JPH08263667A - Method for recognizing rorschach picture - Google Patents

Abstract

PURPOSE: To obtain a Rorschach picture recognizing method where a picture recognition process is made to be a more human-friendly-converted one. CONSTITUTION: A rough scanning within about one-tenth of a finally required max. grid number is taken place over the whole areas of an original picture 1 within about one-tenth of whole required time. A polygonal line in a figure 1 is the locus 2 of the rough scanning. The center coordinate of a characteristic part (the area which becomes a large projection, recession or an island) is detected by the rough scanning. Successively, a priority watching area 3 which includes the characteristic part and is more than about 1% and equal to below 10% of the original picture 1 in an area is set within the rest of time so as to take place the close scanning of the area 3. The characteristic part of the original picture 1 is recognized and analized on the close scanning. The priority watching range 3 has various kinds of forms. They are, for example, a circle, a rectangle and an oval, etc. By this procedure, important picture information is recognized with few scanning resources in a short time.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Rorschach image recognition method, for example, to a Rorschach image recognition method that enables human-friendly feature extraction and highly efficient image data storage.

[0002]

2. Description of the Related Art Conventionally, image data processing techniques, particularly image compression and transmission techniques, have made rapid progress as the socialization of society progresses. In such an environment, there is a demand for more accurate recognition of image data, facilitation of handling, etc., as well as higher speed processing, finer and finer images, and larger sizes. These requirements result in the pursuit of more humane processing, in that the images are perceived by the stimulation of the human retina. Therefore, in order to meet the above demands, it is necessary for the computer to perform so-called human friendly feature extraction processing or output an image having features that are easily accepted by humans.

As a document relating to human reaction to images, there is a research paper by Rorschach. Rorschach, a psychologist, wrote a research paper on psychological studies (Ps
ycho signost) ”. In this paper, it is stated that frequently appearing reactions such as human motion and color reactions are associated with special psychological or behavioral features.

Although the contents of the configuration are different from those of the present invention, Japanese Patent Laid-Open No. 62-198975 and No. 62-3998 disclose a procedure for extracting and processing a characteristic portion of an image.
There is No. 0 publication.

[0005]

However, due to its characteristics, machines excel at simple and high-speed processing, and learning, habits, and
So-called human processing such as judgment, feelings, emotions, hobbies and preferences is not good at it. Machines are used by humans, and pursuing ease of use ultimately leads to the pursuit of human friendliness. As an example, there is a conventional “fuzzy function” which is not directly related to the present invention. Also in image processing, if more convenience is pursued, consideration of human friendliness is essential.

The image recognition function of a computer is not yet industrially and practically tolerable for capturing arbitrary images. Considering the social requirements as described above, there is a pressing need to further develop a computer mechanism that emulates a human image recognition function.

It is an object of the present invention to provide a Rorschach image recognition method in which the image recognition process is made more human friendly.

[0008]

In order to achieve such an object, the Rorschach image recognition method of the present invention uses a rough image of about 1/10 or less of the maximum number of grids finally required in a predetermined image area. A rough scanning step of performing scanning and a large protrusion, depression, or the like in the scanning in this rough scanning step.
A characteristic part region detecting step of detecting the center coordinates of a characteristic part such as an island region, and including the characteristic part based on the center coordinates, the area is about 1% or more of a predetermined image region and about 1
And a characteristic region dense scanning step of densely scanning a focused gaze region having a predetermined shape within 0%.

Further, it is preferable that the focused area having the predetermined shape is circular, rectangular, or elliptical, and the rough scanning step is scanning of the predetermined image area in the vertical direction in the horizontal direction. Furthermore, when a plurality of characteristic portions are detected in the characteristic region detecting step, those having substantially the same height are designated as P1, P2, ... , An are set as the target areas, and the set characteristic areas are set as A1, A2,
The scanning in the characteristic region dense scanning step may be executed in the order of ..., An. Note that the characteristic areas are A1, A2, ..., A
When set as n, the characteristic parts P1, p2, ...
It is preferable that the characteristic portions P1, P2, ..., Pn are exchanged in order so that the path of the polygonal line connecting Pn becomes smaller, and the scanning in the characteristic region dense scanning step is performed in the exchanged order.

