JPS60261276A - Digital picture processing method - Google Patents

Abstract

PURPOSE:To extract accurately structural features even in case of a picture having complicated constitution by retrieving extreme values from the retrieval initial point of each small area and displaying all obtained origins to extract a pattern of the digital picture. CONSTITUTION:In a unit process, unit vectors e1 and e2 orthogonal to each other are set, and a vector sum E=Sb+k.e1 is obtained while increasing a vector from a retrieval initial point Sb in the direction of the vector e1 by an integer (k). If the order of the gradient is in the increase direction, the value (k) is increased. At the time when the gradient is inverted from increase to reduction, the value (k) is increased or reduced to change the gradient in the direction of the vector e2 orthogonal to a vector E1. A maximum position S1 of the gradient is obtained on a basis of the inversion time of a vector E2 in the direction of the vector e2. Vectors E1 and E2 are synthesized to obtain a vector E3, and thus, a unit gradient maximum value E2 is obtained. Origins Sb, S1, S2- are stored in individual unit processes to obtain an extreme value Sp. Origins S1, S2- are obtained from values obtained by this extreme value retrieval to extract a feature pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application The present invention relates to a digital image processing method suitable for extracting characteristic patterns of digital images.

(B) Prior Art Generally, in ultrasonic diagnostic apparatuses and the like, information inside a subject is displayed as a digital image. In order to improve the image quality of this digital image, conventional processes such as image smoothing, edge enhancement, and various filtering methods have been used. However, although this type of processing is effective in terms of image visibility, it does not extract the structural features of the image.

(c) Problems to be solved by the invention Digital images are displayed as grayscale images with a predetermined gradation level, but if the characteristics of the image itself can be extracted from this digital image, it will be an effective means of image analysis and ultrasonic diagnosis. With devices, it becomes possible to make even more accurate diagnoses. However, conventionally, understanding the overall structural features of an image has largely relied on the experience of the diagnostician, and it has been difficult to extract and display those features on the screen.

The purpose of the present invention is to solve such conventional problems and to enable simple extraction of only the structural features of a displayed digital image, thereby making it easier to analyze the structure of the image. shall be.

(d) Means for Solving the Problems In order to achieve the above-mentioned object, the present invention divides the display area of a digital image into a plurality of predetermined small areas, and places a search initial point in each of the divided small areas. Set multiple values, and perform an extreme value search by applying the direct search method for the extrema of a nonlinear function from each search initial point, and calculate the base points that indicate the maximum for each unit process in the process of extremum search. The pattern of the digital image is extracted by displaying all the base points found by extreme value search from each search initial point of each small region.

(e) Examples Hereinafter, the present invention will be explained in detail based on examples shown in the drawings.

