JPH07220066A - Picture processor - Google Patents

Abstract

PURPOSE:To provide the picture processor which accurately extracts edges of an area in a picture by adding the gradation correction processing for each partial area between a picture input means and in edge extraction processing means. CONSTITUTION:The input picture from a picture input means 1 is divided into plural picture blocks by a picture dividing means 2, and picture information of each picture block is obtained by a picture information extracting means 3, and gradation correction is performed by a gradation correction means 4. This means preliminarily learns the relations between picture information and nonlinear gamma curves and discriminates picture information based on the learnt result and corrects the objective picture block by a selected non-linear gamma curve. An edge extracting means 5 extracts edges from the corrected picture block, and a picture composing means 6 composes respective picture blocks into one picture again, and it is outputted from a picture output means 7. By this constitution, edges in the area of shadow which is difficult to extract in the conventional method can be extracted. Since nonlinear gamma curves are selected in accordance with learnt results, the stable processing is always easily performed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus, and more particularly to an image processing apparatus having an edge extraction processing function.

[0002]

2. Description of the Related Art Conventionally, edge extraction of an object in an image is performed by T
The image signal from the V camera is A / D converted and directly input to the edge extracting means using a differential operator or the like to extract the edge. FIG. 11 shows a block diagram of a conventional image processing apparatus that performs edge extraction. Image signal after A / D conversion input from the image input unit 111 (for one image)
Is sent to the edge extraction means 112. The edge extraction means 112 includes a differentiator 112a and a threshold processor 112b,
The image input unit 1 is configured by the thinning processing unit 112c.
The image signal sent from 11 is first inputted to the differentiator 112a. The differentiating unit 112a applies a differentiating operator in parallel to each pixel of one image to differentiate the density value of each pixel. As an example of the differential operator, FIG. 12 shows the Previt operator and the Sobel operator. The differentiated image signal is sent to the threshold processing unit 112b, and the threshold processing unit 112b performs threshold processing on each pixel. That is, if the differential density value of each pixel is larger than a certain value, 1 is output, and if it is smaller, 0 is output. Threshold processing unit 1
The image signal binarized by 12b is the thinning processing unit 112c.
To extract the center line of line width 1. The image thinned by the thinning processing unit 112c, that is, the edge image is output from the image output unit 113.

[0003]

However, since the dynamic range of the TV camera is narrower than that of a natural subject, when the edge extraction is performed with the above-described conventional configuration, the periphery of the shadow area is erroneously extracted, There is a problem that the edges in the shadow area cannot be extracted correctly. A specific example is shown in FIG. FIG. 13A is an indoor image viewed by the human eye. Outside light enters through the window 131, and the inside box 1 is exposed to the outside light.
32 has a shadow. The window 131 is very bright, the interior is medium and the shadow of the box 132 is dark. However, since the human eye has a wide dynamic range, it is possible to recognize the scenery inside the shadow area or outside the window. On the other hand, FIG. 13B is an indoor image captured by the TV camera. As mentioned above, the TV camera has a narrow dynamic range, so the window 1
When the aperture is adjusted to 31, the room becomes very dark and the box 132 in the shadow area cannot be clearly recognized.
On the other hand, when the aperture is adjusted to the room, the brightness of the area of the window 131 is saturated and the area looks white, and the scenery outside the window 131 cannot be recognized. When the edge extraction is performed on the image in such a state as it is, the shadow of the box 132 is erroneously extracted as an edge as shown in FIG. 13C, or the scenery outside the window 131 or the shadow portion of the box 132 is extracted. Often, the edge of cannot be extracted.

In order to solve the above-mentioned problems of the conventional edge extraction method, the present invention adds a partial gradation correction process between the image input means and the edge extraction processing means. An object is to provide an image processing device that can accurately extract edges.

[0005]

An image processing apparatus according to the present invention comprises an image dividing means for dividing an image input from an image input means into a plurality of image blocks, and one or more kinds of image signals from each image block. Image information extracting means for extracting the image signal of, the tone correcting means for enhancing the edge to be extracted by performing the tone correction suitable for each image block based on the image information obtained by the image information extracting means, and the tone It has an edge extracting means for extracting an edge from each image block after correction, and an image synthesizing means for synthesizing the images after the edge extraction and re-synthesizing them into one image.