The scanning in the characteristic region dense scanning step is executed from any one of the upper, lower, left and right ends of the predetermined image region of the characteristic regions A1, A2, ... The part may be determined based on the size of the region of the characteristic part.

A Rorschach image recognition method of another invention is a rough scanning step of performing a rough scanning within about 1/10 of the maximum number of grids finally required in a predetermined image area, and a rough scanning step. Greatly protruded by scanning at
Partial region of a characteristic part such as a region that is depressed or becomes an island, and a characteristic partial region detection step of detecting the central coordinate position of this partial region, and the center of gravity of the land region excluding the partial region from the area of a predetermined image Land area barycentric position detection step of detecting coordinate position, and sequential scanning from the barycentric coordinate position of the land area to one position of the central coordinate position of the partial area, and from this one position to the central coordinate position of another partial area Characteristic partial scanning step, and a multiple refraction portion detection step for detecting a multiple refraction portion whose trajectory in the characteristic partial scanning step is a distance within about 1/10 of the scale of a predetermined image and which is bent twice or more. And a focused gaze area setting step in which the area of the focused gaze area that includes the multiple refraction portion is approximately 1% or more and approximately 10% or less of the area of a predetermined image, and the focused gaze area is closely scanned. Feature area sprinting It is characterized by a step.

Further, in the Rorschach image recognition method described above, the center coordinate positions of the multiple bent portions are connected between the multiple refractive portion detecting step and the focused gaze area setting step to form the contour of the multiple refractive portion. It is advisable to include the multiple refraction partial contour forming step for recognizing the contour of a predetermined image from the contour shape.

[0013]

Therefore, according to the Rorschach image recognition method of the present invention, rough scanning is performed within about one-tenth of the maximum number of grids finally required in a predetermined image area, and this scanning is performed to a large extent. The center coordinates of a characteristic portion such as a protruding, sinking, or island region is detected, and the detected characteristic portion is included and the area is approximately 1 in a predetermined image area.
The dense focused area of a predetermined shape of not less than 10% and not more than about 10% is densely scanned. Therefore, it is possible to quickly detect the characteristic portion of the image by coarse scanning and to quickly detect the characteristic portion of the image by scanning the detected characteristic portion closely.

According to the Rorschach image recognition method of another invention, a rough scan is executed within about one tenth of the maximum number of grids finally required in a predetermined image area, and this scan causes a large protrusion. , A partial region of a characteristic part such as a region that is depressed or an island and a central coordinate position of this partial region are detected, and the barycentric coordinate position of the land region, which is the predetermined image region excluding the partial region, is detected. Scanning is sequentially performed from the detected barycentric coordinate position of the land region to one position of the central coordinate position of the partial region, and from this one position to the central coordinate position of another partial region. Further, a multi-refractive portion in which a locus in this characteristic partial scanning is bent within a distance of about 1/10 of a scale of a predetermined image and is bent twice or more is detected, and the multi-refractive portion is included, and the included emphasis is included. The area of the focused area is set to approximately 1% or more and approximately 10% or less of the area of the predetermined image, and the focused area is closely scanned. Therefore, it is possible to more accurately detect the characteristic portion of the image and more quickly recognize the characteristic portion of the image.

[0015]

Embodiments of the Rorschach image recognition method according to the present invention will now be described in detail with reference to the accompanying drawings. FIG.
2 and FIG. 2, there is shown a diagram for explaining an embodiment of the Rorschach image recognition method of the present invention. These figures are diagrams for explaining the procedure for searching the characteristic portion of a predetermined image.

The outline of Rorschach, which constitutes the name of the present invention, is as described above. The present invention attempts to use the human reaction principle pursued by Rorschach for searching and recognizing a characteristic portion of an image. From these footprints, the present invention uses the name Rorschach to mean recognition of feature points in an image based on human features.

