TWI729338B - Skin detection method and image processing device - Google Patents

Abstract

The disclosure provides a skin detection method and an image processing device. The method includes: retrieving a human face image; retrieving a mask according to a detection item and defining a detection region in the human face image based on the mask; transforming the human face image into channel images; finding out a skin feature corresponding to in the detection region of each channel image; calculating a reference score corresponding to the skin feature of the detection item in each channel image; calculating a detection score of the human face image at the detection item according to the skin feature of the detection item; calculating a calibrated score corresponding to the detection item according to historical detection scores and the detection score.

Description

Skin detection method and image processing device

The present invention relates to a detection method and device, and more particularly to a skin detection method and image processing device.

In recent years, with the improvement of people's living standards, people's awareness of health care and the pursuit of beauty continue to increase. Therefore, skin care and health care have become an increasingly prosperous industry. At the same time, consumers and the industry have more demand for detection devices for various skin characteristics and parameters (such as whiteness, moisture, elasticity, etc.).

Based on this demand, some skin testing instruments have appeared on the market in recent years, but most of them have disadvantages such as expensive, bulky, and difficult to operate, so they are difficult to popularize. Take the full-face skin image analyzer as an example. It can use different illumination modes to detect facial images. However, due to its high price, bulkiness and difficulty in operation, most users will not consider purchasing it. In addition, there are also instruments that use different skin electrical impedance parameters for water content conversion, but they can only perform partial detection and cannot analyze the overall facial features. In addition, the test items that this instrument can perform are also limited and difficult to expand. In addition, there is also a skin tomography scanner on the market, which can scan through light to present the tomographic structure of the skin tissue in real time. However, this skin tomography can only perform partial detection and cannot analyze the overall facial features.

The "prior art" paragraph is only used to help understand the content of the present invention, so the contents disclosed in the "prior art" paragraph may include some conventional technologies that do not constitute the common knowledge in the technical field. The content disclosed in the "prior art" paragraph does not represent the content or the problem to be solved by one or more embodiments of the present invention, and has been known or recognized by persons with ordinary knowledge in the technical field before the application of the present invention.

In view of this, the present invention provides a skin detection method and image processing device, which can be used to solve the above technical problems.

The present invention provides a skin detection method, including: obtaining a face image; obtaining a first mask according to a first detection item, and defining a detection area in the face image based on the first mask; and converting the face image into Multiple channel images; find a skin feature corresponding to the first detection item in the detection area of each channel image; calculate a reference score corresponding to the skin feature of the first detection item in the detection area of each channel image; according to the first The reference score corresponding to the skin feature of the detection item calculates a detection score of the face image in the first detection item; and a first corrected score corresponding to the first detection item is calculated according to the multiple historical detection scores and the detection scores.

In an embodiment of the present invention, the aforementioned step of obtaining a face image includes obtaining an image and performing a face recognition operation on the image to find the face image in the image.

In an embodiment of the present invention, the above-mentioned step of converting a face image into a channel image includes: determining a pre-processing mechanism and a color conversion mechanism according to the first detection item; and using the color conversion mechanism to convert the face image into multiple images. A color channel image; and using a pre-processing mechanism to convert the color channel image into a channel image.

In an embodiment of the present invention, the aforementioned skin feature includes a wrinkle, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each channel image includes: for a first channel image in the channel image , Calculate a wrinkle area of wrinkles in the detection area; calculate a proportion of the wrinkle area in the area of the detection area; and define the ratio as the reference score of the first channel image.

In an embodiment of the present invention, the aforementioned skin feature includes a non-smooth skin, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each channel image includes: for a first channel image in the channel image In other words, calculating a non-smooth area of the non-smooth skin in the detection area; calculating a ratio of the non-smooth area in the area of the detection area; and defining the ratio as the reference score of the first channel image.

In an embodiment of the present invention, each of the above-mentioned channel images further includes a reference area, the skin feature includes a shiny skin, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each channel image includes: For a first channel image in the channel image, calculate a first brightness of the skin oily light in the detection area and a second brightness of the skin oily light in the reference area; subtract the second brightness from the first brightness to generate a difference Value; and the difference is defined as the reference score of the first channel image.

In an embodiment of the present invention, the aforementioned skin feature includes a skin color, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each channel image includes: for a first channel image in the channel image Calculate a skin color standard deviation of the skin color in the detection area, where the detection area does not include the facial features in the face image; and define the skin color standard deviation as the reference score of the first channel image.

In an embodiment of the present invention, each channel image described above further includes a reference area, the skin feature includes a dark circle, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each channel image includes: For a first channel image in the image, calculate a first brightness of the dark circles in the detection area and a second brightness of the reference area; subtract the first brightness from the second brightness to generate a difference; and The value is defined as the reference score of the first channel image.

In an embodiment of the present invention, the above-mentioned channel image includes a first channel image, a second channel image, and a third channel image, and the face image is calculated according to the reference score corresponding to the skin feature of the first detection item. The step of detecting the score of the item includes: multiplying the reference score of the first channel image by a first weight to generate a first value; and multiplying the reference score of the second channel image by a second weight to generate a second value; The reference score of the third channel image is multiplied by a third weight to generate a third value; and the first value, the second value, and the third value are added to form the detection score.