[0006]

In the image processing apparatus configured as described above, when one image signal is input to the image dividing means, the image signal is divided into a plurality of image blocks, and the image information extracting means makes an image for each image block. Send a signal. The image information extraction means extracts and calculates one or more types of image information from the image signal of each image block, and sends it to the gradation correction means.
The gradation correction unit performs gradation correction suitable for each image block based on the image information and emphasizes the edge to be extracted.
The edge extracting means extracts an edge from the image signal of each image block after gradation correction by using a differential operator or the like, and sends the result to the image synthesizing means. The image synthesizing unit synthesizes the edge image for each image block sent from the edge extracting unit into one edge image, and outputs it from the image output unit.

As described above, gradation correction is performed before edge extraction to form an image with a wide dynamic range.
Edges that cannot be extracted by the conventional method can be extracted. For example, by performing gradation correction on the image of FIG. 13B to emphasize the edge of the object in the shadow area, the image of FIG.
As in (d), the edge of the box can be correctly extracted. In addition, although a normal image has both a bright area and a dark area, in order to correct the dark area brightly without affecting the bright area, one kind of gradation correction may be applied to one image. Is insufficient. The image processing apparatus of the present invention is
To correct bright areas as they are and dark areas as bright, divide the image into multiple blocks, perform gradation correction appropriate for each block, and extract edges correctly from both bright and dark areas. be able to.

[0008]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of an image processing apparatus according to the present invention. The configuration will be described below together with the operation. In FIG. 1, 1 is an image input means. An image signal for one image input from the image input unit 1 is divided into a plurality of image blocks by the image dividing unit 2. The image signal of each divided image block is sent to the image information extracting means 3. The image information extracting means 3 extracts one kind or plural kinds of image information from each image block and sends it to the gradation correcting means 4. The gradation correction unit 4 performs gradation correction on the image signal in the image block based on the image information of each image block. The gradation correction unit 4 includes a learning type non-linear gamma curve selection unit 42 and a non-linear gamma correction unit 4.
1 and. The image information obtained by the image information extracting means 3 is sent to the learning type non-linear gamma curve selecting section 42.
The learning-type non-linear gamma curve selection unit 42 has learned the relationship between the image information of various states and the non-linear gamma curve to be used for gamma correction in advance, determines the image information based on the learning result, and uses the non-linear gamma curve to be corrected. Gamma curve (0 to
Select 7). The result of the selected gamma curve and the image signal of the correction target area in the image block are sent to the non-linear gamma correction section 41, and the non-linear gamma correction section 41 uses the selected non-linear gamma curve to gamma-correct the correction target area. Then, the corrected image signal is output. The edge extraction unit 5 extracts an edge from each image block corrected by the gradation correction unit 4. The edge image of each image block obtained by the edge extracting means 5 is sent to the image synthesizing means 6, and the image synthesizing means 6 synthesizes the sent edge image of each image block into one image. The combined image is output from the image output means 7.

FIG. 2 shows an example of a dividing method in the image dividing means 2. As shown in FIG. 2A, one image is divided into several image blocks, and the entire image block (m × n in the figure is calculated based on the image information obtained from the image signal in the image block).
Pixel) is used as a correction target area, and when an image block is extracted so that the image blocks overlap each other as shown in FIG. 2B, based on the image information obtained from the image signal in the image block. An example is a case where the correction process is applied to the central region (k × l pixel in the figure) in the image block. That is, the division method of FIG. 2B can also consider information around the correction target area. Figure 2 (b)
In the case of, k = m / 2 and l = n / 2. Figure 3 (a)
FIG. 3 is a block diagram of an embodiment of the image dividing means 2 in the case of FIG. The image block extraction unit 21 extracts an image block having a size of m × n pixels with a certain point (x, y) as the center of gravity, and sends the image signal in the image block to the image information extraction means 3. In the initial state, both x and y are 0. Further, the image block of size k × l pixels is sent to the gradation correction means 4. After extraction, the values of x and y are updated by the center of gravity setting unit 22. A specific example of the center of gravity setting unit 22 is shown in FIG. X and y sent from the image block extraction unit 21 are route selection units 221a and 22a, respectively.
1b. When the size of one image is X × Y, the path selector 221a sends x to the adder 222a if x <X, adds k and outputs it. x = X
If so, x is sent to the assigner 223a, 0 is assigned to x, and output. The path selector 221b outputs y as it is if x <X. If x = X, the route selector 221
send to c. In the route selector 221c, if y <Y, y is sent to the adder 222b, l is added and output. If y = Y, y is sent to the assigner 223b, -1 is assigned to y, and output. The x ′, y ′ output from the center of gravity setting unit 22 is sent to the switch 23, and if y ≠ −1, x ′, y ′ (newly set center of gravity) is output to the image block extracting unit 21. With the above configuration, it is possible to extract each image block by moving the center of gravity of the image block in the raster scan direction so that the image blocks for image information extraction overlap each other.

