TWI537877B - Method for detecting skin tone and skin tone detection system - Google Patents

Description

Skin color detection method and system

The present invention relates to a method and apparatus, and more particularly to a skin color detecting method and a skin color detecting device.

Skin color detection is an image processing category that can be used to classify whether pixels in a static or dynamic picture are skin tones. Since skin color information can be used to determine the face, hand, and body position, it can be applied in a wide range of applications, such as face detection, gesture recognition, fingertip detection, erotic image recognition, somatosensory games, and human-computer interaction.

However, the conventional technique uses statistical methods to generate skin color probability statistics when offline. However, when the skin color probability statistics table is used as the basis for skin color classification, the detection result may have false positive and detection. Missing detection, noise and incompleteness of skin color detection.

An object of the present invention is to provide a skin color detecting method and a skin color detecting system, wherein the correcting module can instantly reduce the skin color detection false positive rate and detect the missing rate. In addition, the correction screen can be fed back according to the detection result, and the skin color can be added or deleted manually or automatically to expand or reduce the range of detecting the skin color.

In order to achieve the above object, a technical aspect of the present invention relates to a skin color detecting method. The skin color detecting method comprises the steps of: obtaining an image; performing parameter correction on the image according to the calibration reference; and performing a table lookup on the skin color probability table according to the pixels of the image to generate a table lookup result; and checking the table according to at least one automatic correction criterion The result is judged to generate a judgment result; and the skin color probability statistics table is corrected based on the judgment result.

According to an embodiment of the invention, the skin color detecting method further comprises the steps of: feeding back the table lookup result, and correcting the correction reference according to the table lookup result.

According to another embodiment of the present invention, the skin color detecting method further includes the steps of: sampling a plurality of pictures; quantizing the color space of the pictures; and counting the probability that the color of each of the pictures is skin color, To generate a skin color probability chart.

According to another embodiment of the present invention, the skin color detecting method further includes performing manual skin color marking on the plurality of pictures, quantizing the manual skin color marking result, and performing statistical or supervised machine learning to calculate each of the foregoing pictures. The color of the person is the probability of skin color to produce a skin color probability chart.

According to another embodiment of the present invention, the skin color detecting method further includes performing image processing on the image to generate image processing results, and adding or correcting automatic correction criteria according to the image processing result.

According to still another embodiment of the present invention, the step of correcting the skin color probability rate table according to the determination result is performed once every other fixed number of images.

According to still another embodiment of the present invention, the skin color probability statistics table is manually corrected according to the skin color input by the user and the non-skin color samples.

In order to achieve the above object, another aspect of the present invention relates to a skin color detecting system. The skin color detection system includes an image input module, a parameter correction module, a table access module, a table lookup result storage module, and an automatic correction module. In operation, the image input module is configured to obtain an image; the parameter correction module includes a calibration reference, the parameter correction module is configured to perform parameter correction on the image according to the calibration reference; and the automatic calibration module is configured to perform according to at least one automatic correction criterion. The result of the look-up table is judged to generate a judgment result, and the judgment result is transmitted to the form access module, and the form access module corrects the skin color probability statistics table according to the judgment result, thereby making the next judgment result more correct. .

According to an embodiment of the invention, the skin color detection system further includes a feedback correction module, and the table access module is configured to perform a table lookup on the skin color probability statistics table according to the pixels of the image to generate a table lookup result, wherein the feedback correction module is configured by The table access module obtains the result of the lookup table, and returns the result of the lookup table to the parameter correction module, and the parameter correction module corrects the calibration reference of the parameter correction module according to the result of the lookup table.

According to another embodiment of the present invention, the skin color detection system further includes a skin color image sample processing module, a quantization module, and a statistics module. In operation, the skin color image sample processing module is configured to perform automatic or manual skin color marking on a plurality of images; the quantization module is configured to quantize the color of the color space to select a certain number of representative colors. And the statistical module uses statistical or supervised machine learning to calculate the probability that each representative color in each of the aforementioned pictures is marked as skin color and non-skin tone to calculate the skin color probability to generate a skin color probability statistics table.

