in the image described above for a plurality of images correctly detects the target blue sky region. is there. That is, the table shows the relationship between the area information acquired in step S405 as a detection result of the above-described blue sky region and the probability of successful detection. The horizontal axis represents the area ratio of the detected blue sky region in the entire image. The vertical axis represents the ratio between the case where an area other than the blue sky is erroneously detected and the area correctly detected within each range of the detected area ratio.

  It can be read from the graph shown in FIG. 8 that the probability of correctly recognizing when the area ratio of the detected blue sky region is less than 1.0% is approximately 80%. In the case of an area of 1% or more and less than 5%, the possibility of correct recognition is 90% or more, and in the case of a detection area of 5% or more, the detection success rate is 100%.

  As described above, in the detection system in which the detected area ratio and the detection success rate have a certain degree of correlation, even if an attention image is detected, if it is a false detection, the erroneously detected attention image is displayed. If optimized image correction is performed, the image may be deteriorated.

  In general, when a blue sky is detected after correction processing, the image is determined to be highly likely to be an outdoor landscape image, and correction is performed using image correction constants with higher contrast and saturation than normal image processing. Do. However, in the detection system as described above, in consideration of the possibility of erroneous detection, a correction ratio indicating how much to be amplified or reduced from the normal correction is introduced to enable correction in accordance with an actual image. The correction value is corrected as follows. In this embodiment, when the standard correction value is 100% and the area ratio is less than 1%, the amplification amount as the correction ratio is set to 80% in consideration of the detection result of the area ratio in FIG. The correction strength is weakened, and in the region where the detection probability of 1% or more is high, the correction ratio is set to 1 so that the original standard correction value is obtained, and 100% standard correction is applied. Note that the detection success rate and the percentage of the correction amount amplified do not have to match.

  FIG. 26 is a block diagram illustrating a configuration example of the image processing apparatus according to the present embodiment. The image processing apparatus according to this embodiment includes a decoding unit 2610, an image recognition unit (executing first image extraction) 2630 that recognizes an image area to be corrected based on data acquired from the decoding unit 2610, and an image recognition unit 2630. And a tone correction unit 2620 that corrects the recognition area to a desired color. The reproduced and corrected image (BMP) output from the color tone correction unit 2620 is sent to the printer and printed.

  The image recognition unit 2630 receives the decoded image (BMP) from the decoding unit 2610 and detects the target color detection unit 2631 that detects the region of the specified target color (skin color in this example), and the decoded DCT data from the decoding unit 2610. Based on the spatial frequency from the candidate region of the target color detected by the target color detection unit 2631 and the spatial frequency generation unit 2632 that receives and generates a spatial frequency in the candidate region of the target color detected by the target color detection unit 2631 Thus, a target color area selection unit 2633 for selecting a target area for color tone correction is provided. The target color detection unit 2631 includes a decoded image storage unit 2631a that stores the decoded image. However, the decoded image storage unit 2631a does not need to be in the target color detection unit 2631 and may also be used as another processing unit. The target color area selection unit 2633 has a determination table 2633a for selection. The discrimination table 2633 may include a plurality of discrimination tables corresponding to the image size.

  In order to further improve the processing of the present embodiment, the image recognition unit 2630 receives the quantization table value from the decoding unit 2610 and performs color tone correction that prohibits color tone correction processing from determination based on the threshold value 2634a for prohibition. A prohibition unit 2634 is included.

  The color tone correction unit 2620 performs a known color correction process using the color of the selection area selected by the image recognition unit 2630 as a correction target color (skin color in this example) using, for example, the color correction table 2620a. This color tone correction process is prohibited by a color tone correction prohibition signal from the target color area selection unit 2633 or the color tone correction prohibition unit 2634 under a predetermined condition. This correction processing may be performed on the entire image for simplification, but may be correction different for each region or partial correction if the purpose is to improve image quality. Since the feature of the present invention is not in such a color tone correction method, it will be described briefly in this embodiment.

  FIG. 27 is a diagram illustrating a configuration example of hardware and software for realizing image processing according to the present embodiment. FIG. 27 mainly illustrates the image recognition unit 2630 which is a characteristic part of the present embodiment. This apparatus can be realized by a general-purpose computer or a dedicated computer.

