JP2858530B2 - Edge enhancement device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an edge enhancing apparatus for an image, and more particularly to an apparatus for enhancing an edge in a low-quality image whose position, shape, and number are not known in advance and whose direction differs for each pixel.

[0002]

2. Description of the Related Art Conventionally, as a method of enhancing an edge by image processing, a method of performing a filtering process by a local operator such as Laplacian / Gaussian to detect an edge and enhancing the detected edge portion in an original image (“Takagi , Supervised by Shimoda “Image Analysis Handbook”
pp. 550-560, pp. 564-567,199
1. ") And a method of detecting and enhancing edges by Hough transform (" Ballard, DH (Ballard, D.E.
H. ), Pattern Recognition (Pattern)
Recognition), 13, No2, pp. 11
1-112, 1981. , Duda, R.A. O. , Heart, p. E. FIG. (Duda, RO and Hart,
P. E. FIG. ) Communication of ACM (Com
m. ACM), 15, No, 1, pp. 11-15, 1
972. , Duda, R.A. O, heart, P. E. FIG. (Du
da, R .; O. and Hart, P.M. E. FIG. ), "Pattern Classification and Scene Analysis (Pattern Classification and Scene Analysis)
and Scene Analysis) ", John Wiley and Sons
& Sons), 1973. "), A method using the relaxation method (" Zucker, SW, Humme, A., Hammer, RA, Rosenfeld, A ").
l, R. A. , And Rosenfeld, A .; ),
IEEE Transaction (IEEE Tran
s. ), Vol. C-26, pp. 394-403,
1977. Pereg, S.M. Rosenfeld, A .;
(Peleg, S. and Rosenfeld,
A. ), IEEE Transactions (IEEE Tra)
ns. ), Vol. SMC-8, pp. 548-55
4, 1978. Shahater, B.S. J. , Lev, A. ,
Zucker, S.M. W. Rosenfeld, A.S. (Scha
chter, B .; J. Lev, A .; , Zucker,
S. W. , And Rosenfeld, A .; ), Ib
id. vol. SMC-7, pp. 813-816,
1977. "). The method using a local operator is susceptible to noise and needs to be smoothed in advance. However, there is a problem that the structure with fine edges is lost by the smoothing. The method using the Hough transform is relatively resistant to noise, but the shape and number of edges must be known in advance. Even when using the relaxation method, it is necessary to know the function form of the edge in advance.

[0003] In a situation where the S / N ratio is relatively low, the direction of each pixel is different, and the shape and the number of edges are not known in advance. , Or by applying a local operator such as Laplacian / Gaussian to obtain the direction of the edge, smooth it in the direction of the ridge line, perform the second derivative in the direction perpendicular to it, and use it as the weight image. (Supervised by Takagi and Shimoda, "Image Analysis Handbook", p.549, 1991.8) is known to be effective. However, the method using a local operator has the above-mentioned problems. The most effective method is to use the gradient of a luminance surface. However, this method has the following problems at the stage of obtaining the edge direction.

[0004]

[Problems to be solved by the invention]

(1) In the situation of low S / N, the window for smoothing must be enlarged, and in the case of an edge whose shape fluctuates greatly, for example, a saw-like edge, the fluctuation cannot be accurately followed.

(2) When the smoothing window is enlarged, a plurality of edges running side by side cannot be resolved.

(3) Since the connection relationship in the ridge line direction for each pixel is not observed, in situations where noise removal is insufficient, the edge directions pointed to by adjacent pixels may be different and may not be effectively emphasized. .

(4) There is a case where the emphasis processing is performed only once and the result is insufficient.

Due to the above problems, edges have not been effectively emphasized in low S / N images.