In an embodiment of the present invention, when the aforementioned skin feature is a dark circle or a non-smooth skin, the first channel image is a Y channel image, the second channel image is a Cr channel image, and the third channel image is A Cb channel image, and the first weight, the second weight, and the third weight are sequentially decreased.

In an embodiment of the present invention, when the skin feature is a wrinkle or a spot, the first channel image is an R channel image, the second channel image is a G channel image, and the third channel image is a B channel image. Image, and the first weight, second weight, and third weight are all 1.

，其中SC為第一校正後分數，RC為檢測分數，m為平均值，σ為標準差。In an embodiment of the present invention, the step of calculating the first corrected score based on the historical detection score and the detection score includes: calculating an average value and a standard deviation of the historical detection score; and inputting the detection score into an excitation function to generate The first corrected score, where the activation function is:

, Where SC is the first corrected score, RC is the detection score, m is the average, and σ is the standard deviation.

In an embodiment of the present invention, the above method further includes: generating a second corrected score corresponding to a second detection item based on the face image; multiplying the first corrected score by a first specific weight to Generate a first specific value; multiply the second corrected score by a second specific weight to generate a second specific value; and add the first specific value and the second specific value to a comprehensive skin detection score.

The invention provides an image processing device, which includes an image capture circuit, a storage circuit and a processor. The storage circuit stores multiple modules. The processor is coupled to the image capture circuit and the storage circuit, and accesses the modules to perform the following steps: obtain a face image; obtain a first mask based on a first detection item, and based on the first mask on the face Define a detection area in the image; convert the face image into multiple channel images; find a skin feature corresponding to the first detection item in the detection area of each channel image; calculate the first detection in the detection area of each channel image A reference score corresponding to the skin feature of the item; calculating a detection score of the face image in the first detection item based on the reference score corresponding to the skin feature of the first detection item; and calculating a detection score corresponding to the first detection item based on multiple historical detection scores and detection scores A first corrected score of a test item.

In an embodiment of the present invention, the aforementioned processor is configured to obtain an image and perform a face recognition operation on the image to find a face image in the image.

In an embodiment of the present invention, the aforementioned processor is configured to: determine a pre-processing mechanism and a color conversion mechanism according to the first detection item; use the color conversion mechanism to convert the face image into multiple color channel images; and Use the pre-processing mechanism to convert the color channel image to the channel image.

In an embodiment of the present invention, the aforementioned skin feature includes a wrinkle, and the processor is configured to: for a first channel image in the channel image, calculate a wrinkle area of the wrinkle in the detection area; and calculate the wrinkle The ratio of the area in the area of the detection area; and the ratio is defined as the reference score of the first channel image.

In an embodiment of the present invention, the aforementioned skin feature includes a non-smooth skin, and the processor is configured to: for a first channel image in the channel image, calculate a non-smooth skin in the detection area. Smooth area; calculate the proportion of the non-smooth area in the area of the detection area; and define the proportion as the reference score of the first channel image.

In an embodiment of the present invention, each channel image described above further includes a reference area, the skin feature includes a skin shine, and the processor is configured to: for a first channel image in the channel image, calculate the skin shine A first brightness in the detection area and a second brightness in the reference area of the skin oily light; subtracting the second brightness from the first brightness to generate a difference; and defining the difference as the first channel image Reference score.

In an embodiment of the present invention, the aforementioned skin feature includes a skin color, and the processor is configured to: for a first channel image in the channel image, calculate a skin color standard deviation of the skin color in the detection area, wherein The detection area does not include the facial features in the face image; and the skin color standard deviation is defined as the reference score of the first channel image.

In an embodiment of the present invention, each of the aforementioned channel images further includes a reference area, the skin feature includes a dark circle, and the processor is configured to: for a first channel image in the channel image, calculate the dark circle A first brightness in the detection area and a second brightness in the reference area; the first brightness is subtracted from the second brightness to generate a difference; and the difference is defined as the reference score of the first channel image.

In an embodiment of the present invention, the aforementioned channel image includes a first channel image, a second channel image, and a third channel image, and the processor is configured to: multiply the reference score of the first channel image by a first weight To generate a first value; multiply the reference score of the second channel image by a second weight to generate a second value; multiply the reference score of the third channel image by a third weight to generate a third value; The sum of the first value, the second value, and the third value is the detection score.

In an embodiment of the present invention, when the aforementioned skin feature is a dark circle or a non-smooth skin, the first channel image is a Y channel image, the second channel image is a Cr channel image, and the third channel image is A Cb channel image, and the first weight, the second weight, and the third weight are sequentially decreased.

In an embodiment of the present invention, when the skin feature is a wrinkle or a spot, the first channel image is an R channel image, the second channel image is a G channel image, and the third channel image is a B channel image. Image, and the first weight, second weight, and third weight are all 1.

，其中SC為第一校正後分數，RC為檢測分數，m為平均值，σ為標準差。In an embodiment of the present invention, the above-mentioned processor is configured to: calculate an average value and a standard deviation of historical detection scores; and input the detection scores into an excitation function to generate a first corrected score, where the excitation function is :

, Where SC is the first corrected score, RC is the detection score, m is the average, and σ is the standard deviation.

