KR102081752B1 - Video evaluation system and viedo evaluation method - Google Patents

Abstract

The present invention relates to a system for evaluating an image and a method thereof. According to the present invention, disclosed is method for receiving section information, which intends specific emotion, from an image producer and quantifying and providing specific emotion inputted during a corresponding section from a subject watching a corresponding image. According to the present invention, the image produced by the producer may be objectively evaluated.

Description

VIDEO EVALUATION SYSTEM AND VIEDO EVALUATION METHOD}

The present invention relates to an image evaluation system and an image evaluation method, and more particularly, to an image evaluation system and an image evaluation method for objectively providing a score of a match between the intention of an image producer and the emotions of a subject who has viewed the image. .

Recently, the video trend of the entertainment field including the game field is a field using VR (Augmented Reality) and AR (Augmented Reality) video. Since the production of VR and AR takes considerable time and expense, it is hoped that the reaction of the video-experiencers will appear as the planner intended, but it is difficult to accurately measure the reaction of the subjects until the video is released.

The video producer wants the viewer to see the produced video and show the reaction as intended. For example, in the case of horror films, producers think that viewers will be the most surprised at 35 minutes to 1 minute, and they set up the sounds and follow-up scenes needed at that time. As another example, in the case of a TV commercial image, you may want to show an image of a product that is intrigued and sold at the highest point of inquiries. Most of the image evaluation methods used in the past have been conducted in the form of a questionnaire in the form of a questionnaire after the subject has watched the content to the end. This method has a problem in that the subject may not actually remember every moment, and because the subject may not know exactly the feelings felt by the subject, but is expressed as an ambiguous answer rather than an exact answer.

Korean Patent Publication No. 10-2014-0004740 (published Jan. 13, 2014)

An object of the present invention is to provide an image evaluation system and method for objectively expressing an emotion of an experienced person who reacts while watching an image.

The object of the present invention is to quantify the extent to which the subject's response is consistent with the intention of the producer providing the captured image data after receiving the captured image data and the emotion data representing the emotion felt by the subject viewing the captured image data. An image evaluation method provided by a first method, comprising: a first step of receiving a section (between tools) in which an intended emotion appears from a maker and storing the data as intertool data, and using the emotion data input during the tool from the intertool data And a fourth step of calculating a true score and calculating a true score, and a fourth step of providing a true score. The step is provided between steps 1 and 4. Calculate false value by calculating the value for the intended emotion from the emotion data input during the interval (unintentional interval) The method may further include a third step, wherein the fourth step further provides the False score. The first step may be performed before the second step, and normalizes the emotion data to a value between 0 and 1. And a first to third steps for performing morphological processing for converting the emotion data normalized in the first to second times to the duration using the morphology expansion operation. It can be achieved by the image evaluation method, characterized in that using the emotion data is completed morphological processing in steps 1-3.

The object of the present invention is to quantify the extent to which the subject's response is consistent with the intention of the producer providing the captured image data after receiving the captured image data and the emotion data representing the emotion felt by the subject viewing the captured image data. An image evaluation system for providing a video data, wherein the video data, the inter-tool data for a section (between tools) in which the emotions intended by the producer appear in the captured image data, and a memory unit for storing the emotion data are reproduced, while the captured image data is reproduced. True score is the score of the intended emotion included in the emotion data input during the intermediary using the section recognition section for recognizing and notifying the start and end sections of the section between the section tools and the section section notifying the section section section. Calculated as An emotion matching operation unit including a score calculation unit for calculating a false score of the intended emotion included in the emotion data, an image playback area for displaying captured image data, and an inter-tool display area for displaying whether or not it is between and An image generator for generating an image including an emotion level display area for providing a step by step the real-time intended emotion corresponding to the section of the captured image data, and normalizes the value output from the emotion data to a value between 0 and 1 The morphological processing unit for outputting the normalized emotion data input from the normalization processing unit and the normalization processing unit to adjust the delay time by reflecting the intensity by using a morphology expansion operation, the emotion matching operation unit and the image generating unit is the emotion output from the morphological processing unit Image evaluation system characterized by using data It is even possible to achieve.

Using the image evaluation system and method according to the present invention can objectively evaluate the image produced by the producer. In particular, when the VR content creator uses the present invention, it is possible to produce high-quality VR content customized for the consumer, thereby enabling the entry into a global new market with low-cost content production and helping to generate high profits.