[0009]

According to the present invention, there is provided an edge enhancing apparatus comprising: an input image switching unit; a direction image generating unit;
It comprises a detecting means, an image synthesizing means, and an image storing / selecting means.
In the edge enhancement device to be formed, the input image switching means passes an original image, which is an image input to an external image input device, to the direction image generation means, and after sending the original image, the image synthesis means Is passed to the direction image generating means, and the direction image generating means obtains a direction image indicating a local edge direction from the image given from the input image switching means, Based on the directional image, the original image coming from the image input means smoothed in the direction parallel to the edge, to find an edge image subjected to differentiation in the direction perpendicular to the edge,
The image synthesizing unit adds the obtained edge image to the original image according to a preset weight, sends the result to the image storage / selection unit and the image input switching unit, and outputs the image storage / selection unit. Receives the enhanced image obtained by the image synthesizing means as an input,
By evaluating the toned image as a new emphasized image according to an evaluation criterion set in advance , the score of the new emphasized image is evaluated.
And the score of the new enhanced image has already been evaluated.
If higher than the score of the old-enhanced image that holds, the
Retain the new enhanced image instead of the old enhanced image ,
If the score of the new enhanced image is low or the enhancement is repeated
If the number of times has reached the number given from outside in advance
Output the old emphasized image to an external image display device.
And the repetition end control signal is
It is characterized in that it is outputted to the changing means .

[0010] Also, the edge enhancement device of the present invention, the direction
The image generating means includes a window parameter setting unit and a base point.
Window setting part and candidate window setting part and similarity calculation
Section and similarity storage / comparison section, history image storage section and direction image recording
Storage unit, correlated image storage unit, base window setting control unit
And the window parameter setting unit comprises:
The size of the window that needs to be set when cutting out the minute image
And values given externally for the shape and weight in the window
And the base window setting unit determines the input
The base point window on the image received from the image switching means.
The window specified by the window setting control unit
Based on the size and shape determined by the dough parameter setting section
Set a point window and use the image in the
Cut out and save as a point window partial image, and
The coordinate values of the window partial image and its original image are
Sent to the similarity calculation unit, and the candidate window setting unit
Receiving the settings of the base window setting section, the input image
On the image received from the switching means,
A window with high similarity to the window near the dough
Are set, and for all candidate windows,
Window-partitioned partial image as candidate window part
Image and the candidate window partial image and its original image
The coordinate value in the similarity calculator is sent to the similarity calculator.
The arithmetic unit calculates the partial images in all candidate windows,
Externally given between the base window part image
The similarity is calculated according to the calculation method
Pass the window coordinate values to the similarity storage / comparison unit,
The similarity storage / comparison unit is configured by the similarity calculation unit
The degree of similarity to each obtained candidate window and each candidate window
The window coordinates are saved, and the candidate window with the highest similarity is saved.
Window as the next base window candidate, and
In the correlation image held in the storage unit, the next base point c
In the coordinates of the window candidate, the next base window candidate,
What is the base window that is currently the base window?
The previously calculated similarity with another base window is
If present, its past similarities and currently calculated classes
Compare the similarities, and if the past similarity is greater,
Window update signal to the base window setting control unit.
Conversely, if the current similarity is greater,
The degree and the coordinates of the next base window candidate are represented by the correlation image
Sent to the storage unit and further coordinates of the next base window candidate
And the coordinates of the base window are sent to the direction image storage unit.
At the same time, the coordinates of the next base window candidate are
Image base unit, and the base window setting control unit
The base window update signal from the similarity storage / comparison unit
Received, the history stored in the history image storage unit
In the image, the coordinates where the searched index is not
And set it as the coordinates of the new base window.
The coordinates are sent to the base window setting unit, and the similarity
Receiving the base window update signal from the storage / comparison unit
However, the searched index is recorded in all the pixels in the history image.
If it is input, the processing end signal is stored in the direction image storage.
To the base point window from the similarity storage / comparison unit.
Instead of the dow update signal, the next base window candidate
Receives the similarity between the coordinates of the
If it is removed, change the coordinates to the coordinates of the next base window.
And send the coordinates to the base point window setting unit,
The history image storage unit stores a history image having the same size as the input image.
Generates and stores a history image, and stores it in the similarity storage / comparison unit.
Received the coordinates of the next base point window candidate.
The finger indicating the searched end is displayed at the coordinates on the history image.
Mark, and the correlation image storage unit stores the same mark as the input image.
Generate and store the correlation image of the same size, and maintain the similarity
From the existence / comparison unit, the coordinates of the next base window candidate are
If the current similarity is received, the
The similarity is written in coordinates, and the direction image storage unit stores
Generate and hold the direction image of the same size as the input image,
From the similarity storage / comparison unit, the next base window
Received the coordinates of the candidate and the coordinates of the base window.
If the two coordinates are connected by a straight line on the direction image,
Is the angle formed by an externally given axis on the direction image
Calculating the angle on the direction image, the next base point window
C) Fill in the coordinates of the candidate, and set the base window setting controller
If the processing end signal is received from
Output picture