FIG. 4 shows an embodiment of the image information extraction means 3. Here, as the image information, the brightness values of the pixels in the image block are obtained, and these are set to high brightness, medium brightness,
The frequency value is divided into three levels of low brightness to be image information. In FIG. 4, reference numeral 31 denotes a brightness calculation unit, which is used for the R of each pixel of the image block sent from the image dividing unit 2,
A luminance signal Y is created from the G and B signals. Reference numeral 32 denotes a brightness level value storage unit, which is a boundary value of brightness levels (a in the figure,
Remember b). Reference numeral 33a is a brightness level comparator, which compares the brightness value sent from the brightness calculation unit 31 with the value a stored in the brightness level value storage unit 32, and if the brightness value Y is larger, sets 1 to the counter 35a. Add, and if smaller, send 1 to AND circuit 34. 33b is also a brightness level comparator, and similarly to 33a, compares the value b stored in the brightness level value storage unit with the brightness value Y, and if the brightness value Y is larger, sends 1 to the AND circuit 34. If it is smaller, 1 is added to the counter 35c. The AND circuit 34 adds 1 to the counter 35b when the outputs from the brightness level comparators 33a and 33b are both 1. This makes the first
Three types of frequencies: brightness larger than level value a (high brightness), brightness between first level value a and second level value b (medium brightness), brightness smaller than second level value (low brightness) Can be obtained from the counters 35a, 35b, 35c, respectively. Note that the luminance is divided into three levels here, but it can be divided into more levels by connecting a luminance level comparator and an AND circuit in parallel.

The operation of the image processing apparatus configured as described above will be described below.

First, the overall processing flow will be described. The image signal for one image (the size of the image is X × Y) input from the image input unit 1 of FIG. 1 is divided into image blocks by the image dividing unit 2. The operation of the image dividing means will be described below with reference to FIG. The input image signal is sent to the image block extraction unit 21. The image division block extraction unit 21 extracts two types of image blocks from the image signal. One is an image block 1 (m × n pixels) sent to the image information extraction means 3, and the other is an image block 2 (k × l pixels) sent to the gradation correction means 4, where k = m.
/ 2, l = n / 2). These two blocks are concentric rectangles, and the area of image block 1 is
Is four times. That is, for the purpose of performing the gradation correction in consideration of the surrounding situation, by calculating the image information serving as the reference for the gradation correction not only in the area of the gradation correction target but also in the surrounding pixels. There is. Extraction of an image block starts from the upper left of the image (center of gravity (0,0)).
After extracting each image block, the position coordinates (x,
y) is sent to the center of gravity setting unit 22. The center-of-gravity setting unit 22 shifts by k or l in the raster scan direction. For example, if the input centroid is (0,0), the adder 22
In 2a, k is added to the x coordinate and (k, 0) is output. x
During <X, k is similarly added to the x coordinate. When the center of gravity moves to (0, X), the assigner 223a substitutes 0 for the x coordinate, and the adder 222b adds 1 for the y coordinate,
Outputs (1, 0). When the center of gravity becomes (X, Y) by repeating these processes, the substituting unit 223c shifts the y coordinate to −.
Substitute 1 If y = −1, the switch 23 releases the route and ends the image division process. In this way
The image blocks are sequentially extracted so that the image blocks 1 overlap each other, and the image block 1 and the image block 2 are sent to the image information extraction means 3 and the gradation correction means 4, respectively.

The image information extracting means 3 extracts information indicating the state of the image block from the image block of size m × n pixels sent from the image dividing means. In the present embodiment, the brightness of each pixel in the image block is calculated, and the brightness is calculated as 3
Four levels (high brightness, medium brightness, and low brightness are used, and the frequency of each level is used as image information. This will be described below with reference to Fig. 4. The brightness calculation unit 31 sends an image block from the image dividing unit 2. The luminance signal Y is created from the R, G, and B signals of each pixel of 1. The luminance signal Y is obtained by (Equation 1) for the television signal of the NTSC system.