In addition, when the present invention is used to produce a variety of cultural fusion content using the image, it is possible to accurately grasp the reaction of the subject in advance, it is possible to efficiently produce the video content at low cost.

1 is an embodiment of a system for displaying an image and recording an emotion felt by a subject while viewing the displayed image.
2 is a flowchart for explaining a method for evaluating video content according to the present invention.
3 is a graph showing emotion data before and after morphological processing.
4 is an explanatory diagram for a specific example of calculating a true score and a false score.
5 is a block diagram of an image evaluation system according to an embodiment of the present invention.
6 is an example of the resultant image produced by the image generating unit according to the present invention.
Fig. 7 is a graph showing the author's intentions for a plurality of emotions in captured image data having a 103 second running time.
FIG. 8 is a bar graph showing the same image in a group of 10 people, comparing the true values measured for each emotion of a particular subject belonging to the group with the average value of each true score.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprise" or "have" are intended to indicate that there is a feature, number, step, action, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

In addition, in the present specification, "on or above" means to be located above or below the target portion, and does not necessarily mean to be located above the gravity direction. Also, when a portion, such as an area, a plate, etc. is said "on or on top of" another part, it is not only in contact with or spaced apart from other parts, but also in the middle of another part. This includes any case.

In addition, in the present specification, when one component is referred to as "connected" or "connected" with another component, the one component may be directly connected or directly connected to the other component, but in particular It is to be understood that unless otherwise indicated, there may be connected or connected via another component in the middle.

Also, in this specification, terms such as first and second may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another.

Since the present invention relates to a system or method for evaluating measured emotions, techniques for measuring emotions are outside the scope of the present invention. Typical emotions that can be measured include surprise, cozy, attention, cognitive load, good eustress and bad stress.

1 is an embodiment of a system for displaying an image and recording an emotion felt by a subject while viewing the displayed image. Image and sensor data providing device 100 is provided with a plurality of images stored by the manufacturer. The stored image is processed by the graphic processing unit (GPU) and then displayed to the user through a head mounted display (HMD). PPG sensor, GSR sensor, EEG sensor and eye tracking sensor are attached to the user's body. The image signal processed by the GPU is displayed on a separate monitor (HMD Mirroring) through a splitter and stored in the image and sensor data providing apparatus 100 as captured image data using an image capture board. . At this time, the gaze information data input from the eye tracking sensor is synchronized with the captured image data, and the sensor signal output from the EEG sensor GSR / PPG sensor is also synchronized with the captured image data and stored in the image and sensor data providing apparatus 100. do. In the system shown in FIG. 1, an image capturing board is used to capture an image displayed on the HMD and to store the captured image data using the captured image. The image and sensor data providing apparatus 100 may be implemented as a general speaking personal computer having a memory, a GPU, and an image capture board.

The EEG sensor is a sensor for measuring brain waves. In the present invention, a product called MUSE provided by Interaxon was used. It is a brain wave measuring equipment using 4 channels of AF7, AF8, TP9, TP10, can be sampled at 256Hz, uses wireless communication using Bluetooth, and is worn on the head.

The GSR / PPG sensor was used to connect the Grove-GSR Sensor and Laxtha-RP520 to the Arduino board to measure skin response and blood flow. Both instruments can sample at any frequency in the desired range from 100 to 1000 Hz. Measure it by putting it on your finger like a ring.

The eye tracking sensor used FOVE0 as a device for acquiring gaze information. FOVEO is a device in which Fove adds an eye tracking sensor with a gaze tracking function inside the HMD. It can sample the gaze information from tracking the eye at 120Hz, and it is measured by wearing it in the same way as wearing HMD.

Techniques for analyzing what emotions are represented using sensor signals input from various sensors and converting them into emotion data for each emotion are beyond the scope of the present invention. That is, various methods are already known to convert (calculate) emotion data representing emotions such as surprise, stability, concentration, etc. using sensor data output from the plurality of sensors shown in FIG. 1 and are outside the scope of the present invention. The present invention aims to objectively display how faithfully the emotions intended by the subjects are expressed at the time intended by the image producers. For example, the emotion of 'surprise' may be calculated as surprise data using a sudden increase (or decrease) in the amount of change in the raw signal output from the EEG sensor, and the surprise data may be calculated from the EEG sensor value. Since the method is already known, it is not the scope of the present invention. Hereinafter, only the emotion data of surprise will be described, but of course, the same can be applied to other kinds of emotions.

