JPH03502261A - digital data processing - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] digital data processing This invention relates to digital data processing.

Digital data information specifically present in a stored digital image matrix and methods and means for detecting important features in stored information. Ru.

When it is desired to automatically view a particular digital image, code is generally It is necessary to identify the characteristics of This applies to corners and The features of the image correspond to the outline of the object and the conspicuous surface markings present in the memory image. This is to do so. This task provides some low-level information that is used to However, the data extracted at this early stage is of reliable quality. That is the most important thing.

Representation of edge features present in an image is useful only for certain parts of the image It is. Edge joins and shadows on edges are used for other parts of the image. The so-called image corner, which corresponds to a large band and an isolated dot-like feature, What is called is important. For example, if the types of features expected in an image are In the expression of a scene, if it is not limited in any sense, it is simply an edge. Representing an image with only corners or corners alone does not represent the entirety of the information present. I can't grasp the story.

A data processing method has been proposed that extracts both.

The Canny edge filter is one such edge extractor. death However, this device cannot form edge bonds, and this problem In order to overcome this, it is necessary to perform more novel processing operations. Besides, the image In the structural domain, edges sometimes have insufficient and unreliable feature representations. , This is why some people prefer corner expressions.

Horavec, Hagel, Buputs Beaudet and Kitchen and Rosenfeld (Kttchen) & Rosenfeld) have developed some very effective corner detectors. Proposed. However, these corner detectors are not uniform in their response. There is an obvious temporal discrepancy! (11) Unreliable), to create an even stronger image edge We respond too quickly to small imperfections in And it's not ideal.

This invention overcomes the drawbacks for both edge and corner feature detectors. This was done for the purpose of

According to this invention, in a pixel representing an image, its adjacent pixels Digital image matrix for detecting the location of sharp intensity changes present in the bar A method for processing digital data in a computer, the method comprising: The intensity is compared with the intensity of the surrounding secondary pixels, and the primary pixels are edge-shaped. Calculate whether it is classified as having a fixed value or a corner-like constant value, and Furthermore, the characteristics of both edges and corners that exist in the image matrix are A digital data processing process consists of steps that compare this information to obtain the location of the provide a method of management.

Before the above comparison step, the original image matrix is freed from unnecessary details. It is advantageous if the image is processed to form a simple intermediate image. child The intermediate image may be a quaternary image.

In one embodiment, the method further comprises: matrix and any changes in the position of the selected features of the image. The digitally processed area of the second image can be also detected in the first image. Contains steps to compare with that of the page.

As an example, an embodiment of the present invention will be described as a block diagram of a main part of an image motion monitoring detector. The following description will be made with reference to the drawings.

The detection device of the present invention will first be explained with reference to the main parts shown in the drawings, The operation in the feature extraction stage will be explained later.

This monitoring detector supplies information to the digital data processor 2 and outputs the The video camera 1 is operated to provide an output signal to the video camera 3. Processor 2 receives an input signal from the video camera, and this signal is sent to the image digitization stage 4. Ru. After this stage, the input information is sent to the frame store 6 and the pixels in this store are The pixels are then sequentially processed in a corner/edge feature extraction circuit 7.

The data obtained from the feature extraction circuit 7 is then applied to a feature matching block 8. child Parallel to this, the data is further processed for a certain period of time before sending this data to block 8. It is fed to a delay block 9 for holding.

In the feature matching block 8, the extracted features of the data stream are determined in advance. After a certain period of time, the feature set is compared with the compiled set of features. The information obtained is Therefore, it is not possible to show any changes in the selected extracted features that occur within a specific time. I can do it. From this information, the position and image of the moving object observed by video camera 1 can be determined. It is possible to calculate the velocity of

of sharp intensity changes in adjacent pixels representing the image matrix. To detect the presence, we first take the intensity of light at the next pixel. It is necessary to compare the intensity of each surrounding secondary pixel. at least 24, or There are several hundred secondary pixels.

Therefore, a local autophase Ill function for each -order pixel is required, and this The function is a local representation of the image intensity represented by all 25 possible pixels. Whether the batch is approximately constant, has edges, or has corners. This indicates whether the Corners are automatic with sharp peaks It is represented by a correlation function, and edges are represented by a mountain-shaped autocorrelation function. accurate auto It is not necessary to calculate the correlation function, but to determine the quadratic expansion with respect to its original is necessary. The mathematical expression of such a requirement is shown below.