FIGS. 1 and 2 show a Rorschach image and a locus of human viewpoint scanning with respect to the Rorschach image. These drawings conceptually describing the procedure of this viewpoint scanning are illustrated based on the interpretation that a person is paying attention to the curved region of the trajectory 2 of the viewpoint. In the following, this area will be defined as a "focused attention area". The focused gaze area is simply a bent portion of the locus 2 of the viewpoint. On the other hand, the microscopic areas are areas that are prominently projected or depressed, areas that are islands, and areas that are lakes. Therefore, the characteristic gaze area has a conceptual meaning, whereas the focused gaze area 3 is a more specific area of the characteristic gaze area for use in the analysis procedure.

As described above, one of the features of image recognition at the time of performing the Rorschach test method is to use the same procedure as the method of capturing the image information of the subject from the eyeball to the primary visual cortex of the brain. There is a point. That is, instead of uniformly scanning the entire screen as in mechanical raster scanning, the human eye efficiently captures image information by moving the viewpoint at first glance at random.

Therefore, for comparison between the mechanical scanning and the human recognition procedure, the efficiency of image input by these scanning methods will be considered. Taking the scanning of the display screen as an example, the data of both are as follows.

Mechanical Raster Scan: 640 × 480 Rorschach Scan for People: 640 × 10

From the above results, it is clear that the person has adopted the highly efficient image recognition method. In addition, the focused area, which is simply the curved portion of the trajectory of the viewpoint, corresponds to the characteristic area on the figure, such as the area that is significantly projected or depressed, or the area that is an island. ing.
Considering these experimental results, the input mechanism of Rorschach image is analyzed as follows.

A) The input of the image itself to the visual cortex is 0.
Complete within 1 second. B) Recalling a plurality of objects and associating the details thereof with the part of the Rorschach image are executed simultaneously with the scanning of the eyeball. This is referred to as "overall cognition" / Pattern 1. Focusing on each area of the Rorschach image, conversion to an object is performed for each area, and then a story is created between these object groups in about 1 second. This is called "partial cognition" / Pattern 2. C) Unconscious, consciousness-based censorship work and prioritization are performed in approximately 5-10 seconds.

Corresponding to the two image recognition methods, it is expected that there are two types of viewpoint movement on the Rorschach image. One corresponds to global cognition, in which case the viewpoint is focused on the characteristic region of the recognized object. The other corresponds to partial recognition, in which the viewpoint movement is executed before recognizing the object. Therefore, the viewpoint is focused on the mechanically characteristic region, for example, the region where the protrusion or the depression is large.

Based on the analysis of the image recognition procedure of these persons, the present embodiment constitutes a method for extracting the features of the image data (original image). This procedure will be described below with reference to the flowchart of FIG. First, a rough scan, which is a rough scan within about 1/10 of the total required time and within about 1/10 of the maximum number of grids finally required, is spread over the entire area of the original image 1. And execute (S10). By this rough scanning, the center coordinates of a characteristic portion (a large protruding / depressed or island area) are detected (S12). Subsequently, the original image 1 is included in the area including the characteristic portion 3 in the remaining time.
The focused area 3 that is a small area of approximately 1% or more and approximately 10% or less is set, and dense scanning that is dense scanning is performed for the set focused area 3 that has been set (S14). Based on this fine scanning, the characteristic portion of the original image 1 is recognized and analyzed (S
16). There are various forms of the above-mentioned focused area 3 of interest. For example, a circle, a rectangle, a horizontally long ellipse, a vertically long ellipse, and the like.

The above scanning is based on the following theory. That is, there is the fact that not all image information is of equal value to the method when visually recognizing the real space. The information that is actually important is often feature points (first data) on a relatively coarse scale, or more detailed information (second data) near the feature points. As in the case of this requirement item, the first data can be recognized in a short time by first performing the rough scanning within approximately 1/10 of the maximum number of grids finally required. Further, it is intended to efficiently recognize the important information of the second data based on the preliminary knowledge of the first data by scanning the vicinity of the feature point in detail. According to the present embodiment, the initial scanning time is limited to the blinking time of a person within about 0.1 second, and visual recognition can be quickly performed in response to the surrounding situation which is actually changing.