In an embodiment of the present invention, the above-mentioned processor is further configured to: generate a second corrected score corresponding to a second detection item based on the face image; multiply the first corrected score by a first specific Weighting to generate a first specific value; multiplying the second corrected score by a second specific weight to generate a second specific value; and summing the first specific value and the second specific value into a comprehensive skin test fraction.

Based on the above, the skin detection method and image processing device proposed by the present invention only need to be based on a single face image to perform skin detection for different detection items at the same time as required. In this way, a simple and time-saving skin detection mechanism can be provided.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail in conjunction with the accompanying drawings.

In summary, the image processing device and skin detection method proposed by the present invention can define different detection areas in the face image based on the considered one or more detection items after obtaining the face image. Afterwards, the skin feature corresponding to the detection item can be found in the detection area, and the detection score of the aforementioned face image in the considered detection item can be determined accordingly. Then, the present invention can also correct the detection score to a more objective and accurate corrected score through a certain mechanism. In this way, it is possible to easily complete skin detection at a lower cost through image processing without using a specific detection instrument. The detailed description will be given below.

Please refer to FIG. 1, which is a schematic diagram of an image processing device according to an embodiment of the present invention. In this embodiment, the image processing device 100 may be a server, a mobile phone, a smart phone, a personal computer (PC), a notebook PC, a netbook PC, a tablet computer ( tablet PC), or any electronic device with image capture function. In other embodiments, the image processing device 100 may be an electronic device dedicated to detecting skin, but it is not limited to this.

As shown in FIG. 1, the image processing device 100 includes an image capture circuit 102, a storage circuit 104 and a processor 106. The image capture circuit 102 can be any camera with a charge coupled device (CCD) lens, a complementary metal oxide semiconductor transistors (CMOS) lens, or an infrared lens, or it can be available The image capturing device of depth information, such as a depth camera or a stereo camera, does not depart from the scope of the embodiments of the present invention as long as the image capturing device used can capture images and obtain human images from them.

The storage circuit 104 is, for example, any type of fixed or removable random access memory (Random Access Memory, RAM), read-only memory (Read-Only Memory, ROM), flash memory (Flash memory), hard disk Disk or other similar devices or a combination of these devices can be used to record multiple codes or modules.

The processor 106 is coupled to the image capture circuit 102 and the storage circuit 104, and can be a general purpose processor, a special purpose processor, a traditional processor, a digital signal processor, multiple microprocessors, one or Multiple microprocessors, controllers, microcontrollers, Application Specific Integrated Circuits (ASIC), Field Programmable Gate Array (FPGA), any other integrated digital signal processor cores Types of integrated circuits, state machines, processors based on Advanced RISC Machine (ARM) and similar products.

In the embodiment of the present invention, the processor 106 can load the program code or module recorded in the storage circuit 104 to execute the skin detection method proposed by the present invention, which will be further described below.

Please refer to FIG. 2, which is a flowchart of a skin detection method according to an embodiment of the present invention. The method of this embodiment can be executed by the image processing device 100 in FIG. 1. The details of each step in FIG. 2 are described below in conjunction with the components shown in FIG. 1.

First, in step S210, the processor 106 can control the image capturing circuit 102 to obtain a face image. In this embodiment, the processor 106 can control the image capturing circuit 102 to obtain an image for a user, and perform a face recognition operation on the image to find the face image in the image.

Please refer to FIG. 3, which is a schematic diagram of obtaining a face image according to an embodiment of the present invention. As shown in FIGS. 1 and 3, in this embodiment, the image capturing circuit 102 can be controlled by the processor 106 to capture an image 310, and the image 310 includes, for example, a face image of a user whose skin is to be detected. For example, it includes a complete face image. Then, the processor 106 may perform a face recognition operation on the image 310 to find the face image 320 in the image 310. In one embodiment, the processor 106 can find a plurality of feature points 320a on the face in the image 310 through a face recognition operation, and define the recognition area corresponding to the face image 320 through these feature points 320a , Such as identifying areas such as the forehead, corners of the eyes, cheeks, chin, nose, lips, etc. in the face image. Therefore, even if the face in the image 310 is upside down, the processor 106 can still find the face image 320 based on the feature points 320a.

2 again, in step S220, the processor 106 may obtain a first mask according to the first detection item, and define a detection area in the face image based on the first mask. In different embodiments, the above-mentioned first detection item may be one of the detection items such as wrinkles, skin smoothness, skin shine, skin color uniformity, dark circles, spots, etc., but may not be limited thereto. In one embodiment, the image processing device 100 can provide the above-mentioned detection items for the user's reference through a user interface (not shown). Correspondingly, the user can select a desired detection item (for example, wrinkles) as the first detection item through the aforementioned user interface.

In different embodiments, because different detection items correspond to different detection areas in the face image, the processor 106 can define the required detection area in the face image through different masks, as follows: It will be further explained in conjunction with Figure 4.

Please refer to FIG. 4, which is a schematic diagram illustrating the detection area defined based on the masks of different detection items according to an embodiment of the present invention. In this embodiment, it is assumed that the masks 411, 412, 413, 414, and 415 correspond to detection items such as wrinkles, skin smoothness, skin oiliness, skin color uniformity, dark circles, etc., respectively.

