KR101025651B1 - Apparatus and method for identifying video object - Google Patents

Abstract

The present invention relates to an image object recognition apparatus and method for identifying information of another image by storing information representing a specific object in an image.

The present invention stores the Y-weights of the pixels of the specific area of the object, obtains the Y-weights of the pixels of the specific area of the object to be compared, and compares them with each other to identify the object.

Therefore, according to the present invention, when watching a movie, such as "where the main character came from" or when you are curious about the scenes you have seen before, you can check immediately.

       Image, Pixel, Weight, Comparison

Description

Apparatus and method for identifying video object}

1 is a block diagram of one embodiment of the configuration of a display means in which the present invention is implemented;

2 is a flow diagram of one operational embodiment according to the present invention.

3 is a flowchart of another embodiment of the present invention.

4 is another flowchart illustrating performing image object recognition according to the present invention.

FIG. 5 is a flowchart illustrating obtaining and storing the pixel Y-weight of a region in step S405 of FIG. 4.

FIG. 6 is a diagram illustrating a state in which a Y-weight of each pixel of a specific object / area selected by a user and program information related thereto are stored in a memory; FIG.

FIG. 7 is a diagram illustrating an example in which a Y-weight of a corresponding object / area is calculated based on a user's selection while viewing a new image, and is output through comparison with data stored as shown in FIG. 6.

The present invention relates to an image object recognition apparatus and method for identifying information of another image by storing information representing a specific object in an image.

More specifically, the Y-weight of the pixel of the specific area of the object is stored, the Y-weight of the pixel of the specific area of the object to be compared is obtained, and the object is identified by comparison with each other.

Hereinafter, related art will be described.

In general, while watching a movie or a program on TV, the memory of a certain object (person, object, etc.) is often not clear, in which case there was no means to solve the problem when thinking about it. .

There is no way to find out immediately if you can't remember which movie appeared on the air or on a DVD, Blue-Ray or Divx movie.

An object of the present invention is to solve a problem when a movie or a program is watched on an output means such as TV, when the memory of a certain object (person, object, etc.) is not clear.

For example, if you are watching a movie, etc., in case you are wondering, "Where did the main character come from?" It is intended to provide convenience to them. That is, when you are curious about who the person is or what other movie or drama you have, you can find it immediately.

While the video stream is being displayed, the video screen is divided into pixel units (1024x1080), and a block (for example, a pupil or a face) that maintains a certain color and shape at all times (for example, a face of a person) In the case of selecting a pixel of the outline) and storing the Y (Luminance) -weight value of the selected pixel in a memory, and wanting to check an object as seen from other media or broadcasting, the stored Y-weight It is proposed to recognize the object (target) by reading the values and comparing the Y-weight values of the current target stream with each other.

In order to achieve the above object, the image object recognition apparatus according to the present invention includes an output unit for displaying an image; A memory unit for storing a Y-weight associated with an image output from the output unit; An input unit for inputting a user command for deriving the Y-weight; And a controller for deriving a Y-weight of another image corresponding to the image and recognizing an object through comparison with the stored Y-weight.

In addition, the image object recognition method according to the present invention, the step of outputting an image; Selecting a target from an output image; Selecting an area in the object; Calculating a Y-weight of a plurality of pixels constituting the area; Storing information related to the Y-weight and the image; It works including.

In addition, the image object recognition method according to the present invention, the step of outputting the first image; Selecting a target from an output image; Selecting an area in the object; Calculating a Y-weight of a plurality of pixels constituting the area; Storing information related to the Y-weight and the image; Selecting a target of the second image; Selecting an area in the object; Calculating a Y-weight of a plurality of pixels constituting the area of the second image; Comparing the Y-weight of the first image with the Y-weight of the second image; And recognizing an object through the comparison; It works including.

Hereinafter, an image object recognition apparatus and method according to the present invention will be described in detail with reference to the accompanying drawings.

First, the terms used in the present invention were selected as general terms widely used at present, but in some cases, the terms arbitrarily selected by the applicant, which are described in detail in the description of the corresponding invention, the simple terms, It is to be noted that the present invention should be understood as an operation / meaning of a term other than a name of a.