  The CPU 2710 performs arithmetic processing, and the ROM 2720 stores fixed data and programs used by the CPU 2710 (OS and BIOS etc. are here). The RAM 2730 temporarily stores data and programs used by the CPU 2710 in this embodiment. Here, in the present embodiment, the application program is loaded from a later-described external storage unit 2740 into the program load area 132 of the RAM 2730 and executed by the CPU 2710.

The data stored in the data storage area 2731 by the RAM 2730 stores the decoded image data area 2731 a for storing the decoded image decoded by the decoding unit 2610 or the reproduction image whose color tone has been corrected, and the correction target color (skin color in this example) data. Correction target color area 2731b, candidate area storage area 2731c for storing the detected target color area, candidate group area 2731d for storing candidate groups formed from the candidate areas, and a selection area for storing finally selected areas Storage area 2731e, decoded DCT data storage area 2731f for storing the decoded DCT data from the decoding unit 2610, spatial frequency area 2731g for storing the generated spatial frequency, and a discrimination table used for selecting the target color area Discriminating table area 2731h from the decoding unit 2610 Quantization table area 2731i for storing a child table, a quantization coefficient addition value storage area 2731j for storing a value obtained by adding the coefficients of the quantization table, and a threshold value group used for prohibiting color correction, etc. 2731k is included.

  The external storage unit 2740 includes a large capacity or removable medium such as a disk or a memory card, and includes a floppy (registered trademark) disk, a CD, and the like. In the data storage area 2741 of the external storage unit 2740, discrimination tables 1 to n2741a and a threshold value group 2741b are stored. In addition, a database for storing other parameters, image data, and the like may be stored. The program storage area 2742 is roughly classified into a target color area detection module 2741c, a spatial frequency generation module 2741d, a target color area selection module 2741e, a color tone correction prohibition module 2741f, and features executed in the second embodiment to be described later. The part extraction module 2741g is stored.

  Furthermore, the apparatus shown in FIG. 27 may also serve as the decoding unit 2610 and / or the color correction unit 2620. In this case, the color correction table 2741f as data, the color correction module 2741i as a program, The blur correction module 2741j used in the embodiment may be stored.

  In this example, the input interface 2750 inputs the decoded data (BMP) from the decoding unit 2610, the decoded DCT data, the quantization table value, and target color data that is unique to the apparatus or can be designated from the outside. The output interface 2760 outputs a selection area and a color correction prohibition signal. If the apparatus also serves as a color tone correction unit, the output is color tone corrected image data (BMP). Further, the present apparatus may also serve as the decoding unit 2610. In this case, JPEG data is input and color tone corrected image data (BMP) is output. In that case, further data and programs are prepared.

(Second Embodiment)
The second embodiment has the same configuration as the first embodiment described above, and the object of the target image is also a blue sky. However, the distance from the threshold of the chromaticity ratio of the detection determination in the process of detection, and the DCT characteristic The success rate of detection in the image of interest from the distance from the threshold is calculated, and the correction amount is corrected using this value.

  The purpose of the target image detection is to correct the target image or the scene accompanying it to be an optimal image when the target image defined in the image is searched and detected. Therefore, the effect after correction differs depending on the area of the detected target image in the entire image. In other words, if the ratio of the target image to the entire image is large, the influence when the target image is erroneously determined is large, that is, the image quality deteriorates.

  In view of this, FIG. 9 shows specific results when determining the effect ratio of the correction constant of the image correction, that is, the correction ratio, based on the detection success rate of the target image and the area ratio of the target image in the entire image. The horizontal axis represents the occupied area ratio in the entire image of the target image detected in the target image detection. The vertical axis represents the amplification factor which is a correction ratio when the maximum amount amplified from the normal image correction constant by the target image detection is 100%. The polygonal line data is for determining the amplification factor of the corrected image based on the detected target image area ratio. That is, as described above, it is determined which polygonal line to use based on the detection success rate calculated based on the feature information, and after the polygonal line is determined, the amplification factor corresponding to the area ratio is determined based on the graph shown in FIG. It is.

  A broken line indicated by a solid line is a case where the target image detection success rate is higher than 98% for a detection target. In this case, when the area ratio for all images is 70%, the correction amplification factor is set. Set to the maximum of 100%. When the area ratio is 10%, it means that the correction amplification factor only increases up to 70%. This is because the detection success rate is sufficiently high, so that it can be corrected to the maximum if it has a sufficient area ratio of 70%, but if the area ratio is low, even if the detection is successful, it is too much attention If the correction for the image is too much, the quality of the other parts deteriorates, which means that it is necessary to suppress the correction to some extent. The reason why the amplification factor is reduced when the area ratio exceeds 70% is that the area of the image of interest occupies most of the entire image, so that the necessity of performing correction itself is reduced in the first place.