2 is a flowchart illustrating a method of evaluating video content according to the present invention. Captured image data and sensor data are provided from the image and sensor data providing apparatus 100 shown in FIG. 1. The sensor data ST13 is converted to the emotion data ST11 using a separate algorithm (not in the scope of the present invention). After viewing the captured image to a plurality of subjects as a preliminary work for performing the image evaluation, the subject senses the section in which the emotion is felt and stores it as the subject experience section data (ST31). The parameters of the normalization function are obtained using the captured image data and the subject emotion interval data (ST33). The emotion data provided in step ST11 is normalized using a normalization function (ST34), and morphological processing is performed (ST35). After receiving information about the section that the emotion is intended to occur from the video producer, the information is stored as the tool data of the producer's emotion (ST360. Using catcher image data, the tool data of the emotion, and the morphological processing of the emotion data). The true score and the false score are calculated (ST37) and provided as a result image (ST39).

1.Calculation of function parameters for normalization

According to various experiments of the present inventors, an emotional analysis can be performed based on a biological signal of a subject, but there are subjects having a large change in the biological signal when the same emotion occurs in a plurality of subjects, and a subject who does not have the same. For example, if there are subjects A and B, it is assumed that A is a person whose biological signal change is large, and B is a person who does not. If we were equally surprised at the same intervals as A and B, then A would have a large value and B would have a small value. If the analysis is conducted based on these results, an unfair evaluation is obtained, and if the image content is analyzed based on this, reliable results cannot be obtained. Therefore, it is necessary to normalize the emotion data based on the biological signal within the same range. However, there is a part where it is difficult for a person to define the parameter of the normalization function due to the characteristics of the biological signal having a large deviation for each person. In order to solve this problem, the present invention normalized the emotion data.

In order to properly normalize the bio signals for each person, we find the optimal parameter of the normalization function. In the present invention, a sigmoid function such as Equation 1 is used.

The emotional data normalization function model provides the window size, window travel distance, normalization function (assuming normalization function used as sigmoid), slope (e), intensity ( Range of w) Five optimal parameters must be found. In Equation 1, r is emotion data, and S (r) means an output value of the sigmoid function and is output as a value between 0 and 1.

In the case of EEG signals used to measure surprise emotions, judgment is made by looking at the trend of the section instead of using only one signal to analyze emotions at one moment. For example, assuming that the sampling rate is 500hz, if you want to check the results at 1 second intervals, 500 data are covered with a window and the result is calculated. The number of sampled emotion data to view at once is the size of the window, and in the case of the window moving distance, it can be moved by 0.5 second or 0.1 second in the interval of 1 second, so the result will change. It is called Windongwoo moving distance to find the optimum moving distance by changing the movement little by little.

First, the captured image data is watched by a plurality of subjects, and then, each subject is identified, and a section for which a feeling of surprise is sensed is stored and then stored as an experience section data. And when the surprise data is measured in the section in which the subject was surprised (this is called the 'experimental section'), this is calculated using the true score, and the section in which the subject is not surprised (the 'out-of-experimental section') In this case, the surprise data is measured, and the false score is calculated using this.

After normalizing the emotion data from 0 to 1 using the sigmoid function, the true score and the false score are obtained from the average of the sigmoid function output values in the experience section and the average of the non-experimental sections. As described above, the experience section means a section (expressed in units of frames or time) indicated by the subject in the captured image data, and the out-of-experiment section indicates that the subject did not feel surprise in the captured image data. It means a section.

The True Score for each subject is calculated using the Brute Force method, which modifies all parameters within a range (e.g. 0 to 10) and small values (e.g. 0.01) and calculates the results for all cases. Each parameter of the normalization function that is represented by the highest value and the false score is the minimum value is obtained. The Brute Force method takes a lot of time, but it has the advantage of being most accurate because it counts the number of cases. All parameters are modified little by little with brute force, and the model is trained with the goal of maximizing the true score and suppressing the false score. It uses the normalization function to select the function parameters to produce results as close as possible to the emotions actually felt by the user.