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the present invention will be described with reference to the drawings.

FIG. 1 shows an embodiment of the first invention. FIG.
In the embodiment of the first invention, a television camera or the like is used.
External image input means 100 and directional image generating means
Input image switching means 1 for switching an image to be sent to 102
01, a direction image generating means 10 for detecting the direction of a local edge of the input image and generating an image indicating the direction
2, an edge detecting unit 103 for detecting an edge from the original image based on the direction image, an image combining unit 104 for enhancing an edge portion of the original image based on the detected edge,
Save the enhanced image, an image storage compare unit 105 for comparison with the generated enhanced images in the past, finally chosen is
CR that displays the most preferably highlighted image externally
T and other image display means 106 .

The image input means 100 converts the image input by the camera into a digital image signal as an image input switching means 10
1. Send to the edge detecting means 103 and the image synthesizing means 104. The input image switching means 101 is the image input means 10
0 When the image signal sequence is sent from the, first switch initially sends the image signal sequence in the direction the image generation unit 102, thereafter sending the enhanced image sent from the image synthesis unit 104 in the direction the image generating means 102 And the image
When repeatedly receiving the end control signal from the storage / comparison unit 105 , the transmission of the image to the direction image generation unit 102 is stopped. The direction image generating means 102 uses, for example, a method described in detail in the description of the second invention to convert an image having an edge as shown in FIG. 4A into an original image as shown in FIG. Is defined as, for example, the angle between the edge and the direction perpendicular to the image, and the angle is written in the direction image at the same coordinate as the coordinate. The directional image generating means sends the directional image to the edge detecting means 103 after the generation of the directional image is completed. The edge detecting means 103 converts the original image based on the directional image into a direction parallel to the edge and a direction perpendicular to the edge, for example, x in pixel units, σ
= 3 and smoothed by a moving average weighted by Gaussian exp (-x2 / 2.sigma.2), for example, by performing a second derivative in the direction perpendicular to the edge, and performing the above operation on the edge image image combining means 104. To send to. The image synthesizing unit 104 generates an emphasized image by, for example, simply adding the edge image and the original image, and sends the emphasized image to the input unit 101 and the image storage / selection unit 105. The image storage / selection unit 105 stores the emphasized image, compares it with the past emphasized image, and stores the emphasized image if the emphasized image obtained by the current processing is more emphasized. The processed enhanced image is deleted. However, when the emphasis processing reaches the preset number of times, the emphasis image processed in the past is not erased, and the emphasis image is transmitted to the image display means 106 and the input image switching means 101 repeatedly terminates. Send a signal. On the other hand, if the emphasized image processed in the past is more emphasized, the emphasized image is displayed on the image display means.
At the same time as sending to the stage 106 , the emphasized image obtained by the current processing is deleted, and an end signal is repeatedly sent to the input image switching means 101 . The image display means 106 receives the enhancement.
Display an image.