First, the image intensity (preferably smoothed in advance) is set to Ix9. That's it Indicated by A. Note that x and y are Cartesian image coordinates. First of all, two The first-order gradient of is calculated as follows.

Next, a smoothed fourth-order gradient is calculated.

Note that * indicates rotation, and W is, for example, the following smoothing filter.

W=(242) Finally, the corner/edge response R is calculated.

R=(A,B-C2)-k(A+8)2 The negative hook on the coefficient is 0.1 .

The presence of corner regions is indicated by a positive and large value of R. individual vixe at a corner if its response R is greater than the response of each of its 8 neighboring pixels. It is considered that there is. Similarly, edge regions are indicated by negative and large values of R. It will be done. If the response R is determined by which of the linear gradients of X or Y is larger, where X is smaller than the response of two adjacent vixels in either y direction negative), the pixel is considered an edge vixel.

Standard Edge Clean Apple algorithm (used in Canny Edge filters) (similar to the one shown), but with short lines and spurs removed applied to edge images to repair tears in. The result is each Pixels are categorized as adjacent to corners, edges and background. It becomes a four-dimensional (ie, four-state) image. These edges are thin (i.e. they are one pixel in width) and move between corner regions. Questions about bond formation The problem is to find a corner area where the edge ends in the fa area. overcome by the existence of Edge mismatch problem in 11 areas of image set The problem is an image region represented by a region that contains corners and contains few edges. overcome by the area.

One way to perform the necessary calculations is through high- or low-level software. (orr-the-she+r) It is the processing of data by a microcomputer. microcomputer The input to the data is a digitized image, and the output is the desired combination of features. position and image speed. However, due to the amount of computation required, microphone Using a computer is relatively slow, and this slowness is time-critical. Not desirable for applications.

For time-critical applications, intermediate image XSY.

Calculate A, BSC and R as corner or edge pixels, respectively. In order to select an appropriate local maximum or minimum value of the corner/edge response R, a special It is possible to configure hardware for other purposes. For this special purpose proposed The hardware is pipelined and further rotates with the chip and other provided VLs. It uses an SI circuit. This step requires one or more microcomputer What follows on the computer is a stage for edge cleanup and feature tracking. It is.

The scene observed by a video camera changes over time. If it is envisaged, the present invention will take a digital image of the tube 2 after a given period of time. is used to process the image matrix. Processing data for the second image above The data is then compared to the processed data of the first image, thereby making the image detect any change in the position of the selected feature in the selected feature.

The use of digital data processing methods can be used to investigate situations such as burglar alarms. proposed for the production of dynamic equipment. The appearance required for a static investigation camera If the edge and corner features move in unison, it indicates a moving object. On the other hand, unmatched movements are caused by image noise, plants being moved by the wind, etc. This is due to For applications in traffic control, this method requires the presence of moving vehicles. used to detect presence. Digital images use infrared light rather than visible light As a result, the system can operate effectively at night or in bad weather. Ru.

It has also been proposed to utilize this data processing method for automatic scanning of cells.

The above description of one embodiment of the invention is given by way of example only. Various modifications may be made without departing from the scope of this invention as defined in the claims. It is possible. For example, the invention provides data processing in a single image matrix. Although described as a method of It can be applied to processing data from Trix. Note that this also creates a three-dimensional effect. can be obtained.

international search report

Claims (5)

[Claims] 1. The sharpness that exists between adjacent pixels in pixels that represent an image. digital image matrix to detect the location of sharp intensity changes. A method for processing data, comprising: Ratios the intensity of each primary pixel to the intensity of the secondary pixels surrounding this primary pixel. Compare, The above primary pixel has a constant edge-like or corner-like value. Calculate whether the image is classified, and also calculate the errors that exist in the image matrix above. contrasting this information to obtain the location of both edge or corner features; Digital data processing method consisting of steps. 2. Before the comparison step above, an intermediate simplified image with undesirable details removed. The original image matrix is processed to form a The method described in Section 1. 3. 3. The method of claim 2, wherein the intermediate image is a quaternary image. 4. After a given time lightening, adopt a second digital image matrix and Furthermore, any changes in the selected feature positions of the image can be detected. A process for comparing the digitally processed data of the second image with that of the first image. 4. The method according to claim 1, further comprising step. 5. Digital data in a digital image matrix substantially as described above. data processing method.