JPS5819102B2 - Object image extraction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an object image extraction device that extracts and recognizes a specific object image in remote sensing image analysis, microscopic image analysis, three-dimensional scene analysis, and the like.

It is something.

Typical conventional techniques for this type of object image extraction are:
Photographs are taken for each wavelength range using several types of filters for each wavelength range of light, and the film is sampled, quantized, and fed into a computer, and a specific object is extracted and recognized by software using the characteristics of the spectral composition of the specific object. This is the way to do it.

Alternatively, there is a method of performing the above process using analog technology.

This method focuses only on the characteristics of the distribution of concentration values in each wavelength region and recognizes a specific object solely based on the shape of its spectral distribution, so it is difficult to recognize other noise objects that have the same spectral distribution. There are drawbacks to extracting.

The present invention was made with the aim of eliminating the above-mentioned drawbacks of the conventional art, and it actively performs analog calculations between several spectral component images that best represent the spectral characteristics of a specific object. This provides an object image extraction device that extracts a specific object while minimizing the inclusion of noise objects.

The basic idea of the present invention will be explained using FIG.

1. A and B shown in FIG. 1 are different spectral component images in the jc wavelength region.

For example, A is a scene obtained through a red filter,
B is considered to be the same scene obtained through a blue filter.

Assuming that the object to be recognized is 1, instead of the faint shadow of noise object 2 at A, a depression like 3, a large force like 4, and S are formed in object image 1.

On the other hand, in B, instead of the entire object 1 appearing completely, the other noise object 2 also clearly appears.

In the present invention, first, a common component between A and B is extracted in the form of 5 in C, and then all components in B that are connected to 5 in C are extracted to form a specific object image 1. Take out only in a shape like D.

Here, the definitions of having a common basin and being connected are defined as satisfying certain conditions.

& If images A and B have a common component at pixels i and j, we can express that the density value A(i, j) and the density value B(i
, j) is less than or equal to a certain value.

Also, the component connected to 5 in C in B is the pixel 5 [
It can be defined as a component of B for which the absolute value of the difference with the density value in B is less than or equal to a certain threshold value.

The configuration of the device used in the present invention will be explained.

First, as shown in FIG. 2, the object to be imaged is imaged on the surface of the imaging plate using two filters having different passband wavelength bands.

In Figure 2, 6 is the object to be imaged, 7 is a semi-transparent mirror, 8.9
are filters with different passing wavelength bands, and 10.11 is an imaging plate in the imaging device.

To simplify the following explanation, it is assumed that the image pickup plate has the following properties.

In other words, it is assumed that one image pickup plate is a collection of photoelectric conversion elements densely arranged in a grid pattern, and each photoelectric conversion element has a potential holding function or a charge storage function.

Two-dimensional imaging devices with such functions have already been developed and are commercially available, so the above assumption is realistic.

In addition, images such as those shown in FIG. 1 A and H are shown on the imaging plate io in FIG.
'+i, respectively, and the image pickup plate 1 in FIG.
The image of 0 will be displayed as A1, and the image of image pickup plate 11 will be displayed as B.

As an example, assume that A and B have a common component pixel.

A(i, j)≧θa and B(i, j)≧θb and IA(
1, JC-B (i, j) is defined as satisfying the condition 1≦θ1.

To extract B pixels that meet this condition (analog)
A comparator may be used.

FIG. 3 shows a block diagram of two types of comparators and (analog) gate elements.

In the figure, the IN terminal is an input part, the OUT terminal is an output part, and θ and G are condition input parts.

The operation of the comparator 12 shown in FIG. 3a is shown in Table 1 below (
This is as shown.

The operation of the comparator 13 shown in FIG. 3 is as shown in Table 2 below.

The operation of the gate 14 shown in FIG. 3C is as shown in Table 3 below.

Note that in each of the above tables, 0 and 1 are logical symbols, and in an actual circuit, each has a specific level.

Using such a circuit element, B(i
, j) A circuit for extracting pixels is shown in FIG.

In the figure, 12 is a comparator shown in FIG. 3A, 13 is a comparator shown in FIG. 3B, and 14 is a gate shown in FIG. 3C.

Further, 15 is a circuit that outputs the absolute value of the voltage difference between the input terminals 16 and 17 to the comparator 13.

All circuit elements such as those illustrated in FIGS. 3 and 4 can be implemented using conventional circuit technology.

For each pixel of images A and B on the imaging plates io and ii shown in FIG. 2, the output potential 19 obtained from the gate 14 is manually applied to the grid array 21 of potential holding elements through the circuit shown in FIG.

This situation is shown in FIG.

In FIG. 5, 20 is a device having the same number of circuits as shown in FIG. 4 as the number of pixels of image A (or B), realizing the condition of having common components.

Reference numeral 21 denotes an element array that holds the potential of pixels of image B that satisfies this condition, and image C appears in this element array.

For pixels that do not satisfy the conditions, the corresponding pixel potential of image C is set to 0 for convenience.

The potential distribution pattern applied to the element array 21 in FIG. 5 is as shown in FIG. 1.

Next, in Figure 1 or Figure 5, image C is
A circuit for extracting all pixels connected to .

In order to define connectivity, consider the neighborhood of a certain pixel in terms of divisions as shown in FIG.