As shown in FIG. 4, if the detection item is wrinkles, the corresponding detection area 411a is, for example, the forehead in the face image 400. In this case, the mask 411 selected by the processor 106 for the detection item (wrinkle) can cover all areas other than the forehead in the face image 400, leaving only the forehead as the detection area 411a. If the detection item is skin smoothness, the corresponding detection area 412 a is, for example, the cheeks in the face image 400. In this case, the mask 412 selected by the processor 106 for the detection item (skin smoothness) can cover all the areas other than the cheeks in the face image 400, leaving only the cheeks as the detection area 412a. If the detection item is shiny skin, the corresponding detection area 413a is, for example, the forehead, cheeks, nose, and chin in the face image 400. In this case, the mask 413 selected by the processor 106 for the detection item (skin shiny) can mask all the areas other than the cheeks in the face image 400 to define the detection area 413a. Based on the above teachings, it should be possible to deduce the mechanism by which the processor 106 defines the detection areas 414a and 415a based on the masks 414 and 415, respectively, which will not be repeated here.

In other embodiments, the detection area corresponding to each detection item in the face image can be adjusted according to the needs of the designer, and is not limited to the aspect shown in FIG. 4.

Please refer to FIG. 5, which is a schematic diagram illustrating the combination of a mask and a face image to define a detection area according to an implementation of the present invention. As shown in FIGS. 1, 4, and 5, in this embodiment, assuming that the first detection item considered is skin color uniformity, the processor 106 can find the corresponding detection item (skin color uniformity) accordingly. The first mask 510 is combined with the face image to generate the detection area image 520. As shown in FIG. 5, the first mask 510 can be used to cover the facial features (such as eyes, ears, nose, mouth, etc.) in the face image, so that only cheeks, nose, chin, etc. are left in the detection area image 520 The part is used as the detection area 520a, but it is only used as an example and is not used to limit the possible implementation of the present invention.

Please refer to FIG. 2 again. In step S230, the processor 106 may convert the face image into multiple channel images. In one embodiment, the processor 106 may determine the pre-processing mechanism and the color conversion mechanism according to the selected first detection item. The pre-processing mechanism includes, for example, image processing methods such as edge enhancement, blurring, grayscale, noise filtering, edge processing, and area pixel weighting, and the color conversion mechanism is, for example, RGB conversion or YCrCb conversion, but may not be limited thereto. . After that, the processor 106 can use the aforementioned color conversion mechanism to convert the face image into multiple color channel images, and use the aforementioned pre-processing mechanism to convert the aforementioned color channel image into multiple channel images.

For example, assuming that the first detection item considered is skin smoothness or dark circles, the corresponding color conversion mechanism is, for example, YCrCb conversion. Based on this, the processor 106 can convert the face image into multiple color channel images such as Y color channel image, Cr color channel image, and Cb color channel image, and can then convert these color channel images based on the required pre-processing mechanism. It is the aforementioned channel image, but the present invention is not limited to this.

For another example, assuming that the first detection item considered is skin color uniformity or wrinkles, the corresponding color conversion mechanism is, for example, RGB conversion. Based on this, the processor 106 can convert the face image into multiple color channel images such as R color channel image, G color channel image, and B color channel image, and can then convert these color channel images based on the required pre-processing mechanism. It is the aforementioned channel image, but the present invention is not limited to this.

For another example, assuming that the first detection item considered is spots or skin shiny, the processor 106 can use edge enhancement and blurring as the pre-processing mechanisms corresponding to the two, but the present invention is not limited to this. .

2 again, after obtaining multiple channel images through step S230, in step S240, the processor 106 finds the skin feature corresponding to the first detection item in the detection area of each channel image. Next, in step S250, the processor 106 calculates the reference score corresponding to the skin feature of the first detection item in the detection area of each channel image. For ease of description, it will be assumed below that the multiple channel images obtained by the processor 106 include a first channel image, a second channel image, and a third channel image, but the present invention may not be limited to this. Depending on the color conversion mechanism used, the first, second, and third channel images can be R channel images, G channel images, and B channel images, respectively, or Y channel images, Cr channel images, and Cb channel images, respectively. But it is not limited to this. Moreover, since the processor 106 has defined the detection area in the face image, the processor 106 can also define the detection area in the first, second, and third channels accordingly. Based on this, the processor 106 can respectively find the skin features corresponding to the first detection item in the detection areas defined by the first, second, and third channel images, and calculate the first, second, and third features accordingly. Individual reference scores for channel images. The relevant details are as follows.

In different embodiments, the above-mentioned skin characteristics will be different with the selected first detection item, and the way of calculating the reference score will also be different.

In the first embodiment, if the selected first detection item is wrinkles, the skin features found by the processor 106 in the detection regions of each channel image can be wrinkles. In this case, the processor 106 can, for example, find multiple edges in each channel image through edge detection, and define one or more of the edge intensity values higher than a predetermined edge intensity threshold as wrinkles.

For the first channel image (for example, the R channel image), the processor 106 may calculate the wrinkle area of the wrinkle in the detection area of the first channel image, and calculate the wrinkle area in the area of the detection area of the first channel image proportion. After that, the processor 106 can define this ratio as the reference score of the first channel image.