According to another embodiment of the present invention, the skin color detection system further includes an image processing module for performing image processing on the image to generate image processing results, and providing image processing results to the automatic correction module to add or modify automatic correction criteria. . In operation, the image processing module is configured to perform image processing on the image to generate image processing results. For example, the image processing module can perform motion pixel detection, face detection or other skin color detection technologies, and provide image processing. The result is calculated by the detection unit of the automatic calibration module, and the calculation result is stored in the database for changing, intersecting, and collecting the skin color detection result, and then the skin color probability table is corrected by the automatic correction module to expand or reduce the skin color detection. range.

According to still another embodiment of the present invention, the automatic correction module in the skin color detection system transmits the determination result to the form access module every other fixed number of images to correct the skin color probability statistics table.

According to still another embodiment of the present invention, the manual correction module in the skin color detection system can add or delete skin color according to the skin color or non-skin color image provided by the user, so that the skin color pixel recorded by the skin color probability rate table can be Dynamic Adjustment.

Therefore, according to the technical content of the present invention, an embodiment of the present invention provides a skin color detecting method and a skin color detecting system, which can detect a skin color detection error when the skin color detecting system and the method are detected by an automatic correction module and a manual correction module. The skin color probability statistics table can be corrected instantly to make the skin color probability statistics table more complete, so that the skin color detecting system and the method of the embodiment of the present invention are more accurate, and the parameters of the input image can be made by the feedback correcting module. The adjustment is more correct to make the skin detection system result more accurate.

In addition, the skin color detection system and method of the embodiment of the present invention can correct the skin color probability statistics table by the correction mechanism every fixed image number. Therefore, the skin color detection system and method of the embodiment of the present invention can be Correct the skin color probability statistics in the case of excessive use of system resources.

In order to make the description of the present disclosure more complete and complete, reference is made to the accompanying drawings and the accompanying drawings. However, the embodiments provided are not intended to limit the scope of the invention, and the description of the operation of the structure is not intended to limit the order of its execution, and any device that is recombined by the components produces equal devices. The scope covered by the invention.

The drawings are for illustrative purposes only and are not drawn to the original dimensions. On the other hand, well-known elements and steps are not described in the embodiments to avoid unnecessarily limiting the invention.

FIG. 1 is a block diagram showing a skin color detection system 1000 according to an embodiment of the invention. As shown, the skin color detection system 1000 includes an image input module 1010, a parameter correction module 1030, a table access module 1060, a table lookup result generation module 1050, a feedback correction module 1040, an automatic correction module 1080, and The module 1100 is manually corrected.

In operation, the image input module 1010 is used to acquire images. The image input module 1010 refers to any device that can obtain images, such as a camera, a network disk, a database, etc., but the invention is not limited thereto. . The parameter correction module 1030 includes a calibration reference, and the parameter correction module 1030 performs parameter correction on the image according to the calibration reference. For example, if one of the parameters set by the calibration reference is a luminance value of 130, when the luminance value of the image is 50, the parameter correction module 1030 increases the luminance value of the image by 80 according to the calibration reference, thereby making the image The brightness value reaches 130, but the invention is not limited thereto. The parameter correction module 1030 can correct the remaining parameters in addition to the brightness value of the image.

In addition, the table access module 1060 is configured to perform a lookup table according to the pixels of the image to the skin color probability rate table 1070 to generate a table lookup result, and the result of the table lookup may be temporarily stored in the table lookup result storage module 1050, or directly transmitted to the table lookup module 1050. In the embodiment, the table access module 1060 can include a query form, a storage form, a loading form, a color skin color probability, a color skin color probability, and a color setting. Functions such as skin color are not limited to the actual implementation.

The automatic correction module 1080 is configured to determine the result of the lookup table according to the at least one automatic correction criterion to generate a determination result, and transmit the determination result to the form access module 1060, and the form access module 1060 determines, according to the determination result, Correct the skin color probability table 1070. For example, the foregoing automatic correction criterion may be an isolated point judgment method, and the automatic correction module 1080 determines the result of the lookup table according to the isolated point judgment method as described in the following paragraph.