FIG. 2 shows an embodiment of the second invention.
Directional image generating means 102 in the first invention shown in FIG.
Is equivalent to In FIG. 2, a second embodiment of the present invention sets a window parameter setting means 201 for setting the size and shape of a window from which a partial image is cut out on an input image 200, and sets a base window on the input image according to the window parameters. A base window setting means 202 for extracting an image in the window as a base window partial image and setting a plurality of candidate windows in the vicinity of the base window in accordance with the window parameters, and extracting the image in the window to a candidate window part Candidate window setting means 203 for holding as a group of images, and similarity calculating means 20 for calculating the similarity between each candidate window partial image group and the base point window
4, a similarity storage / comparison unit 205 for selecting the candidate window with the highest similarity, a base window setting control unit 206 for setting the candidate window with the highest similarity as a new base window, and the calculation of the similarity is completed. A history image storage unit 207 for generating and storing a history image for checking whether or not a correlation image is generated, and a correlation image for generating and storing a correlation image in which the similarity of the candidate window having the highest similarity is entered in the coordinates of the candidate window. Image storage means 209;
Direction image storage means 208 for generating and storing a direction image indicating the direction of a local edge on the input image, and outputting the direction image as an output image 210 when the local edge direction is obtained for all input images.

In FIG. 2, a window parameter setting means 201 is placed, for example, on a storage device of a computer, and has a size and a shape of a window, for example, 4 pixels by 16 pixels.
The window parameter is determined by reading a file describing that the pixel is a rectangle. The base window setting unit 202 sets the coordinates specified by the base window setting control unit 206 as shown in FIG. 5B with respect to the input image having the edge as shown in FIG.
A base window having the size and shape determined by the window parameter setting means 201 is set, and an image in the base window is cut out and stored as a base window partial image. FIG. 5C shows the candidate window setting unit 203.
As shown in (5), five candidate windows are set every eight pixels in the vicinity of the base window, for example, 4 pixels below the center coordinates of the base window and 24 pixels on the right side from the left 24 pixels of the center coordinates of the base window. Is cut out and saved as a candidate window partial image group. The similarity calculating means 204 raster-scans, for example, the root window partial image from left to right and from top to bottom, arranges the pixel values of the partial image one-dimensionally, sets it as a vector A, and similarly raster scans the candidate window. Then, when the vector is set as a vector B, the similarity is calculated using the respective candidate window partial images and the base window partial image, for example, assuming that AB · | A || B | The similarity storing / comparing means 205 stores all the similarities obtained by the similarity calculating means 204, selects a similarity having the highest value from the similarities, and sets a candidate window given the similarity to a next base window. As a candidate, in the correlation image held in the correlation image storage unit 209, the similarity calculated in the past between the next base window candidate and another base window is set at the center coordinate of the next base window candidate. If there is, the value is compared between the past similarity and the currently calculated similarity, and if the past similarity is larger, a base window update signal is sent to the base window setting control means 206, while If the similarity is larger, the current similarity and the coordinates of the next base window candidate are sent to the correlation image storage means 209, and the coordinates of the next base window candidate are further added. The base window coordinate orientation image storage unit 208
And sends the coordinates of the next base window candidate to the history image storage unit 207 and the base window setting control unit 206. Upon receiving the base window update signal, the base window setting control unit 206
In the history image held in the base image, the coordinates of the base window are determined so that the center coordinates come to the coordinates where the searched index is not entered, and the coordinates are sent to the base window setting means 202. If the searched index is written in all the pixels, a process end signal is sent to the direction image storage means 208. If the coordinates of the next base window candidate are received, the coordinates are determined as the coordinates of the next base window. The coordinates are sent to the base window setting means 202. The history image storage unit 207 generates and holds a history image, and upon receiving the coordinates of the next base window candidate, writes a searched end index at the coordinates on the history image. Correlation image storage means 209 generates a correlation image,
If the coordinates are held and the coordinates of the next base point window candidate and the current similarity are received, the similarity is entered at the coordinates on the correlation image. The direction image storage unit 208 generates and holds a direction image, and when receiving the coordinates of the base window and the coordinates of the next base window candidate, connects the two coordinates with a straight line on the direction image, and Calculate the angle between the direction image and the vertical direction, write the angle in the coordinates of the next base point window candidate on the direction image, and, when a processing end signal is received, convert the direction image to the output image 2.
Output as 10.