For example, the Y-weight referred to in the present invention represents Luminance (color signal) as a combination of R, G and B as follows, Y (luminance signal, luminance level) = KrR + KgG + KbB), In this case, the sum of absolute values of each K factor is defined as Y-weight.

In addition, the Kr, Kg, and Kb values represent weight factors for the respective RGB signals, and the reason why the weighting factors in each pixel are different is because the Y (Luminance) value of each pixel is different. In order to express using B, the coefficients Kr, Kg, and Kb before R, G, and B must be different from each other.

In addition, although the embodiment of the present invention has been described with reference to the image output from the display means, it is also applicable to an object on paper, a photograph, and the like.

In addition, the object or object referred to in the present invention may have the same meaning.

1 is a block diagram 100 of an embodiment of a configuration of a display means in which the present invention is implemented.

As shown in FIG. 1, an output unit 110 on which a stream video is displayed, and a memory in which the Y-weight is automatically stored according to a user selection / or object recognition function setting for the image output from the output unit. The unit 120 and an input unit for inputting a user command for deriving the Y-weight of the selected or selected object, respectively, while viewing the video and other images, and comparing the derived Y-weights to recognize the object. And a control unit 130 to control to operate.

In addition to the above configuration, a processing unit 140 for processing data, a receiving unit 170 for receiving a remote control command or an internal / external signal, and a transmitting unit 150 for transmitting a signal inside / outside are included.

2 is a flow diagram of one operational embodiment according to the present invention.

As shown in FIG. 2 (a), a face 200 of a human is selected as an object among stream images output from an output means (not shown) having a predetermined resolution (pixel), and among the selected objects, as a specific region. Eye 210 was selected. (You can choose your pupil if needed).

That is, in the stream, the human eye is set as an object / object and the Y-weight value of the corresponding pixel is calculated. (S 201).

The calculated Y-weight value is stored in the memory unit 120 of FIG. 1. (S 203).

The name and appearances of the person are mapped to the stored Y-weight value as shown in FIG. 6. (S 205).

If the Y-weight value is calculated in other movies, broadcasts, and programs in the future, it is checked whether the same object (person) is compared with the stored data. (S 207).

On the other hand, when it is difficult to compare only by the Y-weight value derived based on the method as shown in FIG. 2 (a), as shown in FIG. 2 (b), the pixel of the object determined according to FIG. 2 (a) It can be raised to x2, x4, x8, etc. (220) to increase the discrimination power, it is possible to calculate the Y-weight value again. (S 209, S 211). The subsequent steps are the same as S 205 and S 207.

2 (a) and 2 (b), when the accuracy of comparison is inferior, both Y-weight values of the two eyes 210 and 230 are calculated, and the two values are stored as average data.

Therefore, accuracy can be improved as it is the average value of two data. (S 213).

The subsequent steps are the same as S 205 and S 207.

In order to clarify the understanding of the present invention, related contents according to the implementation of the present invention as described above are described as follows.

As described above, the Y-weight is used to distinguish the area of each object.

In general, any luminance (color signal) can be represented by a combination of R, G, and B.

Y = KrR + KgG + KbB, where Kr, Kg, and Kb values are weight factors for each RGB signal, and the sum of the absolute values of each K factor is called Y-weight.

In addition, the Y-weight values for all pixels in a video stream of one frame may be obtained by the processor (video processor) of FIG. 1. That is, each pixel can be assigned a unique Y-weight value.

Therefore, you can get the Y-weight value for a total of 64 pixels by catching pixels in a specific block area

The sum of the Y-weight values for all the pixels of the block can be obtained using the above values. Therefore, since a unique value will be calculated, it may be possible to distinguish or determine the unique Y-weight value.

In other words, when the resolution of the displayed screen is 1024x1080p, the total number of pixels (pixels) is 1024x1080. Each Y-weight value of all pixels may be obtained.

However, because it calculates for too many pixels, it can be decided by the user, or it can be calculated from the beginning at 8x8 (8by8), 16x16 (16by16). For example, if 8by8 is set, 64 pixels are included.