Many variations on the above embodiment are possible. As one example of changes, there are various forms of the region to be scanned and their respective characteristics. For example, if the focused area to be scanned is rectangular, the relationship between the focused area and the position and direction of the entire area is clarified, and the procedure for reconstructing the data is simplified. The horizontally long ellipse makes it similar to the human visual area and can simulate image recognition peculiar to living things. In addition, by forming a vertically long ellipse, it is possible to simultaneously scan a long distance area and a short distance area at the same time. Furthermore, by making the focused gaze area to be scanned amorphous, it is possible to read a wide range of features with a small scanning resource.

As a second variation, the center of the focused area is set in order from the top, and when there are a plurality of objects having the same height, the focused areas are set in order from the left as P1, P2, ..., Pn. , A1, A2, ..., A
Set multiple n. Subsequently set characteristic regions A1, A
Scan in detail in the order of 2, ..., An. Based on this procedure, the scanning order is simple and clear, and a reliable scanning procedure can be realized.

As a third modification, the path of the polygonal line connecting the centers of the focused areas is minimized (P1,
There is a feature extraction system characterized by changing the order of P2, ..., Pn) and scanning A1, A2, ..., An in detail in that order. By minimizing the moving distance, it becomes possible to move quickly between the scanning areas, and efficient image recognition can be executed.

As a fourth variation, (P1, P2, ..., P
n), the order of (P1, P2, ..., Pn) is exchanged so that the path of the polygonal line connecting n) is relatively small, and A1, A2, ..., An are scanned in detail in that order. There is a system. By making the moving distance relatively small, it becomes possible to move quickly between the scanning areas, and at the same time, it is not necessary to strictly search for the shortest distance, so the search time itself can be saved.

According to each of the above-described embodiments or modifications, it is possible to recognize important information of an image with a small scanning resource and in a short time. Therefore, it is possible to configure a method capable of flexibly collecting and recognizing image information and immediately responding to a situation that is changing from moment to moment. It is a very important and useful requirement in the computerized industrial society to detect, recognize, and take action as quickly as possible when a portion of an image contains important information.

It should be noted that the above-described embodiments and modified examples are preferred embodiments of the present invention, but are not limited thereto. Various modifications can be made without departing from the scope of the present invention. For example, the precision scanning start position is TS
Either one end of the P (traveling salesman problem) path can be used as the scanning start point. By minimizing the moving distance, it becomes possible to move quickly between the scanning areas, and efficient image recognition can be executed. Further leftmost,
A simple procedure of scanning is possible by starting the scanning from the four corners of the rightmost, topmost and bottommost areas. Smaller areas can be scanned in a shorter time by scanning from the area of smallest area, so they are part of the details of the characteristic area, but are available early in the total scan time. Scanning from the largest area of area is likely to include important features of the entire image, increasing the likelihood of recognizing important features early in the total scan time. it can. And so on.

[0032]

As is apparent from the above description, the Rorschach image recognition method of the present invention performs a rough scan within a region of a predetermined image within about 1/10 of the maximum number of grids, and this scan is characterized. The center coordinates of the target portion are detected, and the focused gaze area of a predetermined image is densely scanned in an area including the detected characteristic portion. Therefore, the rapid detection of the characteristic part of the image and the quick recognition of the detected characteristic part are to extract and recognize the important information from the image that changes from time to time, and to quickly respond to it. Is possible. With this efficient image processing, a faster computer system can be constructed with a simpler computer system.

According to the Rorschach image recognition method of another invention, a rough scan is executed in a predetermined image area,
This scanning detects the partial area of the characteristic portion and the center coordinate position of this partial area, and detects the center of gravity coordinate position of the land area excluding these partial areas. Sequential scanning is performed from the detected barycentric coordinate position of the land region to the central coordinate position of the partial region, and further, the multiple refraction portion in which the locus in this characteristic partial scanning is bent within a predetermined distance and bent twice or more is detected. Include and scan densely as a defined focus area. Therefore, it is possible to extract a higher characteristic portion from the characteristic portions of the image, perform more accurate detection, and perform recognition of the characteristic portion of the image more quickly.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an embodiment of a Rorschach image recognition method of the present invention, conceptually showing a scanning procedure of a characteristic region of a frame image.