[0015]

[Equation 1]

The calculated brightness value is used as a brightness level comparator 33.
In a, the boundary value a between the high brightness level and the middle brightness level
And if Y ≧ a, the high brightness level counter 35
Add 1 to a. In the brightness level comparator 33b, the boundary value b between the middle brightness level and the low brightness level is compared with the brightness Y, and if Y <b, 1 is added to the low brightness level counter 33c. If Y <a and Y ≧ b, 1 is added to the intermediate brightness level counter 33b. In this way, the pixels in the image block are classified to obtain three types of image information. Since these three types of image information can be expressed by the relational expression (Equation 2), the gradation correction unit 4 has at least 2
It suffices to send the type of information.

[0017]

[Equation 2]

The image information obtained by the image information extracting means 3 and the image block divided by the image dividing means 2 are gradation correction means 4.
Sent to. As described above, the image information obtained by the image information extracting unit 3 is the learning type non-linear gamma curve selecting unit 4
2 is sent. Learning type non-linear gamma curve selection unit 42
Has learned the relationship between image information in various states and a non-linear gamma curve to be used for gamma correction in advance. For example,
Relationship between image information obtained from totally dark image block and non-linear gamma curve that corrects overall brightness, relationship between image information obtained from totally bright image block and non-linear gamma curve that corrects only dark pixels to bright Have been learned in advance (learn more about learning later). Judging image information based on the result of learning in this way,
Select the non-linear gamma curve (0-7) to be used for correction. The result of the selected gamma curve and the image signal of the correction target area in the image block are sent to the non-linear gamma correction section 41, and the non-linear gamma correction section 41 uses the selected non-linear gamma curve to gamma-correct the correction target area. Then, the corrected image signal is output. In this way, by preliminarily learning in the learning type non-linear gamma selection unit 42, the optimum non-linear gamma curve can be selected from these according to the state of the image. Note that FIG. 9 shows an example of a non-linear gamma curve used for correction, and the gamma curve in the figure can be obtained by (Equation 3).

[0019]

[Equation 3]

Each image block corrected by the gradation correcting means 4 including the learning type non-linear gamma curve selecting section as described above is sent to the edge extracting means 5. Edge extraction means 5
Can use a conventional method as shown in FIG. After the edge extraction unit 5 has performed edge extraction, the edge image of each image block is combined into one image by the image combination unit 6 and output from the image output unit.

A concrete example of the above image processing of the present invention is shown in FIG.
Shown in. FIG. 10A shows an original image. This image is window 1
The image is taken with the aperture set to 01, and the room is very dark, and the box 102 in the shadow area cannot be clearly recognized. When this image is divided into four equal parts by the image dividing means 2 and the state of each image block is examined, it can be seen that the lower right image block of the image is entirely dark and the edge of the box 102 cannot be clearly recognized. Therefore, as shown in FIG. 10B, the image block is corrected to be bright as a whole to emphasize the edge of the box 102 in the shadow portion. Figure 10
(C) is each image block after image correction (FIG. 10 (b))
This is the result of edge extraction processing performed on the. As described above, the gamma correction is performed on the lower right image block and the box 10
Since the edges of No. 2 are emphasized, if the edge extraction processing is applied to these, the edge of the shadow portion can be correctly extracted as shown in FIG.

In this way, gradation correction is performed before edge extraction to form an image with a wide dynamic range.
Edges that cannot be extracted by the conventional method can be extracted. Also, by dividing the image into multiple blocks and performing gradation correction suitable for each block, it is possible to correct bright areas as they are and bright areas as dark areas. Can be extracted correctly.

Next, the learning type non-linear gamma curve selecting section 42.
The operation of will be described with reference to the following figures.