So far, the description has described using sigmoid as a normalization function, but the present invention is not limited thereto, and any normalization function that outputs emotion data as a value between 0 and 1 can be used. As an example, the Hyperbolic Tangent function can also be used. In addition, the parameter of the normalization function can be obtained through the machine learning method, for example, in addition to the brute force method.

2. Morphological treatment of emotion point and aftertaste

Normalized emotion data changes every time. Based on this, if the score is calculated as the score of the producer, the result is correct in terms of signal, but a person (image producer, etc.) sees the graph. There is a difficulty in understanding. Because it is assumed that there is a sudden decrease in the biosignal at a certain moment, from the point of view of the signal it can be seen that the emotion changes in an instant, but the emotions that a person actually feels may be in the form of lingering or even the same surprise. Responses may occur at different time points in each subject. Due to these factors, the present invention performs morphological processing rather than expressing emotions according to the emotion analysis algorithm.

As for the point of view and aftertaste, the morphological expansion operation that is frequently used in the morphological processing of images is applied after the point of occurrence of emotion. This assumes that strong emotions last a little longer and perform expansion operations in proportion to the magnitude of the emotional signal.

Biological signals of various people reflect the timing of feeling the individual, the time at which the induced emotion is maintained, and the time of change of the emotion. Because of these various factors, it is not easy to quantify the emotions analyzed using bio signals. The reason is that even if the same emotion is felt, the triggered emotion will be recorded with a high score for long-lasting subjects, and for those who do not, it will be recorded with a small value. If it is visualized and expressed as it is, it can be corrected from the signal point of view, but for those who want to analyze emotions through the results, there is a possibility that it may cause misunderstandings such as the idea that the person whose emotions do not last long has not felt the emotions. For this reason, it is assumed that emotions do not change rapidly within a short time (1 second), so that the stronger the intensity of emotions induced at the time of emotions is expressed, the longer the emotions are maintained, so that all have similar durations. Representation method performed Morphology expansion operation which is widely used in the morphological processing of the image, the stronger the intensity was increased the number of iterations. 3 is a graph showing morphological processing, (a) is input normalized processed emotion data, and (b) is a graph showing emotional data completed until morphological processing.

The morphology expansion equation is shown in Equation 2. Equation 2 is replaced with the largest value among the surrounding values, and as the iteration is repeated, the shape of the graph expands. The stronger the emotion, the more iterative operations are performed.

3. Calculation of matching score according to the intention of the video producer

The video content producer registers a section (frame unit or time unit) in which surprise feelings appear in the video content according to the producer's intention. For example, the time interval between 1:05 and 1:07 is set to a section that you want to be very surprised, and the degree of surprise is registered (1:05 to 1:07, 100% surprise). When registration is completed, a matching score is calculated by comparing the feelings of the actual subject to the content.

There are two matching scores as described above. The first score is a true score that indicates that a corresponding emotion is detected within an intention emotion section (hereinafter, referred to as an 'appraisal tool') set by the content creator. If the True Score is high, it means that the subject actually triggered the emotion in the section intended by the creator, which means that the content was produced as intended by the creator. On the contrary, if the number of benefits is low, it means that the content was not produced as intended. More specifically, when the normalized and completed morphological processing of the emotion data is input in the section intended by the image producer, the average value obtained by accumulating and dividing by the interval is calculated as a true score.

Another matching score is False Score. This is a score indicating that surprise emotions appear in sections other than the intention emotion section set by the content creator (hereinafter, referred to as 'outside the emotion intention section'). A high False score means that a lot of surprise emotions were triggered in the unintentional section. This may or may not affect content directly due to unexpected emotions. More specifically, when emotion data normalized and morphologically processed is input in an unintentional section, the average value is calculated as a false score. Finally, creators can quantitatively evaluate content through these two scores.

4 is an explanatory diagram for a specific example of calculating a true score and a false score. The entire video consists of 60 seconds of running time, and the planner assumed that the intention was to surprise in the Z1 and Z2 sections. In addition, it is assumed that the Z1 section is formed in 10 seconds, the cumulative value of the surprise data measured during the Z1 section is 3.5, the Z2 section is formed in 5 seconds, and the cumulative value of the surprise data measured during the Z2 section is 2.1. It is assumed that the cumulative value of the surprise data measured in the interval is 2.5. In the embodiment of Figure 4 True Score is calculated as' (3.5 + 2.1) / 15 = 0.37, False Score is calculated as' 2.5 / 45 = 0.05 '.