  A broken line shown by a broken line is a case where the success rate of the target image detection is 90% to 98% and the calculated target to be detected. In this case, correction amplification is performed when the area rate for all images is 70%. Set rate to 70% of maximum. When the area ratio is 10%, it means that the correction amplification factor only increases up to 50%.

  A broken line indicated by a one-point difference line represents a case where the success rate of the target image detection is 80% to 90% and the calculated target to be detected. A broken line indicated by a two-point difference line represents a case where the success rate of the target image detection is 60% to 80% and the calculated detection target. It can be seen that the correction amplification factor increases as the success rate increases with an area ratio of 70% as a peak.

  FIG. 10 is a flowchart for determining a correction table from detection determination. In S1001, a detection process using the target image as a blue sky region is executed. After the detection process is completed, the process proceeds to S1002. In S1002, it is determined whether or not a final candidate has been detected as a blue sky detection process result. If not detected, the process proceeds to S1003, and an undetected flag is set and the process ends. If the final candidate is detected, the process proceeds to S1004, where the feature amount of the image area in which the final candidate is detected is acquired, and the detection success rate is calculated.

  In step S1005, it is determined whether the success rate of target image detection is set to 98% or more. If it matches, the process proceeds to step S1006, and a correction intensity table is generated and stored with reference to the broken line represented by the solid line in FIG. If not, go to S1007.

  In step S1007, it is determined whether the success rate of target image detection is set to 80% or more. If it matches, the process proceeds to step S1008, and a correction intensity table is generated and stored with reference to the broken line shown by the broken line in FIG. If not, go to S1009.

  In step S1009, a correction intensity table is generated and stored with reference to the broken line represented by the one-point difference line. In FIG. 9, there is also a broken line having a detection success rate of 60% to 80% for the two-point difference line, but in this embodiment, a detection rate of 80% or less is not subject to correction. However, this is merely an example, and all of those that determine the correction constant corresponding to the area ratio of the image of interest and the probability of detection are included.

  As described above, the correction value conversion table is stored, and in the subsequent image correction processing, the conversion table is read, the correction ratio is determined, and actual correction is performed. Further, in the present embodiment, the correction amount is optimized based on the success rate of detection, so that the image quality is not greatly deteriorated even when the target image is erroneously detected.

(Third embodiment)
Hereinafter, detection using a human skin region as a target image as a target image with the same configuration as that of the first embodiment will be described below.
FIG. 11 is an image sample taken as a portrait image. A human skin region exists in this image.

  The result of performing the human skin region detection process shown in FIG. 3 on the image sample is the image of FIG. In FIG. 12, candidate numbers are assigned to detection candidate groups in descending order of detection area, and information is handed over as data for correction centering on one candidate. The basic detection principle of human skin area detection is to detect those that match a preset chromaticity and luminance range in units of 8 × 8 pixel blocks, and to form candidate areas consisting of adjacent block sets of the detected blocks. Then, the DCT characteristic in that region is calculated to determine whether or not it matches the definition characteristic of the human skin.

  (1) Area information, (2) Luminance information, (3) Chromaticity information, and (4) DCT information calculated in the course of the determination process are stored as feature quantities for use in subsequent correction. The attention image detection process is executed on the condition of the feature amount obtained in this way, and the information of the attention image obtained from the detection result is reflected in the image processing.

  The delivery will be described with reference to FIG. The horizontal axis is the average value (Lave) of the luminance of the input image and takes a range from 0 to 255. The vertical axis represents the target value (NLave) of the luminance average value after conversion, and the range is from 0 to 255.

  The value of NLave is obtained for each luminance average (Lave) of the input image from the conversion table of FIG. 13, and the table of FIG. 14 is created from the result. The creation method plots the luminance average (Lave = A) value of the input image between 0 and 255 on the horizontal axis. A target value (NLave = NA) of the luminance average value after conversion is plotted between 0 and 255 on the vertical axis. The intersection point is searched for vertically from the plot point, and the conversion point is created by connecting the point and (0, 0) and (255, 255) as shown in the figure. By using this conversion table, the luminance value of the entire image can be converted to a desired value.