FIG. 2 is an explanatory diagram of an embodiment of a Rorschach image recognition method of the present invention, conceptually showing a scanning procedure of a characteristic region of a frame image.

FIG. 3 is a flowchart illustrating an embodiment of the Rorschach image recognition method of the present invention.

[Explanation of symbols]

 1 Original image 2 Locus of viewpoint 3 Focus area (characteristic area)

Claims (9)

[Claims] 1. A rough scanning step for performing a rough scanning within about 1/10 of the maximum number of grids finally required in a predetermined image area, and a large protrusion or depression in the scanning in the rough scanning step. A characteristic portion area detecting step of detecting the center coordinates of a characteristic portion such as an island area, and the area including the characteristic portion based on the center coordinates is approximately 1% or more of the area of the predetermined image. A Rorschach image recognition method, comprising: a characteristic region dense scanning step of densely scanning a focused region of a predetermined shape within approximately 10%. 2. The Rorschach image recognition method according to claim 1, wherein the focused area having the predetermined shape is any one of a circle, a rectangle, and an ellipse. 3. The Rorschach image recognition method according to claim 1, wherein the rough scanning step is a scan of the area of the predetermined image in a horizontal direction in a vertical order. 4. When a plurality of characteristic portions are detected in the characteristic region detecting step, those having substantially the same height are designated as P1, P2 ,. , An of the characteristic regions to be set are set as A1, A2, ..., An, and the set characteristic regions are set to A1, A2 ,.
4. The Rorschach image recognition method according to claim 1, wherein scanning in the characteristic region dense scanning step is performed in the order of An. 5. The characteristic regions are A1, A2, ..., An.
, P, p2, ..., P
The order of the characteristic portions P1, P2, ..., Pn is switched so that the path of the polygonal line connecting n is smaller, and the scanning in the characteristic region dense scanning step is performed in the replaced order. 4. The Rorschach image recognition method described in 4. 6. The scanning in the characteristic region dense scanning step is performed from any one of upper, lower, left and right ends of the region of the predetermined image of the characteristic regions A1, A2, ..., An. The Rorschach image recognition method according to claim 5. 7. The Rorschach image recognition method according to claim 6, wherein any one of the upper, lower, left and right ends is determined based on the size of the region of the characteristic portion. 8. A rough scanning step of performing a rough scanning within about 1/10 of the maximum number of grids finally required in a predetermined image area, and a large protrusion or depression in the scanning in the rough scanning step. A characteristic partial area detecting step of detecting a partial area of the characteristic area as an island area and a central coordinate position of the partial area; and a center of gravity of the land area excluding the partial area from the predetermined image area. A land area barycentric position detecting step of detecting a coordinate position, from the barycentric coordinate position of the land area to one position of the central coordinate position of the partial area, and from the one position to the central coordinate position of the other partial area And a characteristic partial scanning step of sequentially scanning, and a multiple refraction portion whose locus in the characteristic partial scanning step is a distance within about 1/10 of the scale of the predetermined image and which is bent twice or more is detected. Multiple bending A partial detection step, a step of setting the focused area of interest including the multiple refraction portion, and setting the area of the included focused area of attention to approximately 1% or more and approximately 10% or less of the area of the predetermined image; Rorschach image recognition method, comprising: a characteristic region dense scanning step of densely scanning a target gaze region. 9. The Rorschach image recognition method further comprises connecting respective center coordinate positions of the multiple bending portions between the multiple refracting portion detecting step and the focused gaze area setting step, and the contours of the multiple bending portions. 9. The Rorschach image recognition method according to claim 8, further comprising: a step of forming a multiple refraction partial contour for forming a contour, and recognizing the contour of the predetermined image from the shape of the contour.