At this time, (1) the Y-weight value can be obtained for each pixel, and after that, each of the 64 Y-weight values is added and the total Y-weight value of the block designated by 8by8 is obtained. Recognizes the object by comparing the total or average of each pixel value)

Meanwhile, (2) each of the 64 Y-weight values can be obtained and read in order, and the values can be stored sequentially. So you can find the new 8by8 block in turn and compare it next. (The object is recognized based on the order of each 64 pixel values)

The reason for selecting the comparison target in the present invention described above is as follows.

In conclusion, one of the most discriminating parts of the human form is selected, which may vary from user to user.

However, for example, I thought that the most unique part of the character was the eye. I think it is unique because each person looks different and the color is different. Because the pupil alone is a very large range when captured with pixel blocks, the Y-weight values will usually converge to the same value.

As another example, each person's face outline is also different, so if you only get the Y-weight value by drawing the pixel that corresponds to the face outline, it will also be unique.

If you want to be able to distinguish the objects more clearly, if you remove all color signal components from the screen to minimize the problems caused by lighting and makeup, only the luminance (Y) signal components remain. Since all the colors have been removed, the color classification due to makeup, etc. disappears, leaving only the eye shape itself. Finding the Y-weight value will be a more reliable way to distinguish.

In the case of eyes / eyes, it may be difficult to distinguish between displays with low resolution, but in case of face outlines, the distinction is clearly distinguished, and the distinction may be clear because the chrominance value between the face color and the non-face color is clearly different.

Therefore, when selecting an object, the side of the eye and the side of the ear may be covered by hair, etc., so it is a good way to compare it using the Y-weight of the face outline, which is lower than the nose.

Meanwhile, as another embodiment of the present invention, when the cells are zoomed in hundreds of times, adjacent cells are almost the same color. So even if the weighting factor gives similar values, the Y-weight is not distinguished. In this case, if you do not compare with each pixel, but select and compare only one pixel every 10 pixels, it may be more discriminating.

3 is a flowchart of another embodiment of the present invention.

The whole screen frame is divided into pixel units. (S 301).

The Y-weight value of each unit pixel is calculated based on Equation 1 above. (S 303).

Set the human eye as an object in the stream and store the Y-weight for the eye block in the memory. (S 305, S 307). If the user wants to save the Y-weight for another object, it is calculated and stored repeatedly. (S 309).

After the above operation, when the user finds the same person in different broadcasts, the Y-weight value of the stream pixel block selected as the Object is calculated and compared with the Y-weight value stored in the above step, Present the data corresponding to the Y-weight value or a plurality of close Y-weight value, so that the user can make a final decision. (S 311-S 319).

Of course, you can also display the target directly.

4 is another flowchart illustrating performing image object recognition according to the present invention.

The first image is output from the display means (FIG. 110) outputting a predetermined resolution. (S 401).

In the image, a specific region (eg, an eye / face outline) of a specific object (eg, a person) is selected. (S 403). Each Y-weight of the pixels of the whole or predetermined portion 8by8 of the specific region is calculated based on Equation 1 above. (S 405).

The derived Y-weight and information related to the image of the value (cast program content, actor name, etc.) are stored in the memory unit of FIG. (S 407).

Then, when a specific region is selected from the second image different from the first image (S 409), the operation of step S 405 is performed to obtain the Y-weight of the specific region of the second image to obtain Y of the stored first image. Compare to -weight. (S 411). The object information of the first image corresponding to the error preset by the user is displayed. (S 413).

The user then decides. (S 415).

FIG. 5 is a flowchart illustrating obtaining and storing the pixel Y-weight of the area in step S405 of FIG. 4.

Among the above areas, the Y-weights of some pixels selected by the user may be obtained and added (S501), or the Y-weights of each pixel may be obtained and summed to obtain Y-weights for the entire area. (S 503). The sums obtained may be compared, or values may be stored in order one by one, and each object in two images may be compared based on the order.

FIG. 6 is a diagram illustrating a state in which a Y-weight of each pixel of a specific object / region selected by a user and program information related thereto are stored in a memory.

FIG. 7 illustrates an embodiment in which a Y-weight of a corresponding object / area is calculated based on a user's selection while viewing a new image, and is output through comparison with data stored as shown in FIG. 6. Here, the object may be displayed immediately, or the object may be selected by the user by showing several objects within an error range.