FIG. 3 shows an embodiment of the image storage / selection means 105 in FIG. 1 showing an embodiment of the first invention. In FIG. 3, an image storage / selection unit binarizes an input image 300 and detects an edge candidate fragment group.
And labeling means 3 for labeling each edge candidate fragment
02, an edge candidate connecting means 303 for connecting edge candidate fragments based on the position information to form an edge candidate, an edge determining means 304 for erasing candidates that do not meet the criteria among the edge candidates, and an edge obtained just now. A score is calculated from the image based on a certain criterion, compared with the score of the edge image obtained in the past, the result is sent to the emphasized image storage means 306, and a score calculation / signal for transmitting a signal regarding repetitive control according to the result is obtained. Comparing means 305 and an emphasized image storing means 306 which holds the emphasized image and its score and outputs the emphasized image as an output image 307 in accordance with the score comparison performed by the score calculating / comparing means 305.

In FIG. 3, when the maximum value of the luminance of each pixel in the image data is max and the minimum value is min, the image binarizing means 301
(x-min) × 0.1 is set as a threshold value, and a value higher than the threshold value is binarized to 1, and a value lower than the threshold value is set to 0. In the binarized image, the labeling means 302 applies a label to each pixel, which is called an edge candidate portion, to which a value of 1 is given, and a different numerical value is given to each pixel. The edge candidate connection means 303 regards two edge candidate portions that are within ± 1 pixel in the horizontal axis direction on the image and 1 pixel in the vertical axis direction on the image as one edge, and considers two edges as two edges. The labels of the candidate portions are replaced, and the edge candidate portions are connected to form edge candidates. Edge determination means 30
Numeral 4 measures the length of each edge candidate. For example, a threshold value is set to 80 pixels. Edge candidates shorter than the threshold value are regarded as noise and erased, and long edge candidates are regarded as edges and determined as edges. The score calculation / comparison means 305 sums up, for example, the lengths of all detected edges in the edge image obtained by the edge determination means, and obtains the result as a score. Compared to the score, if the obtained score is higher, an enhanced image update signal is sent to the enhanced image storage means 306, and if it is lower, a repetition end signal is input to the enhanced image storage means 306.
This is sent to the force image switching means 101. In FIG. 3,
The transmission to the input image switching means 101 is illustrated.
Not. The enhanced image storage unit 306 deletes the previously stored input image when receiving the enhanced image update signal, stores the newly input image, and stores the currently stored input image when repeatedly receiving the end signal. The output input image is output as an output image 307.

[0018]

【The invention's effect】

(1) When noise is not sufficiently removed by only one smoothing process, the obtained local edge direction largely differs from the true edge direction, and the edge may not be effectively emphasized. However, by performing the repetitive processing, the degree of smoothing increases, and the mismatch can be reduced. However, when the number of repetitions is increased, the smoothing becomes excessive, so that not only a fine structure of the edge shape is lost, but also an edge having a low S / N and a short length is buried in noise and cannot be detected. In the first invention, the emphasis process is repeated, for example, the image storage / selection unit 105
By making it possible to select an emphasized image based on the criterion described in the embodiment, there is an effect that an image which is most preferably emphasized under the criterion is output.

(2) When calculating the gradient of the local brightness curved surface of the image to determine the direction of the local edge, in the case of excessive noise, an image having an edge as shown in FIG. For example, as shown in FIG. 6B, the direction of the edge may be completely different between adjacent pixels, and the edge is not effectively emphasized. In the present invention, since the condition that the edge is continuous is included, the direction of the edge is changed as shown in FIG.
(C) is obtained, and there is an effect that the direction of the edge does not largely differ between adjacent pixels.