FIG. 1 is a flowchart showing the dental image processing method of the present invention. In this digital image processing method, first,
For example, in the case of an ultrasonic diagnostic device, the image data is captured using a digital scan converter (step Nl).
), a display area to be subjected to image processing is determined from among the captured image data (step N2). Here, as an example, a 6-bit two-dimensional image with 512×512 pixels is assumed to be the display area to be image-processed. Next, the second display area is divided into a plurality of predetermined small areas. For example, the area is divided vertically and horizontally into 16 areas, and each of the N=256 small areas is made up of 32×32 pixels (step N3). Then, the count 1 corresponding to one search initial point is set to 1 without changing it (step N4). Next, a plurality of search initial points sb are set within each small area (step N5). Then, from each search initial point sb, a direct search method for extreme values of the nonlinear function is applied to perform an extreme value search (step N6). This extreme value search is performed as follows. As shown in Figure 2, first, in the case of a 6-bit pixel,
Each pixel is ranked according to the gradation level from 0 to 63. Next, for one unit process, unit vectors e1 and e2 that are orthogonal to each other are set, and the vector is Sum E = Sb + -e,
, and compare the order of the gradation levels before and after each time. That is, when the ranking of gradations tends to increase (or decrease), the steps k are sequentially increased or decreased. When the order of the gradation levels reverses from an increasing (or decreasing) tendency to a decreasing (increasing) tendency, the previous and subsequent gradations are adjusted in the same manner as above in the direction of the unit vector e2 orthogonal to the vector E1 at that time. While successively comparing the size of , k is increased or decreased. Then, as the unit vector e and k in the direction increase or decrease, the gradation tends to increase (decrease) to decrease (increase).
The maximum (minimum) position Sl of the gradation indicated by the vector E at the time when the trend reverses is stored. This completes one unit process, so in the next 9
A unit vector e3 indicating the direction of vector E3 obtained by adding and combining l and E2, and a unit vector e orthogonal to this,
Then, use the same procedure to find the vector E5. Then, a base point S2 indicating the maximum (minimum) value of the gradation at the end point position is stored. In this way, the extreme value Sp is calculated while sequentially storing the round base points sb, s, , S2, . . . for each unit process. Then, when the maximum search is completed for all the initial points Sll set in one small region (step N7), the next step is to compare the number N of small regions obtained by dividing the force-Kunt number 1 (step N8). The above operation is repeated while incrementing the count number 1 by 1 until the extreme value search is completed over the entire small area. Then, when all the base points S1, S7, S3, . When the number N matches the number N, the trajectory is recombined for each small area based on this memorized base point (step N9), and each small area is further arranged in 1.6 x 1.6 pieces to form one image screen. Configure (step N9). Then, C
The recombined extracted pattern is output to a display device such as an RT or a printer (step N10). As a result, the characteristic patterns obtained in the process of extreme value search are displayed, as shown in FIGS. 3 to 5, respectively. In these figures, the symbol sb indicates the initial search point, and sp indicates the extreme value. Figure 3 shows a peak-like pattern, Figure 4 shows a ring-like pattern, and Figure 5 shows a ridge-like pattern. This is a case where each pattern is extracted. Note that although this image extraction method has been described for a two-dimensional image in this example, it is possible to use a two-dimensional image, and the region can be divided arbitrarily. Furthermore, the search initial point may be selected systematically or randomly.

Also, if you repeat the search by trial and error depending on the search results, increasing or decreasing the number, you can obtain an accurate pattern.

(g) Effects As described above, according to the present invention, the display area of a digital image is divided into a plurality of predetermined small areas, a plurality of search initial points are set in each of the divided small areas, and each search Extrema search is performed from the initial point by applying the direct search method for extrema of the nonlinear function, and all base points indicating the maximum are memorized for each unit process in the process of extremum search, and the base points of each small region are Since the pattern of the digital image is extracted by displaying all the base points found by extreme value search from each search initial point, it is possible to accurately identify the structural features even in images with complex configurations. can be extracted into therefore,
Structural analysis of images becomes even easier, and it can be widely applied to diagnosis of medical digital images used in ultrasound diagnostic equipment and the like.

[Brief explanation of drawings]

The drawings show an embodiment of the present invention; Fig. 1 is a flowchart for explaining the method of extracting features from digital images; Fig. 2 is an explanatory diagram of the extreme value search method; and Figs. 3 to 5 show the features. FIG. 3 is a schematic diagram showing extracted patterns. sb...search initial point, S,, Sl, Sl...base point, Sp...extreme value. Applicant Shimadzu Corporation Representative Patent Attorney Oka 1) Kazuhide el Figure 5 (a) (b) Figure 3 (a) (b) Figure 4 (a) (b)

Claims (1)

[Claims] (1) Divide the display area of the digital image into a plurality of predetermined small areas, set multiple search initial points in each divided small area, and directly search for the extreme value of the nonlinear function from each search initial point. Perform extreme value search by applying the method, memorize all the base points that indicate the maximum for each unit process in the process of this maximum search, and find the base points by searching for the extreme value from each search initial point of each small region. A digital image processing method, characterized in that a pattern of the digital image is extracted by displaying all of the base points.