In the pixel matrix of a nine-square grid in the above image, if the result of the lookup of the isolated pixel in the middle is skin color, and the results of the surrounding eight pixels are not skin color, then according to the rule of the isolated point judgment method, it is determined The isolated pixel in the middle is also not the skin color. The result of the above judgment is that the RGB value of the isolated pixel in the middle does not represent the skin color, and the result of the determination is transmitted back to the table access module 1060, and the table access module 1060 determines the result according to the judgment. The above-described isolated pixel RGB values in the skin color probability table 1070 are corrected to be non-skin color. However, the above automatic correction criterion is not limited to the isolated point judgment method, and those skilled in the art can selectively adopt an appropriate judgment rule as the above-described automatic correction criterion.

The above automatic correction method can be implemented by a series of detecting units 1082 for automatically calculating false positive pixels and false non-skin detections, and correcting skin color probability statistics through the table access module 1060. table. The detection modes of several detection units 1082 are listed below:

1. In the detection result, one pixel is a skin color pixel, and its adjacent n pixels (nine squares) are not skin color pixels, indicating that the isolated point is a false skin color (false positive), then the color is corrected to be non- For skin tone, n can be a constant value (for example: 7 or 8).

2. In the detection result, one pixel is a non-skinning pixel, and the adjacent n pixels are skin color pixels, indicating that the isolated point is a false non-skin detection pixel, and the color can be tested by other laws. Whether it is skin color, if it is corrected, this color is skin color, and n can be a constant value (for example: 7 or 8).

3. In the detection result, the noise detection technology is used to detect the skin color pixel, and the noise is a false skin color pixel, and the color is corrected to be non-skin color. For example, it is possible to perform noise removal on the skin image, and subtract the original image from the noise-removed image to obtain noise, and the skin color probability table 1070 can be corrected according to the image processing result.

4. In the detection result, use the skin color pixel erosion, expansion operation, and then test whether the color after expansion is skin color by other rules, and then correct this color to skin color.

In addition, the skin color detection system 1000 further includes a manual correction module 1100 for adding or deleting colors marked as skin color in the skin color chance statistics table 1070 according to the skin color or non-skin color images provided by the user. Accordingly, the skin color detection system 1000 of the embodiment of the present invention provides a mechanism for the user to actively correct the skin color probability rate table 1070.

In an embodiment, the feedback correction module 1040 corrects the calibration reference of the parameter correction module 1030 according to the result of the lookup table. Therefore, the parameter correction module 1030 can perform white balance, automatic exposure, soft and hard parameters, and the like on the original image. Make corrections so that the input of the skin tone rate table is more correct. For example, the table access module 1060 looks up the skin color probability table 1070 according to the RGB value of each pixel in the image to determine whether each pixel in the image is a skin color, and then generates a table lookup result. The result of this lookup is fed back to the feedback correction module 1040, by which the parameter correction module 1030 corrects the calibration reference. However, the pixel representation of the image of the present invention is not limited to the RGB mode, and may be selectively implemented using different color spaces.

As described above, the skin color detection system 1000 of the embodiment of the present invention includes a correction mechanism. When the skin color detection system 1000 detects a skin color detection error, the skin color detection system 1000 can immediately correct the skin color detection system 1000. The skin color probability statistics table 1070 makes the skin color probability rate table 1070 more complete, and instantly reduces the skin color detection false positive rate and the detection leakage rate, thereby making the judgment of the skin color detection system 1000 of the embodiment of the present invention more accurate. The feedback correction module 1040 can be used to make the parameter adjustment of the input image more correct, so that the result of the skin color detection system is more accurate.

In the embodiment of the present invention, the skin color detection system 1000 is required to correct the skin color probability rate table 1070 because the detection result may have a false positive and a miss detection, resulting in a complex skin color detection. News and incomplete. Each of the above correction modules can instantly reduce the skin color detection false positive rate and detect the leakage rate. In addition, it can be manually corrected or automatically corrected according to the detection result to add or delete skin color, thereby expanding or reducing the range of detecting skin color. These corrections can be made in a continuous or intermittent manner depending on the actual application requirements.

In an embodiment, the skin color detection system 1000 further includes a skin color image sample processing module 1110, a quantization module 1120, and a statistics module 1130. In operation, the skin color image sample processing module 1110 is configured to perform automatic or manual skin color marking on a plurality of images; the quantization module 1120 is configured to quantize the automatic or manual skin color marking result; and the statistical module 1130 adopts statistical or supervised The machine learning method calculates the probability that each of the aforementioned pictures is a skin color to generate a skin color probability table 1070.