  For example, it is assumed that the image data is a person standing on a white silver slope (not shown). If the average luminance of the entire image is B (= 188) and the average luminance of the human skin detected in the target image detection is A (= 64), the image correction exposure process is performed when the target image detection process is not performed. Then, correction is performed to reduce the brightness of the entire image from 188 to 170. However, by correcting in this way, the face of the person who wants to see most as a subject is corrected in a darker direction than 64, so the image correction is performed instead. It gets worse. However, it can be reflected in the image correction value by using the output of the detected region of the target image. Therefore, if 64, which is the average luminance value of human skin, is the subject of exposure correction, exposure correction is performed by raising 64 to 100, and it is possible to improve the quality of the portion of the image data to be viewed.

  15 and 16 are flowcharts of operations according to this embodiment. The overall flow will be described with reference to the above figures. In step S1501, the human skin region detection shown in FIG. 3 is executed.

  FIG. 17 is a sample image for explaining the flowcharts shown in FIGS. 15 and 16. FIG. 18 shows a luminance histogram of this image. The horizontal axis represents the range from 0 to 255 at the left end 0. The vertical axis represents the luminance distribution for all images, and represents the relative amount with respect to the configuration of all images. The average brightness of the sample image in FIG. 17 is 84, but a peak of 240 or more bright parts and a dark part of the peak around 35 appear, and the brightness is distributed in two bipolar parts. I understand that. The face area of a person who attracts attention is about 90 and is dark.

  FIG. 19 is a diagram showing a result of correcting an image of this sample image by a conventional image correction method that does not extract a feature amount. First, regarding the contrast correction, since data is distributed in the dark part and the bright part, it is determined that the image is a good image that does not need to be corrected because the contrast is sufficiently detected. With regard to exposure correction, the average brightness is increased to about 90 by performing a correction that slightly increases the brightness level using the tables of FIGS. 13 and 14 as described above, but there is almost no change in the brightness of the face area of a person who attracts attention. Not improved.

Therefore, the correction of this embodiment is performed.
FIG. 20 is a diagram showing a result of area detection performed on the image sample of FIG. 17 by the flow of FIG. In this sample image, there are 1 to 5 candidate areas that match the chromaticity ratio range defined as the human skin color area and the DCT characteristics. It is considered that the candidate area is divided in the face and hand parts because the chromaticity ratio in the part between them is not suitable. The feature amount of the human skin area in this example is acquired from the detected candidate area.

  In step S1502, it is determined whether or not a human skin region is detected. If the human skin area cannot be detected, the process advances to step S1503 to set an undetected flag for the human skin area and end the human skin detection process flow. When the human skin region is detected in the image and the feature amount is acquired as in the image sample of FIG. 17, the process proceeds to area ratio determination 1 of the acquired image in S1504.

  In step S1504, a determination is made based on the area ratio of the detected human skin region to the entire image, and a table to be applied is determined. If the area ratio is 0.3% or less, the process proceeds to S1505, and otherwise, the process proceeds to S1506. In step S1506, it is determined whether or not the area ratio of the detected human skin region to the entire image is 0.3% to 5.0%. When the area ratio of the detected human skin region is 0.3% to 5.0%, the process proceeds to S1507, and otherwise, the process proceeds to S1508.

  Step S1507 is a case where the area ratio of the detected human skin region is 0.3% to 5.0%, the exposure setting table 1 is set, and the process proceeds to S1601 in FIG.

  FIG. 21 shows the exposure setting table 1. Here, the APD (Attention Picture Detection) detection value in the horizontal column uses the average luminance information in the detection region among the feature amounts detected from human skin as the target image. According to the detected luminance, it is divided into four, “0 to 35”, “36 to 80”, “81 to 120”, and “121 to 299”. The vertical column is average luminance information of all images. Depending on the brightness, it is divided into three parts, “0-60”, “61-120”, “121-180”.

  In the configuration of the exposure setting table, a certain ratio is determined from the relationship between the luminance of the detected human skin area and the luminance of the entire image, and exposure setting is performed using the human skin area detected at that ratio. As described above, by inserting the value calculated in the exposure setting table in place of the value of “Lave” in FIG. 20 described above, it is possible to increase the effect on the human skin area where the image correction is detected.