FIG. 5 shows a learning type non-linear gamma curve selecting section 42.
Is an example of the unit recognition unit 421 that constitutes the network of FIG. Signal input unit 4 in FIG.
The unit 211 is composed of the quantizer 4212, the quantizer 4212, and the route selection unit 4213. The signal input unit 4211 inputs the feature data to be recognized, which is input via the signal input terminal 4211a, to the quantizer 4212. The quantizer 4212 quantizes the input feature data and inputs the quantized value to the route selection unit 4213. 4213a
Is a path input terminal, and 4213b1 and 4213b2 are path output terminals, which are connected to each other when the unit recognition units are combined to form a network. The route selection unit 4213 includes a quantizer 4212.
Based on the value input from the path input terminal 4213a
And the route output terminal 4213b1 or 4213b2. Structure storage unit 4
Reference numeral 214 stores the quantization range of the quantizer, the number of quantizations, the number of route input terminals and the number of route output terminals of the route selection unit. The internal state storage unit 4215 stores the average, variance, and total number of inputs of signals input so far as the internal state of the unit recognition unit. Duplication section 4
216, when the internal state stored in the internal state storage unit reaches a certain value, divides the quantization range of the unit recognition unit stored in the structure storage unit into two, and copies the unit recognition unit. Is configured to be created.

The unit recognition unit can be configured as follows in addition to the above. (B) shows the route selection unit of (a), which includes a route input unit 4213a having one route input terminal 4213a1 and two route output terminals 4213b1.
4213b2 and a path output unit 4213b and a switch 4213c. The switch 4213c is configured to switch the connection method between the route input terminal 4213a1 of the route input unit 4213a and the route output terminal 4213b1 or 4213b2 of the route output unit 4213b based on the value input from the quantizer 4212. (C) shows the route selection unit 4213 of (a) including a route input unit 4213a having one route input terminal 4213a1 and two route output terminals 421.
Path output unit 4213b having 3b1 and 4213b2
And the path load part 4213c. Loads 4213c1 and 4213c
2 is a route output terminal 4213b of the route output unit 4213b.
1 and 4213b2 are weights to be added to the path output signals to be output to 4213b2, and the weighter 4213c0 changes these weights according to the value output from the quantizer 4212. Load 42
13c1 and 4213c2 weight the route signal input from the route input unit, and the route output unit 4213b outputs the weighted route signal to the route output terminals 4213b1 and 4213b2.

The operation of creating a copy of the unit recognition unit configured as described above will be described below with reference to FIG. In order to actually make a copy, it is necessary to prepare an empty unit recognition unit that is not used and has no connection with other unit recognition units. A signal having a value of 1 to 10, for example, is input from the signal input terminal 4211a of the signal input unit 4211 of the unit recognition unit serving as the duplication source, and the quantizer 4212 has a quantization range of 1
It is set to ~ 10. The internal state storage unit 421421 calculates and stores the average, variance, and the number of inputs of data signals that are sequentially input each time a data signal is input. And
When the product of the number of inputs and the variance exceeds a certain value, the copy creation unit 4216 causes the structure storage unit 4214 to store the quantization range of the quantizer, the number of quantizations, and the route selection unit 4213. By referring to the number of route input terminals and the number of route output terminals, the information of this unit recognition unit is copied to an empty unused unit recognition unit, and the unit recognition unit including the connection with other unit recognition units is copied. Make an exact duplicate. At that time, divide the required quantization range from 1 to 10,
For example, set the quantization range of the quantizer of the base unit recognition unit to 1 to 5 and the quantization range of the quantizer of the duplication unit recognition unit to 6 to 10, A copy is created so that the functions are shared with the unit.

FIG. 6 shows an embodiment of the threshold processing unit 422. 4221 route input unit.
Reference numeral 4222 denotes an adder, which has a plurality of path input terminals
Input signals from 42211n (n is an arbitrary natural number) are added. The threshold processor 4223 performs threshold processing on the signals added by the adder 4222, and determines how to output to the path output terminal 4224.

FIG. 7 shows a learning type non-linear gamma curve selecting section 42.
FIG. 1 shows an embodiment of the present invention, in which unit recognition units are interconnected in a multi-layered hierarchical form to form a network. The unit recognition units n11 to n12, n21 to n24, and n31 forming the first layer, the second layer, and the third layer.
5 to n38 use, for example, the unit recognition unit shown in FIG. 5A, and as described above, the route selection unit 4 is used.
213 includes a route input unit 4213a having one route input terminal 4213a1 and two route output terminals 4213b.
1 and 4213b2, and a route output unit 4213b and a route selection unit. Further, the units n41 to 43 of the fourth layer use the threshold value processing unit shown in FIG. 6, and the threshold value processing unit has a path input section having a plurality of path input terminals as already described. 4221
And an adder 42 for adding the input signal from the path input terminal
22 and a thresholder 4 for thresholding the path signal
223 and the route output unit 4224. The learning type non-linear gamma selection processing unit shown in FIG.
Based on three types of characteristic data (information obtained by the image information extraction means) consisting of two types, eight types of gamma curves are classified and recognized.