In detail, the skin color image sample processing module 1110 performs automatic or manual skin color marking on a plurality of pictures, and then the quantization module 1120 quantizes the color of the color space to select a certain number of representative colors. Then, the statistical module 1130 calculates the skin color probability of each representative color by means of statistical or supervised machine learning. If the calculated skin color probability is greater than a threshold, the color is determined to be the skin color, and then each color is determined. A table is created for the determination result of the skin color, and the above table is the skin color probability table 1070.

For example, the skin color probability statistics table 1070 can be used to automatically or manually circle or mark skin color pixels, and then determine the number of colors recorded in the statistical table, and use a quantization algorithm to determine the statistical table. represent color. Then, each pixel color in the picture is calculated by a quantization algorithm to determine which color of the statistical table is, and the number of times belonging to the skin color or non-skin color is recorded, and then the probability that each color belongs to the skin color is calculated. Finally, set a threshold. If the color of any color is greater than the threshold, set the color to skin tone, otherwise set it to non-skin tone. The above skin color probability statistics table 1070 is as shown in Table 1.

As shown in Table 1, the leftmost color label, C1, C2, and C3 represent the three channels of the color space, Positives(a) represents the number of times the color is recorded as skin color, and Negatives(b) represents the color recorded as non-skin. The number of times, Prob (a / a + b) represents the ratio of the color to the skin color, and Results (th = 0.4) is the result of the final determination of whether the color is the skin color.

For example, if the color numbered 0 in Table 1 is recorded as 50 times for skin color and 100 times for non-skin color, the ratio of skin color is 0.33. Here, a threshold is set ( Th) is 0.4, and since the skin color ratio of the aforementioned color is less than 0.4, the aforementioned color is judged to be non-skin tone (No). However, the values used in the above thresholds are not intended to limit the invention, and those skilled in the art can selectively implement the invention with appropriate thresholds.

In another embodiment, the skin color detection system 1000 further includes an image processing module 1090 for performing image processing on the image to generate image processing results, and providing image processing results to the automatic correction module to add or correct automatic corrections. Guidelines. In operation, the image processing module 1090 is configured to perform image processing on the image to generate image processing results. For example, the image processing module 1090 can perform motion pixel detection, face detection, or other skin color detection technologies, and provide The image processing result is calculated by the detecting unit 1082 of the automatic correction module 1080, and the calculation result is stored in the database 1084 for changing, intersecting, and collecting the skin color detection result, and then the skin color probability table 1070 is corrected by the automatic correction module 1080. Further expand or reduce the skin color detection range.

In an optional embodiment, the automatic correction module 1080 in the skin color detection system 1000 transmits the determination result to the table access module 1060 every other fixed number of images, for example, every 60 images, and saves the table by the table. The module 1060 retrieves the skin color probability table 1070 based on the determination result. However, it is not intended to limit the present invention, and those skilled in the art can selectively implement the present invention by using appropriate sampling intervals. Thus, the skin color detection system 1000 of the embodiment of the present invention can be borrowed every fixed number of images. The skin color probability rate table 1070 is modified by the correction mechanism. Therefore, the skin color detection system 1000 of the embodiment of the present invention can correct the skin color probability rate table 1070 without excessively utilizing system resources, so that the skin color probability rate table 1070 is further improved. The integrity is such that the judgment of the skin color detection system 1000 of the embodiment of the present invention is more accurate.

In an embodiment, the automatic correction module 1080 of the skin color detection system 1000 includes a detection unit 1082 and a database 1084, wherein the database 1084 includes an automatic correction criterion, and the detection unit 1082 obtains an automatic correction criterion from the database 1084, and The result of the look-up table is judged according to the automatic correction criterion to generate a judgment result.

In another embodiment, the skin color detection system 1000 further includes a manual correction module 1100 for adding or deleting colors marked as skin color in the skin color probability table 1070 according to the skin color or non-skin color image provided by the user. . Accordingly, the skin color detection system 1000 of the embodiment of the present invention provides a mechanism for the user to actively correct the skin color probability rate table 1070.

