JPH0395686A - Image processing method - Google Patents

Abstract

PURPOSE:To suppress the deterioration of a picture, and to improve picture quality by obtaining the center of sway, and re-arranging a swaying partial picture or a swaying picture element at this center of the sway. CONSTITUTION:Provided that a measuring sampling interval is sufficiently shorter than the cycle of the sway, the center 10 of the sway can be well approximated by using the center of the motion 9 of a very small part 8 with time, and by shifting the very small part 8 on the picture gr to its center 10 of the sway, the sway can be corrected. By executing this processing for the very small parts enough to fill up the picture gr, the sway of the whole picture gr can be reduced.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to an image processing method for reducing scattering on continuous images. Conventional technology Cumulative averaging has generally been used to reduce fluctuations in continuous images. This cumulative averaging involves processing at a certain time a using the following procedure. Each image of the continuous images gt (x, y) is sampled in the reverse chronological direction starting from time a, that is, at sampling intervals of 1 (1 is the time-series parameter of the image, and generally represents the time at which the image was measured), and cumulatively adds each pixel for a certain set number of sampling times T to obtain a cumulative image st(x, y) shown by equation (1).

SL (x+y) = Σ g t (x, y
) -----・=(1) By dividing the image St (x, y) formatted by the cumulative addition value by the cumulative number of images T, which is the addition parameter, as in equation (2),
An averaged image f t (x,y) with reduced scattering is obtained.

r c (x, y) − s t (x, y) / T
・・・・・・・・・・・・・・・・・・・・・(2) Problems to be solved by the invention The above-mentioned averaging process using cumulative addition can solve the problem of suddenness that occurs on an image or It was developed as a method to remove random noise, and has become widely used due to its ease of processing and applicability to real-time processing.
The basic process is that the value of the event is accumulated, and the value obtained by dividing the accumulated value by the cumulative parameter is determined according to the statistical probability of occurrence according to the density. Therefore, this process is very effective when the probability of occurrence is concentrated around a certain value, or when the values of the event are evenly distributed in positive and negative directions around a certain value, but For two-dimensional events (N images) that have a large variance value and a relatively wide probability density distribution, such as seeding, the blurring that occurs as a result of averaging processing is a processing issue that must be solved. It is taken up as In order to solve this problem, the present invention proposes a fluctuation noise reduction processing method that causes less image deterioration such as image blurring.

Means for Solving the Problems The present invention solves the above problems by using the following processing means. In other words, the fluctuation command that exists in continuous time-series images is extracted as a motion vector between images,
This motion vector is tracked to find the center position of the fluctuation, and the partial image or pixel that is the starting point of the fluctuation is relocated to this center position. Using motion vector detection processing, the fluctuation command is extracted as a motion vector from between the time-series images in which the action is spread, and this motion vector (fluctuation or division) is extracted on consecutive images in the direction of the time series. By tracking the motion vector, it is possible to trace the transition of fluctuation in the partial image (or pixel) associated with the motion vector, and the center position of the fluctuation can be calculated from the transition progress. A method is used in which the partial image (or pixel) that is the starting point of the drawing is relocated to the center position of the calculated ti-rag. In this way, fluctuations are captured as movements of partial images (or pixels), and in order to relocate the partial images (or pixels) to the center of the movement, image deterioration due to fluctuation averaging is reduced to the movement of partial images (or pixels). It is possible to suppress the size within a range.

EXAMPLE The image processing method of the present invention will be described below with reference to the drawings. FIG. 1 shows the structure of an example of a measurement system for carrying out the image processing method according to the present invention. In this measurement system, the measurement object 1 is measured by the TV camera 3 through the flame 2, and the images measured by the TV camera 3 are at times t and 2.
It is measured as a continuous image measured discretely at sampling interval t at t, 3t..., but as it is, it looks like the unprocessed image 6 due to the fluctuation of the light incident on the TV camera 3 due to the flame 2. It is measured as an image sequence of about tg. An image processing device 4 is provided to correct this fluctuation, and an image without fluctuation as shown in the processed image 7 is displayed on the monitor 5.

The method of fluctuation reduction processing of the image processing device 4 in this measurement system will be explained using FIG. 2. FIG. 2 shows a series of continuous images measured going back in time starting from a certain time T. The measurement sampling interval is t, and g? indicates an image measured at time T. Focus on minute portion 8 on image gt. @Assuming that the small portion 8 is smaller than the fluctuation width, the temporal movement 9 of the small portion 8 is determined by a motion vector such as block matching or gradient method between gr-c + gt-gt... By performing detection processing, it can be obtained as a motion vector. If the measurement system is static, that is, the measurement object and the TV camera are stationary, and the measurement sampling interval is sufficiently shorter than the period of fluctuation, then using the center of the temporal movement 9 of this minute portion 8, The center of fluctuation 10 can be well approximated, and the image g
, the fluctuation can be corrected by moving the upper minute portion 8 to the center 10 of the fluctuation. By performing this processing on a minute portion that completely fills g, the fluctuation of the entire image gt is reduced. Image gt-
You can do exactly the same thing for t+g r-tt'... Effects of the Invention As described above, the present invention departs from the processing that uses the statistical probability density of event values that ignores the nature of fluctuations in averaging by cumulative addition, and instead processes fluctuations in a time-series manner. Capturing the movement of partial images or pixels between images, extracting it as a motion vector, and analyzing the transition of fluctuation to find the center of the fluctuation, and relocate the partial image or pixel that is in the corner to the center of the fluctuation. By doing so, it is possible to suppress image deterioration due to fluctuation noise reduction to the extent that the upper limit is the size of the microscopic area when dividing the image for dynamic vector detection, and this effect greatly contributes to improving image quality. give birth to

[Brief explanation of drawings]

FIG. 1 is a conceptual diagram showing an example of a measurement system for carrying out the image processing method according to the present invention, and FIG. 2 is a diagram showing the principle of the image processing method according to the present invention. 1...Measurement target,
2...flame, 3...TV camera, 4...
...Image processing device, 5...Monitor, 6...
...Unprocessed image, 7...Processed image, 8...
...Minute part, 9...Temporal movement of minute part, 10...Center of fluctuation.

Claims (1)

[Claims] This method extracts the fluctuations that exist on continuous time-series images as a motion vector between images, traces the motion vector to find the center of the fluctuation, and relocates the starting point of the fluctuation to this center position. Featured image processing method.