Next, the learning operation of the learning type non-linear gamma curve selecting section 42 shown in FIG. 7 will be described.

The unit recognition units n11 and n of the first layer
First, 1 is given as the route signal to the 12 route input terminals. The first series of feature data of the recognition object is input to the signal input terminals of the quantizers of these units. (In the case of this figure, the two first feature data are respectively input to two unit recognition units.) These first feature data are quantized by the quantizers n11 and n12. And based on this quantized value, the paths p11 and p
12 is selected by the switch shown in FIG.
The route signal "1" is sent to the route input terminals of the unit recognition units n22 and n23 of the layer. A second series of feature data of the recognition object is input to the signal input terminals of the quantizers of these units. (In the case of this figure, 2
The second feature data are input to the two unit recognition units n22 and n23, respectively. ) These second characteristic data are quantized by the quantizers of n22 and n23, and the paths p21 and p22 are selected by the switch shown in FIG. 5B based on the quantized values. To the path input terminals of the unit recognition units n34 and n36 of the fourth layer,
The route signal "1" is sent. At the signal input terminals to the quantizers of these units, a teacher input signal indicating which of the three items the recognition object belongs to, that is, n41
Input a signal indicating which one of n48 is selected. (In the case of this figure, the teacher input signal is input to the two unit recognition units n34 and n36.)
The quantizers of n34 and n36 quantize, and based on the quantized values, the paths p31 and p32 are calculated as shown in FIG.
It is selected by the switch shown in.

Next, the operation of the entire apparatus when each unit recognition unit of the learning type recognition and judgment apparatus according to the present invention makes a copy will be described.

When the unit recognition units of the layers up to the fourth layer make a copy, it is necessary to copy the unit recognition unit including all the unit recognition units connected to the lower layer of each unit. FIG. 8 shows a state of duplication including all the unit recognition units connected to the lower layer. Duplicate detection unit S
Detects all the unit recognition units that have performed the duplication operation, and gives an instruction to perform duplication including all the unit recognition units connected to the lower layer of each unit that creates the duplication.
For example, as shown in FIG. 8, the unit 81 is duplicated including the unit recognition units 811 and 812 connected to the lower layer, so that the duplicated unit recognition unit 82 and the unit recognition units 821 and 822 of the lower layer are duplicated. It is created. Therefore, not only is network learning performed on the input data, but the network structure can be automatically changed, constructed, and self-organized adaptively according to the input signal. is there.

Next, the recognition operation will be described.

In the recognition process, the internal state of the unit recognition unit stored in the internal state storage unit does not change, and therefore, the copy creation unit does not operate.

In the unit recognition units n11, n12 and n22, n23 from the first layer to the second layer, the quantizer quantizes the input feature data in exactly the same way as the learning operation.
Based on this, the switches are changed to the paths p11 and p1.
2 and p21 and p22 are sequentially selected. In the recognition operation, the teacher input signal is not input to the signal input terminals of the unit recognition units n34 and n36 of the third layer. Therefore,
The states of the switches at the time of learning are held, the routes p31 and p32 are selected according to the states of these switches, and the route signal "1" is sent to the route input terminal of the unit recognition unit n42 of the fourth layer. To be The adder in the path input section of this unit adds the path signals input through the paths p31 and p32. The signal "1" is input to the signal input terminal of the signal input unit, the quantizer quantizes it, and the route selector enables the route output. (When the signal "0" is input , The path selector switches to a state in which the path is not output.) The added path input signal is sent to the path output device. The path output device thresholds this signal and outputs it to the path output terminal. Therefore, if the value of the added signal is larger than a certain threshold value, an output is made. In this way, the classification and recognition judgment of the recognition target object are performed based on the input recognition target object feature data. It can be done. A sigmoid function, a step function, or the like can be used as the function for performing the threshold processing. As described above, in the copy creation process of the learning type non-linear gamma curve selection unit, the internal state of each unit recognition unit changes according to the input data, and the copy creation unit operates and each unit recognition unit operates. , It is possible to automatically change, construct, and self-organize the network structure adaptively according to the input signal, and generalize to new unlearned data input. It is also excellent in sex.
Also, in the learning process, various feature data of the object are input to the signal input section of each unit recognition unit constituting the multilayer hierarchical network, and the unit recognition units are connected to each other according to the output of the quantizer. It suffices to switch the coupling paths of the above and determine the selection path to the lowermost layer by inputting the teacher signal in the frontmost layer of the lowermost layer. Therefore, the learning process can be performed very quickly. Further, in the recognition process, various feature data of the object are input to the signal input section of each unit recognition unit constituting the multilayer hierarchical network, and the unit recognition units are connected to each other according to the output of the quantizer. The recognition result can be obtained only by switching the connection paths of the lowermost layer and by determining the selection path to the lowermost layer based on the connection path set in the learning process in the lowermost layer. Based on the result, the recognition process can be performed very fast.