FIG. 2 is a schematic flow chart of a skin color detecting method 200 according to another embodiment of the present invention. As shown, the skin color detection method includes the steps of: acquiring an image (step 210); performing parameter correction on the image according to the calibration reference (step 230); and performing a table lookup to the skin color probability table according to the pixel of the image to generate Checking the result of the table (step 240); feeding back the result of the lookup table, and correcting the correction reference according to the result of the lookup table (step 250); determining the result of the lookup table according to at least one automatic correction criterion to generate a judgment result (step 260); The result is to correct the skin tone probability table (step 270).

Please refer to Figure 1 and Figure 2 together. In step 210, the step of acquiring an image may be performed by the image input module 1010. The image input module 1010 generally refers to any device that can obtain images, such as a camera, a network disk, a database, etc., but the present invention Not limited to this. Referring to step 230, the parameter correction module 1030 can perform parameter correction on the image according to the calibration reference. The parameter correction module 1030 includes a calibration reference. For example, if one of the parameters set by the calibration reference is a brightness value of 130, when the brightness value of the image is 50, the parameter correction module 1030 increases the brightness value of the image by 80 according to the calibration reference, thereby making the brightness of the image. The value reaches 130, but the invention is not limited thereto. The parameter correction module 1030 can correct the remaining parameters in addition to the brightness value of the image.

In step 240, the table access module 1060 can perform a lookup table according to the pixels of the image to the skin color probability rate table 1070 to generate a lookup result. In addition, the result of the lookup may be temporarily stored in the lookup result storage module 1050, or directly fed back to the parameter correction module 1030, and the parameter correction module 1030 may correct the calibration reference according to the result of the lookup table.

For example, the table access module 1060 can perform a lookup to the skin color probability table 1070 according to the RGB value of each pixel in the image to determine whether each pixel in the image is a skin color, and then generate a table lookup result. However, the pixel representation of the image of the present invention is not limited to the RGB mode, and may be selectively implemented using different color spaces.

Referring to step 260, the auto-correction module 1080 can determine the result of the look-up table according to the at least one automatic correction criterion to generate a determination result. In step 270, the table access module 1060 can correct the skin color according to the judgment result. Probability statistics table 1070. For example, the foregoing automatic correction criterion may be an isolated point judgment method, and the manner in which the automatic correction module 1080 determines the result of the look-up table according to the isolated point judgment method has been described in detail in the correlation of FIG. 1 . In the description, it will not be described here.

As described above, the skin color detecting method 200 of the embodiment of the present invention can implement the correction mechanism. By using the correcting mechanism, when the skin color detecting method 200 finds an error, the error can be corrected immediately, and the skin color probability statistics table can be synchronously corrected. 1070, the skin color probability statistics table 1070 is more complete, so that the judgment of the skin color detecting method 200 of the embodiment of the present invention is more accurate.

The reason why the skin color detecting method 200 disclosed in the embodiment of the present invention corrects the skin color probability meter 1070 is that the detection result may have false positive and miss detection, resulting in skin color detection. The noise is not complete. Each of the above correction modules can instantly reduce the skin color detection false positive rate and detect the leakage rate. In addition, the correction input screen can be fed back according to the detection result, and the manual correction or the automatic correction can be added to add or delete the skin color, thereby expanding or reducing the range of detecting the skin color. These corrections can be made for each picture, or intermittently.

The above automatic correction method can be implemented by a series of detecting units 182 for automatically calculating false positive pixels and false non-skin detections, and correcting them through the table access module 1060 (for example: Add or remove) skin color probability statistics. The detection mode of the detecting unit 182 has been described in detail in the description of FIG. 1 and will not be described herein.

In an embodiment, the skin color detecting method 200 further includes the following steps: performing image processing on the image to generate an image processing result, and correcting the skin color probability table 1070 according to the image processing result (step 280).

In step 280, image processing is performed on the image to generate image processing results, and image processing results are provided to the automatic correction module to add or correct automatic correction criteria. The image processing module 1090 is configured to perform image processing on the image to generate image processing results. For example, the image processing module 1090 can perform motion pixel detection, and the image processing module 1090 can be executed by the image processing module 1090. Face detection or other skin color detection technology, and provide image processing results to the detection unit 1082 of the automatic correction module 1080, and the calculation results are stored in the database 1084 for changing, intersecting, and collecting skin color detection results, and further The skin color probability table 1070 is corrected by the automatic correction module 1080 to further enlarge or reduce the skin color detection range.