  Here, in “0 to 35” and “121 to 299” in the horizontal column, all items related to the vertical column are “OFF”. This is because the feature amount of the detected target image and the entire image are displayed. In this case, it is determined that it is better to perform normal exposure correction. In this case, exposure correction is performed using the tables of FIGS.

  In “36 to 80” and “81 to 120” in the horizontal column, the designated amount is calculated based on the ratio of the exposure setting table based on the value of the area detected as human skin. For example, when the luminance of the APD detection value is about 90 as in the image sample, the horizontal column is “81 to 120” and the luminance of the entire image is 84, so the vertical column is “61 to 120”. Then, “× 1.5” is selected as the designated ratio of this range. As a result, a value obtained by multiplying the average luminance of the detected human skin region by 1.5 is calculated, and the result obtained by calculating the value as the Love value in the table of FIG. 13 is used. For other columns, exposure correction is performed by calculating the relationship as described above.

  The area ratio of the human skin area detected in step S1504 is 0.3% or less. In this case, the exposure setting table 2 is set, and the process proceeds to S1601 in FIG. FIG. 22 shows the exposure setting table 2.

  In step S1508, it is determined whether the area ratio of the detected human skin region is 5.0% or more. The exposure setting table 3 is set and the process proceeds to S1601 in FIG. FIG. 23 shows the exposure setting table 3.

  In step S1601, the relationship that is set to “OFF” in the part for obtaining the constant for calculating the exposure setting table based on the detection result of the human skin region and the luminance information of the entire image, or the luminance of 0 to 255 If the value is not set in the column, such as when the value is outside the table range that does not require special table settings that reflect the detection result of the human skin area, the procedure does not proceed to S1602, and normal correction is performed. Exit to do. If it is within the table correspondence range, the process proceeds to S1602.

  In step S1602, a value calculated based on the exposure setting table is set. In the above description, in the image sample, the brightness value is 90 × 1.5 to 135 as the calculated value of Level. From this value, NLLave is derived from the table of FIG. From this result, a conversion table FIG. 14 can be created to generate an exposure correction table. Thereafter, image correction is performed in S104.

  FIG. 24 is a diagram illustrating an image correction result in which a result of human skin area detection is reflected on an image sample. The average image brightness is 129, and the brightness of the face of the person of interest is also corrected to a brightness of around 150.

  FIG. 25 is a luminance histogram of the image of FIG. The graph configuration is the same as the histogram shown in FIG. It can be seen that the peak of the dark part rises to near luminance 90 and the entire image is bright.

  In the present embodiment, the feature amount is calculated based on the result of the target image detection, and the correction value is corrected using the ratio of the detected area to the entire image in order to use the feature amount for correction. It will have a certain impact on the whole. For this reason, the correction amount that makes use of the feature amount is changed according to the area ratio. However, the determination method based on the area ratio may be a continuous change based on a mathematical expression for the area ratio, regardless of a plurality of threshold values.

  In this embodiment, the exposure correction setting amount is determined based on the region where the target image is detected and the brightness value of the entire image. However, this determination need not be based on a plurality of threshold values. Furthermore, the correction in image correction may be performed based on a mathematical expression of the relationship between the three parties based on the region feature amount of the target image detected, the area ratio in the entire image, and the feature amount of the entire image.

  As described above, the attention image in the present embodiment is not limited to the human skin, the blue sky, the night view, and the like described above, but it is needless to say that the attention image can be a target if it is a specific attention area. For example, the concept of the attention image such as a specific color gamut or a specific luminance range is also included as the attention area.