5 is a block diagram of an image evaluation system according to an embodiment of the present invention. As shown in FIG. 5, the image evaluation system according to the present invention includes a memory unit 10, an emotion matching unit 20, a normalization processing unit 21, a morphology processing unit 23, a normalization function parameter calculating unit 25, and an image. It is composed of a generation unit 30 and the display unit 35.

The normalization processing unit 21 is a module for normalizing the emotion data of the subject to a value between 0 and 1 using the normalization function as described above. The morphology processing unit 23 is a module for processing the emotion data output from the normalization processing unit 21 to reflect the afterimage (duration time) according to the intensity. As described above, the normalization function parameter calculating unit 25 receives the captured image data and which section of the captured image is felt by the plurality of subjects who have viewed them, and uses the experience section data in which information about the section is recorded. Module for calculating the parameters of a normalization function. The normalization processor 21, the morphology processor 23, and the normalization function parameter calculator 25 are generally provided as software modules, but may be implemented in hardware if necessary.

The memory unit 10 inputs the captured image data provided to the image and sensor data providing apparatus 100 and the experience section data and input devices received from a plurality of subjects through an input device not shown in the drawing. Save the data between the receiving tools. The captured image data, the inter-tool data, and the morphologically processed emotion data are input to the emotion matching operation unit 20.

Emotion matching unit 20 is composed of a section recognition unit and a score calculation unit. The section recognition unit counts the frame unit or playback time while playing the inputted captured image data, and notifies the score calculation unit of the start and end of the frame belonging to the inter-tool data or the start and end time of the playback time. The score calculation unit accumulates the emotion data input when the video corresponding to the unintentional section is played and stores it in the False storage unit. When the video corresponding to the intention section is played, the score calculation unit accumulates and stores it in the True storage unit. do. When the image playback ends, the cumulative value stored in the false storage unit is divided by using the total number of frames constituting the unintentional section or the related total playing time, and the false score of the subject is calculated. Similarly, when the image playback ends, the cumulative value stored in the True storage unit is divided by using the total number of frames constituting the intended section or the associated total playback time to calculate the true score of the subject. The image generator uses the captured image data, inter-tool data, morphologically processed emotion data, and the false scores and the true scores calculated by the score calculator to construct an image in the image forming unit and displays them on the display unit 30. .

The emotion matching operation unit 20 is also used to calculate the parameter of the normalization function. The section recognition unit counts the frame unit or playback time while playing the inputted capture image data, and notifies the score calculation unit of the start and end of the frame belonging to the experience section data or the start and end time of the playback time. The score calculation unit accumulates the emotion data input when the video corresponding to the non-experimental section is played and stores it in the False storage. When the video corresponding to the experience section is played, the score data is accumulated and stored in the True storage. do. When the video playback ends, the cumulative value stored in the false storage unit is divided by using the total number of frames constituting the non-experimental section or the related total playing time, and the false score of the subject is calculated. Similarly, when the image playback ends, the true score of the subject is calculated by dividing the cumulative value stored in the True storage unit by using the total number of frames constituting the experience section or the related total playback time. The normalization function parameter calculation unit 25 calculates a true score and a false score of a plurality of experienced persons using the experience interval, and calculates a parameter of the normal function in which the true score is the highest value and the false score is the lowest value.

4. Provide result video

The emotion result of the measurer (subject) and the intention emotion of the planner can be easily observed in a device (computer) equipped with a program explaining the above-described process. However, it is inconvenient to access the program in order to see the results of the measurer, and the disadvantage is that you cannot see the analyzed results from the outside. In order to solve this problem, the results of the analysis of the emotion of the measurer and the intention of the planner were produced and provided at a glance. 6 is an example of the resultant image produced by the image generator according to the present invention. Using the resultant image as shown in FIG. 6, the analyzed result can be confirmed anytime and anywhere through the extracted image, and the result can be confirmed as a video without a program from the outside. The produced video has three characteristics different from the general video.

(1) A gaze history is indicated based on gaze information collected when a subject views an image in an area (area reproducing area) where captured image data is reproduced (a portion indicated by a green square in FIG. 6).