(3) When finding the direction of the edge, the similarity is calculated and the cross-sectional shape of the edge is checked to connect the two points. For example, a diagram in which two edges A and B intersect In the case of FIG. 7 (a), the edge to be connected before and after the intersection is mistaken as shown in FIG. 7 (b), or it is erroneously determined that A is interrupted at the intersection as shown in FIG. 7 (c). This has the effect of making it harder.

(4) In the case of a sine wave-like edge that continues up and down as shown in FIG. 8A, in the method of finding the direction of the edge by the gradient of the luminance surface, a smoothing window is set in FIG. If a large value is taken like A in b), the edge structure is lost, while B is used to maintain the edge structure.
However, the accuracy of the similarity degree depends only on the area of the window, and the shape thereof does not matter.
By making the window horizontally long like C and D in (c), the accuracy can be improved without losing the edge structure.

(5) As shown in FIG. 9 (a), when a plurality of edges run close to each other in the vertical direction, the edge direction is determined by the gradient of the luminance curved surface. If the window of FIG. 9B is made large as in A in FIG. 9B, the three edges are not decomposed. If the window is vertically elongated as in B in order to maintain the spatial resolution, the accuracy of the horizontal component of the gradient decreases. In the present invention, C, D in FIG.
By making the window vertically long as in the above, the accuracy can be improved without losing the edge structure.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the first invention.

FIG. 2 is a block diagram showing an embodiment of the second invention.

FIG. 3 is an embodiment of an image storage / selection means in the above invention.

FIG. 4 is a diagram showing a specific example of an edge image and a direction image in the above invention.

FIG. 5 is a diagram showing an example of an input image and a window setting in the above invention.

FIG. 6 is a diagram showing the effect of the present invention as compared with a conventional method.

FIG. 7 is a diagram showing the effect of the present invention as compared with the conventional method.

FIG. 8 is a diagram showing the effect of the present invention as compared with the conventional method.

FIG. 9 is a diagram showing the effect of the present invention as compared with a conventional method.

[Explanation of symbols]

REFERENCE SIGNS LIST 100 image input means 101 input image switching means 102 directional image generation means 103 edge detection means 104 image synthesis means 105 image storage / selection means 106 image display means 200 input image 201 window parameter setting means 202 base window setting means 203 candidate window setting means 204 Similarity calculation means 205 Similarity storage / comparison means 206 Base window setting control means 207 History image storage means 208 Direction image storage means 209 Correlation image storage means 210 Output image 300 Input image 301 Image binarization means 302 Labeling means 303 Edge Candidate connection means 304 edge determination means 305 score calculation / comparison means 306 enhanced image storage means 307 output image

Claims (2)