For the second channel image (eg, G channel image), the processor 106 may calculate the wrinkle area of the wrinkle in the detection area of the second channel image, and calculate the wrinkle area in the area of the detection area of the second channel image proportion. Afterwards, the processor 106 can define this ratio as the reference score of the second channel image.

For the third channel image (for example, the B channel image), the processor 106 may calculate the wrinkle area in the detection area of the third channel image, and calculate the wrinkle area in the area of the detection area of the third channel image proportion. Afterwards, the processor 106 may define this ratio as the reference score of the third channel image.

In the second embodiment, if the selected first detection item is a spot, the skin feature found by the processor 106 in the detection area of each channel image can be a spot. In this case, the processor 106 may perform binary large object (BLOB) detection on each channel image, for example. In one embodiment, in order to improve the performance of performing BLOB detection, the processor 106 may use sharpening as a pre-processing mechanism for converting each color channel image into a channel image, but it is not limited to this. After performing BLOB detection on each channel image to find the BLOB in each channel image, the processor 106 may further filter out skin areas that are not blobs based on the area, color, and circular deformation index of the BLOB. In this way, the processor 106 can find spots in the detection area of each channel image.

For the first channel image (for example, the R channel image), the processor 106 may calculate the spot area of the spot in the detection area of the first channel image, and calculate the area of the spot in the detection area of the first channel image proportion. After that, the processor 106 can define this ratio as the reference score of the first channel image. The calculation method of the individual reference scores of the second channel image (such as the G channel image) and the third channel image (such as the B channel image) is similar to the calculation method of the reference score of the first channel image, and will not be repeated here.

In the third embodiment, if the selected first detection item is skin smoothness, the skin feature found in the detection area by the processor 106 may be non-smooth skin. In this case, the processor 106 can find multiple edges in the detection area of each channel image by way of edge detection, so as to quantify the smoothness of the skin in the detection area. The following is a further description with Figure 6.

Please refer to FIG. 6, which is a schematic diagram of finding skin features in a channel image according to a third embodiment of the present invention. In this embodiment, it is assumed that the first detection item considered is skin smoothness, and the corresponding pre-processing mechanism and color conversion mechanism are, for example, noise filtering and YCrCb conversion, respectively. In this case, after the processor 106 obtains a face image (not shown), the processor 106 may first convert the face image into multiple color channel images (ie, Y color channel image, Cr color channel image, etc.) based on YCrCb conversion. Channel image and Cb color channel image), and the Y color channel image is, for example, the color channel image 610 in FIG. 6.

In FIG. 6, the processor 106 can find the detection area 610 a based on the mask corresponding to the skin smoothness (the first detection item), and perform noise filtering on it to generate the channel image 620. After that, the processor 106 can find multiple edges in the detection area 610a of the channel image 620 through edge detection, so as to quantify the smoothness of the skin in the detection area 610a. After that, the processor 106 may define an edge with a higher edge strength as a non-smooth skin, but the present invention may not be limited to this.

After that, for the first channel image (for example, the Y channel image), the processor 106 may calculate the non-smooth area of the non-smooth skin in the detection area of the first channel image, and calculate the detection of the non-smooth area in the first channel image. A proportion of the area of a region. Then, the processor 106 can define this ratio as the reference score of the first channel image. The calculation method of the reference scores of the second channel image (such as the Cr channel image) and the third channel image (such as the Cb channel image) is similar to the calculation method of the reference score of the first channel image, and will not be repeated here.

In the fourth embodiment, if the selected first detection item is shiny skin, the skin feature found by the processor 106 in the detection area can be shiny skin. For the first channel image, the processor 106 may calculate the first brightness of the skin oily light (the first detection item) in the detection area of the first channel image, and the second brightness of the skin oily light in the reference area. In this embodiment, the aforementioned reference area is, for example, a full face area, but it may not be limited to this. After that, the processor 106 subtracts the second brightness from the first brightness to generate a difference value, and defines the difference value as the reference score of the first channel image. The calculation method of the reference score of the second channel image and the third channel image is similar to the calculation method of the reference score of the first channel image, and will not be described here.

In the fifth embodiment, if the selected first detection item is skin color uniformity, the skin feature found by the processor 106 in the detection area can be skin color. For the first channel image (for example, the R channel image), the processor 106 may calculate the skin color standard deviation of the skin color (the first detection item) in the detection area, and use the skin color standard deviation as the reference score of the first channel image. The calculation method of the reference scores of the second channel image (such as the G channel image) and the third channel image (such as the B channel image) is similar to the calculation method of the reference score of the first channel image, and will not be described here.

In the sixth embodiment, if the selected first detection item is dark circles, the skin feature found by the processor 106 in the detection area may be dark circles. For the first channel image (for example, the Y channel image), the processor 106 may calculate the first brightness of the dark circles (the first detection item) in the detection area of the first channel image and the reference area (for example, the full face area). The second brightness. Then, the processor 106 may subtract the first brightness from the second brightness to generate a difference value, and define the difference value as the reference score of the first channel image. The calculation method of the individual reference scores of the second channel image (such as the Cr channel image) and the third channel image (such as the Cb channel image) is similar to the calculation method of the reference score of the first channel image, and will not be repeated here.