(2) The type of emotion felt by the subject (area indicated by (2)) is indicated, and on the right side, the intensity is expressed in the form of hi-fi equalizer (area indicated by (3), emotion level display area).

(3) In order to be able to check whether the plan is intended for the emotion section at that time, the LED lights up (indicated by (4), the inter-instrument display area) when the circle lights up and the corresponding emotion triggers section. It is implemented to make it easy to see that it is not an emotion section.

Until now, only specific emotions of surprise have been described. However, as described above, similar emotions such as stability and concentration can be numerically expressed at once. Fig. 7 shows the producers of surprise, cozy, attention, cognitive load, good stress and bad stress on the captured image data with 103 seconds of running time. Graph showing the tools in. As shown in FIG. 7, it can be seen that an intention section of a producer, which is expected to show various intended emotions in one image, can be input. FIG. 8 is a bar graph showing the same image in a group of ten people, comparing the group value represented by averaging each True score and the True score measured for each emotion of a specific subject belonging to the group. As shown in FIG. 7 and FIG. 8, it can be seen that several emotions that appear while watching a reproduced image can be numerically presented at once.

While preferred embodiments of the present invention have been described above using specific terms, such terms are only for clarity of the present invention, and embodiments and described terms of the present invention are departed from the spirit and scope of the following claims. It is obvious that various modifications and changes can be made without a change. Such modified embodiments should not be understood individually from the spirit and scope of the present invention, but should fall within the claims of the present invention.

10: memory unit 20: emotion matching operation unit
21: normalization processing unit 23: morphology processing unit
25: Normalization function parameter calculation unit 30: Image generation unit
35: display unit 100: image and sensor data providing device

Claims (7)

delete delete An image evaluation method of numerically providing a capture image data and an emotion data indicating an emotion felt by a subject who has viewed the captured image data, and then quantifying the degree to which the subject's response matches the intention of the producer who provided the captured image data. As
A first step of receiving a section (between tools) in which an intended emotion appears from the producer and storing the data as tools between tools;
A second step of calculating a true score by calculating a value for the intended emotion using the emotion data input during the inter-tool between the inter-tool data and
A fourth step of providing the True score,
As a step provided between the first step and the fourth step,
Comprising a third step of calculating the value for the intended emotion from the emotion data input during the interval (excluding the intentional interval) excluding the inter-instrumental data from the inter-instrumental data to calculate a false score,
In the fourth step, the False score is further provided.
As the step performed before the second step
1 to 2 normalizing the emotion data to a value between 0 and 1;
A first to third steps of performing morphological processing for converting the emotion data normalized in the first to second times into a duration using morphological expansion operations,
In the step after the second step, the image evaluation method, characterized in that using the emotional data is completed in the first to third morphological processing.
The method of claim 3,
The first and second steps are performed using a normalization function.
The method of claim 4, wherein
The normalization function is an image evaluation method, characterized in that the sigmoid function.
delete An image evaluation system for receiving the captured image data and the emotion data representing the feelings of the subject who viewed the captured image data and quantifying the degree to which the reaction of the subject matches the intention of the producer who provided the captured image data. As
A memory unit for storing inter-tool inter-data and the emotion data for a section in which the emotion intended by the producer appears in the captured image data, the captured image data;
While the captured image data is reproduced, the emotion data inputted between the tool using the interval recognition unit for recognizing and notifying the start and end intervals between the tool between the tool between the tool and the tool notified to the section recognizer. Emotion coincidence including a score calculator that calculates the score of the intended emotion included as a true score, and calculates the score of the intended emotion included in the emotion data input for the interval except for the intentional period (false). With an operation unit,
An image including an image playback region to play back the captured image data, an inter-tool display region to indicate whether the captured image data is between, and an emotion level display region to gradually provide a real-time intended emotion corresponding to a section of the reproduced captured image data Including an image generating unit for generating,
A normalization processor for normalizing the value output from the emotion data to a value between 0 and 1;
A morphological processing unit for outputting the normalized emotion data input from the normalization processing unit by adjusting the delay time by reflecting the intensity using a morphology expansion operation, the emotion matching operation unit and the image generating unit is output from the morphological processing unit Image evaluation system, characterized in that using the emotion data.

Images