As described above, the image processing apparatus of the present invention is characterized by performing the non-linear gamma correction process as a pre-process of the conventional edge extraction process, thereby improving the dynamic range of the image. It is possible to extract edges that cannot be extracted by conventional processing, such as edges of objects in shadow areas. Also, instead of performing a fixed gamma correction on one image, the image area is divided into a plurality of blocks, and a non-linear gamma curve suitable for that block is selected and corrected to suit the detailed situation. Correction processing can be performed, and a more accurate edge image can be obtained from these corrected images. Furthermore, by using the learning type non-linear gamma curve selection unit, high-speed learning and recognition are possible, and the user does not need to determine the relationship between the image information and the non-linear gamma curve by trial and error.

Each means of the present invention can be realized by software using a computer, or can be realized by using a dedicated hardware circuit having these functions.

[0038]

As is apparent from the above description, the image processing apparatus of the present invention includes an image dividing means for dividing one image signal into a plurality of image blocks, and one or a plurality of image signals in the image block. Image information extraction means for extracting seed image information, gradation correction means for performing optimum gradation correction on the image block based on the image information, and edge for edge extraction for the image signal after gradation correction By providing the extracting means and the image synthesizing means for re-synthesizing the image signal after the edge extraction into one image, the dynamic range of the image is improved in detail, and the edge of the object in the shadow area is conventionally used. It is possible to extract edges that cannot be extracted by the processing of.

Further, the relationship between the image information and the non-linear gamma curve is formed by learning in advance, and the user judges the relationship between the image information and the non-linear gamma curve by learning. There is no need to make a decision by trial and error, and the processing is very easy.

[Brief description of drawings]

FIG. 1 is a diagram showing an embodiment of an image processing apparatus according to the present invention.

FIG. 2 is a diagram showing an example of an image dividing method by the image dividing means of FIG. 1 (a) An example in which an image block for extracting information and an image block for performing gradation correction are the same (b) Example in which the image block for extracting information and the image block for performing gradation correction are different

3A is a diagram showing an embodiment of the image dividing means of FIG. 1 (in the case of FIG. 2B). FIG. 3B is a diagram showing an embodiment of the center of gravity setting unit of FIG. 3A.

FIG. 4 is a diagram showing an embodiment of the image information extraction means of FIG.

5 (a) is a diagram showing an embodiment of a unit recognition unit forming the network of the learning type non-linear gamma selection unit of FIG. 1 (b) is a route input unit and a route output unit of the route selection unit of FIG. (C) A diagram showing an example in which the route selection unit in FIG. 10A is configured by a route input unit, a route output unit, and a route load unit.

FIG. 6 is a diagram showing an embodiment of a threshold processing unit constituting the network of the learning type non-linear gamma selection unit of FIG.

7 is a diagram showing an embodiment of the learning type non-linear gamma selection unit of FIG.

8 is a diagram showing the network replication operation of the learning-type non-linear gamma selection unit in FIG. 1 during learning.

FIG. 9 is a diagram showing an example of a non-linear gamma curve used in the gradation correction unit of FIG.

FIG. 10 is a diagram showing a specific example of image processing according to the present invention.

FIG. 11 is a general configuration diagram of a conventional edge extraction device.

FIG. 12 (a) is a diagram showing a Prewitt operator. (B) is a diagram showing a Sobel operator.

FIG. 13 is a diagram showing a specific example of conventional edge extraction processing.