It is a figure which shows the flowchart showing the flow of the whole correction | amendment process concerning one Example of this invention. It is a figure which shows the flowchart which performs execution selection with respect to the some detection target of the attention image detection process concerning one Example of this invention. It is a figure which shows the flowchart which performs the information acquisition from the person skin detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the flowchart which performs the information acquisition from the blue sky detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the flowchart which performs the information acquisition from the night view detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the image sample used as the object of the blue sky detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the image of the result of having performed the attention image detection process to the original image sample concerning one Example of this invention. It is a figure showing the graph which shows the relationship between the detection area rate of the blue sky detection in the attention image detection process concerning one Example of this invention, and a detection success rate. It is a figure which shows the response | compatibility to the correction value by the detection probability and detection area rate when the attention image concerning one Example of this invention is made into a blue sky detection. It is a figure which shows the flowchart of correction of the correction value by the blue sky detection success rate in the attention image detection process concerning one Example of this invention. It is a figure which shows the image sample which performs the person skin detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the detection result which performed the attention image detection process to the original image sample concerning one Example of this invention. It is an example of the exposure correction amount determination table concerning one Example of this invention. It is a figure which shows the example of the look-up table for exposure correction concerning one Example of this invention. It is a figure which shows the flowchart which corrects a correction value from the person skin detection in the attention image detection process concerning one Example of this invention. It is a figure which shows the flowchart which corrects a correction value from the person skin detection in the attention image detection process concerning one Example of this invention. It is an image sample used as the object of human skin detection in the attention image detection process concerning one Example of this invention. It is the detection result sample 3 which performed the attention image detection process with respect to the original image sample 3 concerning one Example of this invention. It is a figure which shows the histogram of the image sample of FIG. 17 concerning one Example of this invention. It is a figure which shows the result of having performed normal image correction to the sample image concerning one Example of this invention. It is a figure which shows the detection result which performed the attention image detection process to the image sample concerning one Example of this invention. It is a figure which shows the optimization exposure correction table in case the area ratio of the human skin detection in the attention image detection process concerning one Example of this invention is more than 0.3% and less than 5%. It is a figure which shows the optimization exposure correction table in case the area rate of the person skin detection in the attention image detection process concerning one Example of this invention is 0.3% or less. It is a figure which shows the optimization exposure correction table in case the area rate of the person skin detection in the attention image detection process concerning one Example of this invention is 5% or more. It is a figure which shows the result of having corrected based on the information of attention image detection with respect to the sample image concerning one Example of this invention. It is a histogram of the FIG. 24 image concerning one Example of this invention. It is a block diagram which shows the structural example of the image processing apparatus of this embodiment. It is a figure which shows the structural example of the hardware and software which implement | achieve the image processing of this embodiment.

Explanation of symbols

2610 Decoding unit 2620 Color tone correction unit 2630 Image recognition unit 2631 Target color region detection unit 2632 Spatial frequency generation unit 2633 Target color region selection unit 2634 Color tone correction prohibition unit 2710 CPU
2720 ROM
2730 RAM
2731, 2741 Data storage area 2732, 2742 Program load area 2740 External storage unit 2750 Input interface 2760 Output interface

Claims (4)

A step of extracting a target image included in the input image data; and a correction step of correcting the input image data using a standard correction value calculated based on the feature amount of the target image to generate output image data. In the image processing method,
A step of analyzing the region of the noted image, obtains the feature information including the area ratio ratio is the ratio of the area of the target image with respect to the entire region of the input image data,
Calculating a success rate of detection of the target image from chromaticity information or DCT information included in the feature information;
A correction step of correcting the calculated standard correction value using the area ratio and the success rate to obtain an actual correction value;
Calculating an average luminance in the target image from luminance information in the region of the target image, wherein the correction step corrects input image data using the actual correction value, and the correction step includes the area ratio. And the greater the success rate, the larger the correction ratio of the standard correction value ,
The image processing method according to claim 1, wherein in the correction step, luminance correction is performed on the input image data based on the area ratio, the average luminance of the target image , and the average luminance of the input image data .   The image processing method according to claim 1, wherein the target image is a blue sky.   The image processing method according to claim 1, wherein the attention image is human skin. Extraction means for extracting an attention image included in the input image data, and correction means for correcting the input image data using the standard correction value calculated based on the feature amount of the attention image and generating output image data In the image processing apparatus provided,
Analyzing the region of the target image and obtaining feature information including an area ratio that is a ratio of the region of the target image to the entire region of the input image data;
Calculating means for calculating a success rate of detection of the target image from chromaticity information or DCT information included in the feature information;
Correction means for correcting the calculated standard correction value using the area ratio and the success rate to obtain an actual correction value;
Means for calculating an average luminance in the target image from luminance information in the region of the target image, the correction unit corrects input image data using the actual correction value, and the correction unit includes the area ratio. And the greater the success rate, the larger the correction ratio of the standard correction value ,
The image processing apparatus , wherein the correction unit performs luminance correction on the input image data based on the area ratio, average luminance in the target image , and average luminance of the input image data .

Images