That is, pixel P. The neighboring pixels are P1 to P8.

Pixel Po is connected to its neighboring pixels Pkth, for example.

This means that the absolute value of the density value P of and the density value of P is less than or equal to a certain threshold value θ2.

A function is added to the element array 21 that allows the potential or charge to be moved (shifted) in the vertical and horizontal directions between each element of the element array 21 shown in FIG.

A sequential shift method is used in which elements that protrude as a result of shifting are moved to empty elements on the opposite side.

This situation is shown in FIG.

That is, FIG. 7 shows an example of a 4×4 charge arrangement, where a is the original state and b is the state after being shifted to the left.

Such shifts can be made any number of times.

The same applies to shifts in other directions. Image C in image B
To extract all pixels connected to 5 in
A control mechanism may be added to the circuit shown in FIG.

This is shown in FIG. In FIG. 8, 22 is a control circuit, in which a control signal 26 controls the shift of the element array 21, a control signal 23 controls switching of the input section of the image pickup plate 10, and a control signal 24 controls the threshold value to the threshold circuit group 20. Performs various controls such as settings.

The circuit elements shown in FIG. 4 included in the circuit group 20 are now used for extracting connected components in image B.

Each output of the image pickup plate 10 receives the shifted output of each of the five elements of the image C of the element array 21 by the control circuit 22.

The threshold values θa, θb, θ1 in FIG. 4 are reset by the control circuit 22 to the threshold value θCθbb, 4 for extracting connected components.

The state of each circuit element in the circuit group 20 in the case of connected component extraction is shown in FIG. 9 again.

The input/output relationship shown in FIG. 9 examines the connectivity between the circuit group 20 shifted by one pixel in the left direction and image B on the imaging board, and pixel B (i ,j)
A situation is shown in which the connectivity between C(i, j+1) and the left-shifted pixel C(i, j+1) of image C is examined.

When B(i, j)≧θbb and C(i, j+1)≧θC and l B(i, j)−C(i, j+1)l≦θ2, the (i, j) component of image B and ( It is determined that the component of (i, j+1) is connected, and the potential of the (i, j+1) pixel of the image B is output from the gate output 19 at the position of the (i, j) pixel of the image C.
transferred via.

After such texture has been applied to the entire circuit group 20,
For element array.

Then, the control circuit 22 executes a rightward shift by one pixel.

By this operation, the image B (i, j+1) component connected to the (i, j) component of image B is copied to the (i, j+1) component of image C.

The control circuit 22 performs the same processing as above for the right, top, bottom, upper left,
Run in the following order: bottom left, top right, bottom right.

As shown in Figure 6, (i, j
) The connectivity between a pixel and its neighborhood can be examined.

The above-described cycle is repeated under the control of the control circuit 22 until there is no change in the potential distribution pattern of each element of the element array 21.

With the circuit group and its functions described above, a potential distribution pattern (image D) corresponding to the specific object image, which is the final processing result, is obtained in the element array 21 shown in FIG.

In the above, the circuit operation of the device of the present invention has been described in the form of analog electrical signal processing, but this is only one example. After quantizing everything and converting it into a digital signal,
It is also possible to completely realize the above structure using only digital logic circuits.

Furthermore, conditions regarding connectivity and common components are just one example;
It is also easy to replace it with another conditional expression.

Using the device of the present invention, object recognition can be achieved in a shorter time, and furthermore, if current microcomputer, firmware technology, LSI technology, etc. are adopted, a device with better performance can be realized.

The device of the present invention can be applied to various fields such as three-dimensional object recognition, microscopic image analysis, and remote sensing image texture.

[Brief explanation of drawings]

FIG. 1 is a drawing for explaining the basic idea of the present invention, and A to D show images. FIG. 2 is an explanatory diagram of the human-powered portion of the device of the present invention. 3, 4 and 9 are block diagrams showing the basic circuit elements of the device of the present invention. FIGS. 5 and 8 are block diagrams showing the overall configuration of the apparatus of the present invention. FIG. 6 is a diagram for explaining the concept of neighborhood in continuity, and shows pixel arrangement. FIG. 7 is an explanatory diagram of the forward shift and shows the pixel arrangement. Note that the same reference numerals in the figures indicate the same or equivalent parts. 1: Object image, 2: Noise object, 3: Dent, 4: Hole, 5: Common component, 6: Object, 7: Semi-transparent mirror, 8, 9 two filters, 10
, 11: Image pickup plate, 12, 13: Comparator, 14: Gate, 15: Voltage difference output circuit, 16 to 19: Signal, 20
: circuit group, 21: element arrangement, 22: control circuit, 23-2
5: Signal.

Claims (1)

[Claims] 1. At least two filters with different wavelength passbands, at least two sets of imaging circuits each having a photoelectric conversion function, a potential holding function, and a voltage input switching function; a circuit for extracting a common component pixel of an image, an array circuit that holds the extracted common component and has a potential holding function by sequentially shifting vertically and horizontally under the control of a control circuit; Transferring the image held in the array circuit to one of the image pickup circuits under the control of a circuit, and repeating extraction of connected components in the other image pickup circuit until there is no change in the image held in the array circuit. The feature is that the structure is configured to perform. Object image extraction device.