[Explanation of symbols]

 1 image input means 2 image division means 21 image block extraction section 22 center of gravity setting section 221a to 221c route selectors 222a to 222b adders 223a to 223b substituter 3 image information extraction means 31 brightness calculation section 32 brightness level value storage section 33 brightness Level comparator 34 AND circuit 35 Counter 4 Gradation correction means 41 Non-linear gamma correction section 42 Learning type non-linear gamma curve selection section 421 Unit recognition unit 4211 Signal input section 4211a Signal input terminal 4212 Quantizer 4213 Path selection section 4213a Path input section 4213a1 Path input terminal 4213b Path output section 4213b1 to 4213b2 Path output terminal 4213c Switch 4213d Path load section 4213d0 Loader 4213d1 to 4213d2 Load 4214 Structure storage section 4215 Internal state storage section 4216 Duplicate creation section 422 Threshold processing unit 4221 Path input section 42211 to 4221n Path input terminal 4222 Adder 4223 Threshold value processor 4224 Path output section 5 Edge extraction means 6 Image combining means 7 Image output means 81 units Tsu DOO 82 replicate 101 window 102 Box 111 image input unit 112 the edge extracting unit 112a differentiating section 112b thresholding unit 112c thinning processing unit 113 an image output unit 131 a window 132 boxes

Front page continued (72) Inventor Susumu Maruno 1006 Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd.

Claims (5)

[Claims] 1. An image input unit for inputting an image, an image dividing unit for dividing the image input from the image input unit into a plurality of image blocks, and an image for each image block sent from the image dividing unit. Image information extraction means for extracting at least one type of image information from the signal, gradation correction means for performing gradation correction on the image signal of the image block based on the image information, and correction by the gradation correction means An edge extracting means for extracting an edge from each of the image blocks, an image synthesizing means for re-synthesizing the edge image extracted by the edge extracting means into one image, and an image for outputting the image synthesized by the image synthesizing means. An image processing apparatus comprising: an output unit. 2. The image dividing means is a coordinate (x,
y) is the center of gravity, and an area of m × n pixels and an area of k × l pixels (where k ≦ m, l ≦ n) are extracted as image blocks, and the image block of m × n pixels is extracted as the image information. Means for sending the image block of k × l pixels to the gradation correcting means, and a center of gravity setting unit for calculating the center of gravity of the image block to be extracted next. The image processing apparatus according to claim 1, wherein the output of the setting unit is input to the image block extracting unit to repeatedly perform the image block extracting process. 3. The gradation correction unit forms a relationship between the image information sent from the image information extraction unit and a non-linear gamma curve to be used for gradation correction by learning, and the image information is based on the learning. A learning-type non-linear gamma selecting unit that determines the non-linear gamma curve to be used for the gradation correction and a non-linear gamma correction that corrects the image signal using the non-linear gamma curve selected by the learning-type non-linear gamma selecting unit. The image processing apparatus according to claim 2, further comprising: 4. A network is formed by combining unit recognition units in a multi-layered structure, and an adder for adding input signals from at least a plurality of path input terminals and the input terminals and an output signal of the adder are thresholded. A thresholding unit constituted by a thresholding unit for value processing and a path output terminal for outputting the output of the thresholding unit is arranged at the lowermost layer and is located at the front layer of the lowermost layer of the network. A learning type non-linear gamma selection unit is configured by providing a teacher signal input terminal for inputting a teacher signal to the signal input unit of the unit recognition unit, and quantization is performed according to the input signals from the signal input unit and the signal input unit. A unit unit composed of a quantizer, a single or a plurality of path input terminals, a single or a plurality of path output terminals, and a path selection unit for selecting a path according to the output of the quantizer. , The average of the structure storage means for storing the number of ranges and quantization of the quantization of the quantizer as the quantization constant input signal to the unit unit, dispersing,
The unit recognition unit includes an internal state storage unit that stores the number of inputs as an internal state, and a duplication unit that divides the quantization range into two based on the stored internal state and creates unit units before division. The image processing apparatus according to claim 3, wherein the image processing apparatus comprises: 5. The image information extraction means compares the luminance signal extracted by the luminance signal of the image signal of each image block with the luminance signal extracted by the luminance extraction section with a level value within a predetermined range to obtain 1 or The image processing apparatus according to claim 1, comprising a plurality of brightness level comparators that output 0, and a plurality of counters that count the outputs of the brightness level comparators.