(57) [Claims] An input image switching means and a direction image generating means are provided.
Step, edge detecting means, image synthesizing means, image storage / selection means
And the input image switching means passes an original image, which is an image input to an external image input device, to the direction image generating means, and after transmitting the original image, Passing the enhanced image coming from the image synthesizing unit to the directional image generating unit, wherein the directional image generating unit obtains a directional image indicating a local edge direction from the image given from the input image switching unit, Edge detection means, based on the direction image, for the original image coming from the image input means smoothing in the direction parallel to the edge, to obtain an edge image subjected to differentiation in the direction perpendicular to the edge, The image synthesizing unit adds the obtained edge image to the original image according to a preset weight, and outputs the result to the image storage / selection unit and the image input switching unit. Ri, the image storage and selection means inputs the enhanced image obtained by the image synthesizing means, the enhanced image in which the resulting new
Evaluate according to preset evaluation criteria as a new emphasized image
By calculating the score of the new enhanced image,
The score of the new enhanced image has already been evaluated and held
Old is higher than the score of the enhanced image, the old emphasis image that
The new enhanced image held in place of the image, the new
If the score of the emphasized image is low or the number of
If the number of times given from outside is reached beforehand,
Outputting old emphasized images to an external image display device
The repetition end control signal is supplied to the input image switching means.
An edge enhancement device for outputting to an edge. 2. The apparatus according to claim 1, wherein the direction image generating means includes a window parameter.
Meter setting section, base window setting section, and candidate window
C Setting section, similarity calculation section, similarity storage / comparison section, and history screen
Image storage unit, direction image storage unit, correlation image storage unit,
And a window setting control section. The window parameter setting section cuts a partial image.
Window size and shape that need to be set when
Determine weights in windows according to externally given values
And the base window setting unit controls the input image switching operation.
Control of the base window setting on the image received from the column
The window parameters to the coordinates specified by the
-The base window of the size and shape determined by the setting unit
Set the image in the base window as the base window
Cut out and save as a minute image, and
The image and its coordinate values in the original image are sent to the similarity calculation unit.
The candidate window setting unit is configured to set the base window setting
Unit setting, and receives the setting from the input image switching unit.
On the image to be trimmed, the window
Set multiple window candidates with high similarity to windows
Then, for all candidate windows,
The obtained partial image is used as a candidate window partial image,
Coordinate values in the complementary window partial image and its original image
Sent to the similarity calculation unit, and the similarity calculation unit
The image is external to the base window partial image.
Calculate the similarity according to the given calculation method and calculate
Save the result and the coordinate values of each candidate window to the similarity
Handover to the comparison unit, and the similarity storage / comparison unit
The degree of similarity to each obtained candidate window and each candidate window
The window coordinates are saved, and the candidate window with the highest similarity is saved.
Window as the next base window candidate, and
In the correlation image held in the storage unit, the next base point c
In the coordinates of the window candidate, the next base window candidate,
What is the base window that is currently the base window?
The previously calculated similarity with another base window is
If present, its past similarities and currently calculated classes
Compare the similarities, and if the past similarity is greater,
Window update signal to the base window setting control unit.
Conversely, if the current similarity is greater,
The degree and the coordinates of the next base window candidate are represented by the correlation image
Sent to the storage unit and further coordinates of the next base window candidate
And the coordinates of the base window are sent to the direction image storage unit.
At the same time, the coordinates of the next base window candidate are
The base window setting control unit sends the similarity degree storage / ratio
When the base window update signal is received from the comparison unit,
In the history image held in the history image storage unit,
Select the coordinates where the searched index is not entered, and
The coordinates of the new base point window are set as the coordinates of the base point window.
Window to the window setting section and forward from the similarity storage / comparison section.
Even if a base window update signal is received,
If the searched index is entered for all pixels in the image
If a processing end signal is sent to the direction image storage unit,
From the similarity storage / comparison unit, the
Instead, the coordinates of the next base window candidate and its current
If we receive the similarity with the base window of
The coordinates are determined as the coordinates of the next base window, and the coordinates are determined.
Is sent to the base window setting unit, and the history image storage unit stores a previous image having the same size as the input image.
Generates and stores a history image, and stores it in the similarity storage / comparison unit.
Received the coordinates of the next base point window candidate.
The finger indicating the searched end is displayed at the coordinates on the history image.
Mark, and the correlation image storage unit stores a mark of the same size as the input image.
The correlation image is generated and stored, and stored in the similarity storage / comparison unit.
The coordinates of the next base window candidate and the current similarity
, The similarity to the coordinates on the correlation image
And the direction image storage unit stores the same size as the input image.
Generates and stores the notational image, and stores it in the similarity storage / comparison unit.
The coordinates of the next base window candidate and the base window
If the coordinates of the window are received,
The coordinates are connected by a straight line, and the straight line is
Angle with the given axis, and calculate the angle
On the image, fill in the coordinates of the next base window candidate and
Receiving the processing end signal from the base window setting control unit.
If the image is removed, the direction image is output as the output image 210.
2. The edge enhancement device according to claim 1, wherein the edge enhancement device applies force .