As described above, according to the present invention, while the video stream is being displayed, the video screen is divided into pixel units (1024x1080), and an area of a specific object (for example, a constant color and shape among faces of a person is always present). Select the pixel of the block (e.g., the outline of the eyes or face) to keep, and check the object that seems to be seen from other media or broadcasting while storing the Y (Luminance) -weight value of the selected pixel in memory. In order to recognize the object (target), the stored Y-weight value is called and the Y-weight value of the current target stream is compared with each other.

As mentioned above, preferred embodiments of the present invention described above are disclosed for the purpose of illustration, and those skilled in the art can improve other various embodiments within the spirit and technical scope of the present invention disclosed in the appended claims below. Changes, substitutions or additions will be possible.

According to the present invention, when watching a movie, such as, "Where did the main character come from?" Or when you are curious about the scenes you have seen before, you can check immediately.

As a result, the user's knowledge can be improved and the output image can be effectively enjoyed.

Claims (14)

Outputting an image to an output unit; Selecting at least one object from the output image; Selecting at least one area of the object; Calculating a Y-weight of a plurality of pixels constituting the area; And Storing information representing the Y-weight and the image; Including; In the following formula, all Luminance (color signals) are represented by a combination of R, G, and B, where Y = KrR + KgG + KbB and Y-weight is the same as the sum of the absolute values of each K Factor. A video object recognition method, characterized in that defined as weight. Outputting a first image to an output unit; Selecting at least one object from the output image; Selecting at least one area of the object; Calculating a Y-weight of a plurality of pixels constituting the area; Storing information representing the Y-weight and the image; Selecting an object corresponding to the object of the first image from the second image output to the output unit; Selecting an area corresponding to a target area of the first image from the target; Calculating a Y-weight of a plurality of pixels constituting the area of the second image; Comparing the Y-weight of the first image with the Y-weight of the second image; And      Recognizing an object through the comparison; Including; In the following formula, all Luminance (color signals) are represented by a combination of R, G, and B, where Y = KrR + KgG + KbB and Y-weight is the same as the sum of the absolute values of each K Factor. A video object recognition method, characterized in that defined as weight. The image object recognition method according to claim 1 or 2, wherein the Y-weight calculation of the pixels constituting the area is performed for all pixels of the area. The image object recognition method according to claim 1 or 2, wherein the Y-weight calculation of the pixels constituting the area is performed on at least one or more pixels of all pixels in the area. The image object recognition method according to claim 2, wherein when the calculated and stored Y-weight is difficult to recognize, the object or pixel is enlarged by a predetermined multiple to derive the Y-weight to perform recognition. The image object according to claim 2, wherein when it is difficult to recognize the calculated and stored Y-weight, the image object is performed by using the derived Y-weight of another object / pixel corresponding to the object or pixel. Recognition method. delete The image object recognition method of claim 1, wherein the Kr, Kg, and Kb values are weight factors for respective RGB signals. The method of claim 8, wherein the weight factors in each pixel are different because the Y (Luminance) value of each pixel is different. Image object recognition method characterized in that the coefficients are different from each other. The image object recognition method according to claim 1 or 2, wherein the storing of the information representing the image includes at least one of a performer's name or information on the work to be performed. The image object recognition method of claim 2, wherein the user is able to select a plurality of pieces of target information by comparing the Y-weight of the first image with the Y-weight of the second image. The image object recognition method of claim 2, wherein when comparing the Y-weight of the first image and the Y-weight of the second image, the target information is displayed by comparing the total sum of the corresponding pixels. The image object recognition of claim 2, wherein when comparing the Y-weight of the first image and the Y-weight of the second image, target information is displayed based on the Y-weight order of the corresponding pixels. Way. An output unit for displaying an image; A memory unit for storing a Y-weight indicating an image output from the output unit; An input unit for inputting a user command for deriving the Y-weight; And a controller configured to derive a Y-weight of another image corresponding to the image and to recognize an object through comparison with the stored Y-weight. In the following expressions, all Luminance (color signals) are a combination of R, G, and B, where Y = KrR + KgG + KbB and Y-weight is the same as the sum of the absolute values of each K Factor. A video object recognition device, characterized in that defined as weight.

Images