In step 270 of the skin color detecting method 200, according to the step of determining the skin color probability table according to the determination result, the automatic correction module 1080 can transmit the determination result to every fixed number of images, for example, every 60 images. The form access module 1060, if the form access module 1060 correspondingly fixes the number of images, adjusts the skin color probability table 1070 according to the determination result. However, it is not intended to limit the invention, and those skilled in the art will be able to implement the invention selectively using the appropriate sampling spacing.

In this manner, the skin color detection method 200 of the embodiment of the present invention can correct the skin color probability rate table 1070 by using a correction mechanism every fixed image number. Therefore, the skin color detection method 200 of the embodiment of the present invention can be However, in the case of utilizing the system resources, the skin color probability table 1070 is corrected to make the skin color probability table 1070 more complete, so that the judgment of the skin color detecting method 200 of the embodiment of the present invention is more accurate.

FIG. 3 is a flow chart showing a method 300 for generating a skin color probability rate in a skin color detecting method according to another embodiment of the present invention. The method 300 for generating the skin color probability table includes the following steps: performing automatic or manual skin color marking on a plurality of pictures (step 310); quantifying the color of the color space to select a certain number of representative colors. (Step 320); calculating the probability that the color of each of the aforementioned pictures is the skin color using statistical or supervised machine learning (step 330); and generating a skin color probability statistics table (step 340).

Referring to step 310, the skin color image sample processing module 1110 may perform automatic or manual skin color mark on the plurality of pictures. Then, in step 320, the step of quantizing the automatic or manual skin color mark results of the pictures may be borrowed. </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> </ RTI> The module 1130 generates a skin tone probability table 1070.

In detail, the skin color image sample processing module 1110 can perform automatic or manual skin color mark on the collected skin color training samples (the plurality of pictures), and then automatically or manually color the skin color training samples by the quantization module 1120. The result of the labeling is quantified, and then the statistical module 1130 uses statistical or supervised machine learning to calculate the skin color probability of each color. If the calculated skin color probability is greater than a threshold, the color is determined to be the skin color. Then, the determination result of whether each color is skin color is made into a table, and the skin color probability table 1070 is generated by the above steps.

In still another embodiment, the skin color detecting method 200 further includes the step of manually correcting the skin color probability statistics table according to the input skin color and the non-skin color samples. In this step, the skin color can be added or deleted according to the skin color or non-skin color image provided by the user by the manual correction module 1100, so that the skin color pixel recorded by the skin color probability meter can be dynamically adjusted. Accordingly, the skin color detection method 200 of the embodiment of the present invention provides a mechanism for the user to actively correct the skin color probability rate table 1070.

The skin color detection method 200 as described above can be performed by software, hardware, and/or carcass. For example, if the execution speed and accuracy are the primary considerations, the hardware and/or the carcass may be mainly used; if the design flexibility is the primary consideration, the software may be mainly used; or At the same time, the software, hardware and carcass work together. It should be understood that the above examples are not intended to limit the present invention, and are not intended to limit the present invention. Those skilled in the art will need to design elastically at that time.

Moreover, those skilled in the art can understand that the steps in the skin color detection method 200 are named according to the functions they perform, only to make the technology of the present invention more obvious and understandable, and not to limit such step. It is still an embodiment of the present disclosure to integrate the steps into the same step or to split into multiple steps, or to replace any of the steps into another step.

It will be apparent from the above-described embodiments of the present invention that the application of the present invention has the following advantages. The embodiment of the present invention provides a skin color detection system 1000 and a skin color detection method 200. When the skin color detection system 1000 and the method 200 detect a skin color detection error, the skin color detection system 1000 and the method 200 can immediately detect The skin color probability table 1070 is corrected to make the skin color probability table 1070 more complete, and the skin color detection false positive rate and the detection loss rate are instantly reduced, thereby making the skin color detecting system 1000 and the method 200 of the embodiment of the present invention more Accurate, in addition, the correction input screen can be returned according to the detection result to make the table lookup method more correct.