After calculating the reference scores of the first, second, and third channel images based on the above teachings, the processor 106 may calculate the reference scores of the face images in the first detection item according to the reference scores corresponding to the skin features of the first detection item in step S260. Detection score.

In an embodiment, the processor 106 may be configured to: multiply the reference score of the first channel image by the first weight to generate a first value; and multiply the reference score of the second channel image by the second weight to generate the second value. Numerical value; multiplying the reference score of the third channel image by the third weight to generate a third value; and adding the first value, the second value, and the third value to the detection score. In short, the processor 106 may add the weighted reference scores of the first, second, and third channel images to the detection scores of the face images in the first detection item.

In different embodiments, the above-mentioned first weight, second weight, and third weight may be determined according to the selected first detection item. For example, if the first detection item is dark circles or skin smoothness, the corresponding first channel image, second channel image, and third channel image are, for example, Y channel image, Cr channel image, and Cb channel image, respectively. In this case, the first weight, the second weight, and the third weight can be set to decrease sequentially. That is, the processor 106 may use the highest weight for the Y channel image and the lowest weight for the Cb channel image, but the invention is not limited to this.

For another example, if the first detection item is wrinkles or spots, the corresponding first channel image, second channel image, and third channel image are, for example, R channel image, G channel image, and B channel image, respectively. In this case, the first weight, the second weight, and the third weight can all be set to 1, but the invention is not limited to this.

In different embodiments, since the quality of the face image may change due to the brightness at the time of shooting or other environmental factors, the obtained detection score may not objectively and accurately reflect the skin condition of the user. Therefore, the present invention further proposes the following mechanism to appropriately correct the detection score, so as to obtain a more objective and accurate skin detection result.

Specifically, after obtaining the detection score of the face image in the first detection item, the processor 106 may perform step S270 to calculate the first corrected score corresponding to the first detection item according to the multiple historical detection scores and the detection scores. In this embodiment, the aforementioned historical detection scores are, for example, multiple detection scores calculated based on other face images taken in various environments, but it is not limited to this.

表示），並將人臉影像在第一檢測項目的檢測分數輸入至一激勵函數以產生第一校正後分數。在本實施例中，所述激勵函數例如是：

其中SC為第一校正後分數，RC為人臉影像在第一檢測項目的檢測分數。In one embodiment, the processor 106 may calculate an average value (represented by m) and a standard deviation (represented by

Represents), and input the detection score of the face image in the first detection item into an excitation function to generate the first corrected score. In this embodiment, the excitation function is, for example:

Where SC is the first corrected score, and RC is the detection score of the face image in the first detection item.

After the detection score (ie, RC) of the first detection item is converted into the first corrected score (ie, SC) through the above-mentioned excitation function, the first corrected score can be used to present the skin detection result more accurately and objectively. And to eliminate the detection errors caused by different environmental factors (for example, lighting).

It can be seen from the above that the method and device proposed by the present invention only need to be based on a single face image to perform skin detection for different detection items at the same time as required, instead of separately targeting each partial face area as in the conventional practice. Take measurements. In this way, a simple and time-saving skin detection mechanism can be provided.

In one embodiment, if the user selects more than two detection items at the same time, the processor 106 may calculate the corresponding corrected scores based on these detection items, and provide the corrected scores corresponding to these detection items through the user interface. For the user's reference.

In addition, the processor 106 may further generate a comprehensive skin detection score based on the corrected score of each detection item to comprehensively characterize the user's skin detection result (for example, skin type and age).

In one embodiment, after the processor 106 generates the first corrected score corresponding to the first detection item according to the teaching in the previous embodiment, the processor 106 may then generate a second detection score corresponding to the same face image. The second corrected score of the item. Thereafter, the processor 106 may be configured to: multiply the first corrected score by a first specific weight to generate a first specific value; multiply the second corrected score by a second specific weight to generate a second specific value; and The sum of the first specific value and the second specific value is the aforementioned comprehensive skin test score. In short, the processor 106 may add the weighted first corrected score and the second corrected score to a comprehensive skin detection score to comprehensively characterize the user's skin detection result.

In addition, although the above embodiment only describes the calculation method of the comprehensive skin detection score based on two detection items, in other embodiments, the processor 106 can also be based on the corrected scores of more detection items (different detection items can correspond to Based on different specific weights) to calculate the comprehensive skin detection score. In an embodiment, if the selected detection item includes wrinkles, the processor 106 may set the corresponding specific weight to be the highest, but the present invention is not limited to this.

Please refer to FIG. 7, which is a schematic diagram of a skin detection result according to an embodiment of the present invention. FIG. 7 is, for example, a screen displayed by the user interface of the image processing device 100, which provides corrected scores 711, 712, 713, 714, and 715 corresponding to different detection items, and is calculated based on the corrected scores 711 to 715 The comprehensive skin test score of 720. In other embodiments, the image processing device 100, for example, further includes a display unit (not shown), coupled to the processor 106 (as shown in FIG. 1), for displaying the corresponding items corresponding to different detection items as shown in FIG. The screen of the corrected scores 711, 712, 713, 714, 715, and the comprehensive skin detection score 720 calculated based on the corrected scores 711 to 715.

In summary, the skin detection method and image processing device proposed by the present invention can perform skin detection for different detection items at the same time based on a single face image, and generate corrected scores corresponding to different detection items. . In this way, it is possible to easily complete skin detection at a lower cost through image processing without using a specific detection instrument.