In addition, the skin color detection system 1000 and the method 200 of the embodiment of the present invention can correct the skin color probability rate table 1070 by using a correction mechanism every fixed number of images. Therefore, the skin color detection system 1000 of the embodiment of the present invention and The method 200 can correct the skin tone probability table 1070 without excessively utilizing system resources.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and the present invention can be modified and modified without departing from the spirit and scope of the present invention. The scope is subject to the definition of the scope of the patent application attached.

200. . . Skin color detection method

210～280. . . step

300. . . Method for generating skin color probability statistics

310～330. . . step

1000. . . Skin color detection system

1010. . . Image input module

1030. . . Parameter correction module

1040. . . Feedback module

1050. . . Query result storage module

1060. . . Table access module

1070. . . Skin color probability statistics

1080. . . Automatic calibration module

1082. . . Detection unit

1084. . . database

1090. . . Image processing module

1100. . . Manual calibration module

1110. . . Skin color image sample processing module

1120. . . Quantization module

1130. . . Statistical module

The above and other objects, features, advantages and embodiments of the present invention will become more apparent and understood.

FIG. 1 is a block diagram showing a skin color detecting system according to an embodiment of the invention.

FIG. 2 is a schematic flow chart of a skin color detecting method according to another embodiment of the present invention.

FIG. 3 is a schematic diagram showing a method for generating a skin color probability rate table in a skin color detecting method according to still another embodiment of the present invention.

200. . . Skin color detection method

210～280. . . step

Claims (12)

A method for detecting a skin color includes: obtaining an image; performing a lookup table on a skin color probability table according to pixels of the image to generate a lookup result; determining the result of the table according to at least one automatic correction criterion to generate a As a result of the determination, the automatic correction mode of the automatic correction criterion includes detecting that the pixel of the image is a false skin color pixel or a false non-skin color pixel; and correcting the skin color probability statistics according to the determination result. The skin color detecting method of claim 1, further comprising: performing parameter correction on the image according to a calibration reference; and feeding back the lookup result, and correcting the calibration reference according to the lookup result. The skin color detecting method of claim 1, further comprising: performing automatic or manual skin color marking on the plurality of pictures; quantifying the automatic or manual skin color marking result; and calculating the method by using statistical or supervised machine learning. The color of each of the pictures is the probability of skin color to produce the skin color probability table. The skin color detecting method of claim 1, further comprising: performing image processing on the image to generate an image processing result, and adding or correcting the automatic correction criterion according to the image processing result. The skin color detecting method according to claim 1, wherein the step of correcting the skin color probability statistics according to the determination result is performed once every fixed number of images. The skin color detecting method of claim 1, further comprising: manually correcting the skin color probability statistics according to the input skin color and the non-skin color samples. A skin color detection system includes: an image input module for acquiring an image; and a table access module for performing a table lookup to a skin color probability table according to pixels of the image to generate a lookup result And an automatic correction module, configured to determine the result of the lookup according to the at least one automatic correction criterion to generate a determination result, and transmit the determination result to the form access module, the automatic correction criterion is automatic The calibration method includes detecting a pixel of the image as a false skin color pixel or a false non-skin color pixel, wherein the form access module corrects the skin color probability statistics according to the determination result. The skin color detection system of claim 7, further comprising: a parameter correction module, comprising a calibration reference, wherein the parameter correction module performs parameter correction on the image according to the calibration reference; and a feedback correction mode a group for obtaining the result of the lookup table by the form access module, and feeding back the result of the lookup table to the parameter correction module, wherein the parameter is corrected The positive module corrects the calibration reference based on the results of the lookup table. The skin color detection system of claim 7, further comprising: a skin color image sample processing module for automatically or manually coloring the plurality of pictures; and a quantization module for marking the automatic or manual skin color The results are quantified; and a statistical module that uses statistical or supervised machine learning to calculate the probability that each of the images is a skin color to produce the skin color probability table. The skin color detection system of claim 7, further comprising: an image processing module for performing image processing on the image to generate an image processing result, and adding or correcting the automatic correction according to the image processing result Guidelines. The skin color detection system of claim 7, wherein the automatic correction module transmits the determination result to the form access module every other fixed number of images, and the form access module corrects the result according to the determination result. Skin color probability statistics. The skin color detection system of claim 7, further comprising: a manual correction module for adding or deleting the skin color probability rate table according to the skin color or non-skin color image input to the skin color detection system The color of the skin tone.