In addition, since the corrected scores of each test item are converted by a specific function that considers multiple historical test scores, the skin test results can be presented more objectively and accurately. In addition, the present invention can also generate a comprehensive skin detection score based on the corrected score of each detection item to comprehensively characterize the user's skin detection result.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the relevant technical field can make some changes and modifications without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be subject to those defined by the attached patent application scope.

100: Image processing device 102: Image capture circuit 104: Storage circuit 106: Processor 310: Image 320: Face image 320a: Feature point 400: Face image 411, 412, 413, 414, 415: Mask 411a, 412a, 413a, 414a, 415a, 520a, 610a: detection area 510: first mask 520: detection area image 610: color channel image 620: channel image 711, 712, 713, 714, 715: corrected score 720: comprehensive Sexual skin test scores S210~S270: steps

FIG. 1 is a schematic diagram of an image processing device according to an embodiment of the invention. Fig. 2 is a flowchart of a skin detection method according to an embodiment of the present invention. FIG. 3 is a schematic diagram of obtaining a face image according to an embodiment of the present invention. FIG. 4 is a schematic diagram illustrating a detection area defined based on masks of different detection items according to an embodiment of the present invention. FIG. 5 is a schematic diagram illustrating a combination of a mask and a face image to define a detection area according to an implementation of the present invention. FIG. 6 is a schematic diagram of finding skin features in channel images according to the third embodiment of the present invention. Fig. 7 is a schematic diagram of a skin detection result according to an embodiment of the present invention.

S210~S270: steps

Claims (26)

A skin detection method includes: obtaining a face image; obtaining a first mask according to a first detection item, and defining a detection area in the face image based on the first mask, wherein the first detection item includes At least one of wrinkle detection, skin smoothness detection, skin shine detection, skin color uniformity detection, dark circle detection, spot detection; convert the face image into multiple channel images; in the detection area of each of the channel images Find a skin feature corresponding to the first detection item; calculate a reference score corresponding to the skin feature of the first detection item in the detection area of each of the channel images; according to the skin of the first detection item The reference score corresponding to the feature calculates a detection score of the face image in the first detection item; and calculates a first corrected score corresponding to the first detection item according to a plurality of historical detection scores and the detection score. For the method described in item 1 of the scope of patent application, the step of obtaining the face image includes: obtaining an image, and performing a face recognition operation on the image to find the face image in the image. For the method described in item 1 of the scope of patent application, the step of converting the face image into the channel images includes: Determine a pre-processing mechanism and a color conversion mechanism according to the first detection item; use the color conversion mechanism to convert the face image into multiple color channel images; and use the pre-processing mechanism to convert the color channel images into the Some channel images. The method according to claim 1, wherein the skin feature includes a wrinkle, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each of the channel images includes: For a first channel image in the image, calculate a wrinkle area of the wrinkle in the detection area; calculate a ratio of the wrinkle area to the area of the detection area; and define the ratio as the first The reference score of the channel image. The method according to claim 1, wherein the skin feature includes a non-smooth skin, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each of the channel images includes: For a first channel image in the channel images, calculate a non-smooth area of the non-smooth skin in the detection area; calculate a ratio of the non-smooth area to the area of the detection area; and define the ratio Is the reference score of the first channel image. According to the method described in claim 1, wherein each of the channel images further includes a reference area, the skin feature includes a skin shiny, and the skin feature corresponding to the first detection item in each of the channel images is calculated The steps of this reference score include: For a first channel image among the channel images, calculate a first brightness of the skin oily light in the detection area and a second brightness of the skin oily light in the reference area; subtract the first brightness Removing the second brightness to generate a difference; and defining the difference as the reference score of the first channel image. The method according to claim 1, wherein the skin feature includes a skin color, and the step of calculating the reference score corresponding to the skin feature of the first detection item in each of the channel images includes: For a first channel image in the image, calculate a skin color standard deviation of the skin color in the detection area, where the detection area does not include the facial features in the face image; and define the skin color standard deviation as the first The reference score of the channel image. According to the method described in claim 1, wherein each of the channel images further includes a reference area, the skin feature includes a dark circle, and the skin feature corresponding to the first detection item in each of the channel images is calculated The step of the reference score includes: for a first channel image of the channel images, calculating a first brightness of the dark circle in the detection area and a second brightness of the reference area; The first brightness is subtracted from the second brightness to generate a difference; and the difference is defined as the reference score of the first channel image. For the method described in claim 1, wherein the channel images include a first channel image, a second channel image, and a third channel image, and according to the first channel image The step of calculating the detection score of the face image in the first detection item by the reference score corresponding to the skin feature of the detection item includes: multiplying the reference score of the first channel image by a first weight to generate a first weight A value; multiply the reference score of the second channel image by a second weight to generate a second value; multiply the reference score of the third channel image by a third weight to generate a third value; and The first value, the second value, and the third value are added to form the detection score. The method according to claim 9, wherein when the skin feature is a dark circle or a non-smooth skin, the first channel image is a Y channel image, and the second channel image is a Cr channel image, The third channel image is a Cb channel image, and the first weight, the second weight, and the third weight are sequentially decreased. The method according to claim 9, wherein when the skin feature is a wrinkle or a spot, the first channel image is an R channel image, the second channel image is a G channel image, and the third channel image is The channel image is a B channel image, and the first weight, the second weight, and the third weight are all 1. For the method described in claim 1, wherein the step of calculating the first corrected score based on the historical detection scores and the detection score includes: calculating an average value and a standard deviation of the historical detection scores; The detection score is input to an excitation function to generate the first corrected score, where the excitation function is:

, Where SC is the first corrected score, RC is the detection score, m is the average, and σ is the standard deviation. The method described in item 1 of the scope of patent application further includes: generating a second corrected score corresponding to a second detection item based on the face image; and multiplying the first corrected score by a first specific weight To generate a first specific value; multiply the second corrected score by a second specific weight to generate a second specific value; and sum the first specific value and the second specific value to form a comprehensive skin Detection score. An image processing device includes: an image capturing circuit; a storage circuit storing a plurality of modules; and a processor, coupled to the image capturing circuit and the storage circuit, accessing the modules to perform the following steps : Controlling the image capturing circuit to obtain a face image; obtaining a first mask according to a first detection item, and defining a detection area in the face image based on the first mask, wherein the first detection item includes Wrinkle detection, skin smoothness detection, skin shine detection, skin color uniformity detection, At least one of dark circle detection and spot detection; converting the face image into a plurality of channel images; finding a skin feature corresponding to the first detection item in the detection area of each of the channel images; calculating each A reference score corresponding to the skin feature of the first detection item in the detection area of the channel images; calculating the face image in the first detection item according to the reference score corresponding to the skin feature of the first detection item A detection score of a; a first corrected score corresponding to the first detection item is calculated based on a plurality of historical detection scores and the detection score. For the image processing device described in claim 14, wherein the processor is configured to obtain an image and perform a face recognition operation on the image to find the face image in the image. For example, the image processing device of claim 14, wherein the processor is configured to: determine a pre-processing mechanism and a color conversion mechanism according to the first detection item; use the color conversion mechanism to convert the face image Is a plurality of color channel images; and using the pre-processing mechanism to convert the color channel images into the channel images. The image processing device according to claim 14, wherein the skin feature includes a wrinkle, and the processor is configured to: For a first channel image of the channel images, calculate a wrinkle area of the wrinkle in the detection area; calculate a ratio of the wrinkle area to the area of the detection area; and define the ratio Is the reference score of the first channel image. The image processing device according to claim 14, wherein the skin feature includes a non-smooth skin, and the processor is configured to: for a first channel image among the channel images, calculate the non-smooth skin A non-smooth area of the skin in the detection area; calculating a proportion of the non-smooth area in the area of the detection area; and defining the proportion as the reference score of the first channel image. The image processing device according to claim 14, wherein each of the channel images further includes a reference area, the skin feature includes a skin shine, and the processor is configured to: for one of the channel images For the first channel image, calculate a first brightness of the skin oil in the detection area and a second brightness of the skin oil in the reference area; subtract the second brightness from the first brightness to generate a Difference; and define the difference as the reference score of the first channel image. The image processing device according to claim 14, wherein the skin feature includes a skin color, and the processor is configured to: For a first channel image among the channel images, calculate a skin color standard deviation of the skin color in the detection area, where the detection area does not include the facial features in the face image; and define the skin color standard deviation Is the reference score of the first channel image. The image processing device according to claim 14, wherein each of the channel images further includes a reference area, the skin feature includes a dark circle, and the processor is configured to: for one of the channel images For the first channel image, calculate a first brightness of the dark circle in the detection area and a second brightness of the reference area; subtract the first brightness from the second brightness to generate a difference; and The difference is defined as the reference score of the first channel image. For the image processing device described in claim 14, wherein the channel images include a first channel image, a second channel image, and a third channel image, and the processor is configured to: The reference score is multiplied by a first weight to generate a first value; the reference score of the second channel image is multiplied by a second weight to generate a second value; the reference score of the third channel image is multiplied A third weight is used to generate a third value; the first value, the second value, and the third value are added to form the detection score. The image processing device according to claim 22, wherein when the skin is characterized by a dark circle or a non-smooth skin, the first channel image is a Y channel image, and the second channel image is a Cr channel Image, the third channel image is a Cb channel image, and the first weight, the second weight, and the third weight are sequentially decreased. For example, the image processing device of claim 22, wherein when the skin feature is a wrinkle or a spot, the first channel image is an R channel image, the second channel image is a G channel image, and the The third channel image is a B channel image, and the first weight, the second weight, and the third weight are all 1. For the image processing device described in claim 14, wherein the processor is configured to: calculate an average value and a standard deviation of the historical detection scores; and input the detection scores into an excitation function to generate the first A corrected score, where the excitation function is:

, Where SC is the first corrected score, RC is the detection score, m is the average, and σ is the standard deviation. The image processing device according to claim 14, wherein the processor is further configured to: generate a second corrected score corresponding to a second detection item based on the face image; Multiplying the first corrected score by a first specific weight to generate a first specific value; multiplying the second corrected score by a second specific weight to generate a second specific value; and the first specific weight The sum of the value and the second specific value is a comprehensive skin test score.

Images