JPH01138858A - Picture processing method - Google Patents

Abstract

PURPOSE:To obtain a video picture with less noise by applying movement correction to plural consecutive video pictures so as to apply calculation processing to the picture element data of the same position. CONSTITUTION:The movement is detected (2) between the video picture 1 and other video pictures 2-N in N-set of video pictures 1-N. Based on the detected movement, movement correction 3 is applied to each video picture to generate video pictures 2'-N' having no time difference from the video picture 1. Then the arithmetic mean value 4 of picture element data of the same position as the video pictures 1, 2'-N' is calculated and the result is used as the picture element data of the position, thereby obtaining a video picture 5 of the lower noise.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to an image processing method for producing high quality still images for hardcopy etc. from a plurality of consecutive video images.

BACKGROUND ART From the viewpoint of printing video images, there is a strong desire to improve the image quality of video images, and in particular, it is essential to reduce noise contained in video images. Even if a video image containing noise appears to have sufficient image quality when viewed as a moving image on a monitor television, the image quality is not satisfactory when hard-copied as a still image.

Conventionally used methods for reducing noise include, for example, a method in which noise in only flat areas of a video image is removed by averaging processing (Method 1, Bengami, Sayanagi; 1 Insha). (Refer to "Image Analysis", Kyoritsu Shuppan, pp. 318-319) and a method of removing noise by performing temporal interframe processing (Method 2, Fukinukisanon: "Digital signal processing of images" 1 Nikkan Kogyo) Newspaper company, p.
p., 115-118 (1986)).

As an example of a conventional noise removal method, method 2 will be explained using FIG. 3. When calculating the average of pixel data at the same position of N video images, the signal components have strong correlation between frames, so the values remain unchanged.
Since the noise component has no correlation between frames, the amplitude is 215.
reduced to In this way, noise is reduced for static parts of the image, but when averaging is performed for moving parts, motion blur occurs as shown in Figure 4. , shows the averaging process when a steep edge is moving. Since motion blur occurs in this way, motion detection 6 is performed, and noise removal as described above is not performed for moving parts.

Problems to be Solved by the Invention However, the conventional methods have the following problems. In method 1, there is a possibility that the gradation waveform with low contrast in the flat part becomes blurred, and noise in parts other than the flat part cannot be reduced because it causes a decrease in resolution. In method 2, noise in moving parts is not reduced, so the noise is not noticeable when viewed as a moving image on a television monitor, but becomes noticeable when hard-copied as a still image. In view of the above problems, the present invention provides an image processing method for reducing noise to an extent that does not cause deterioration of image quality even in hard-copied still images.

Means for Solving the Problems The technical means of the present invention for solving the above problems is to extract high-quality video from a plurality of continuous moving video images (video image 1 to video image NUN are integers of 2 or more). In an image processing method for obtaining an image, the amount of movement of each part of the image is determined between video image 1 and each of the other video images 2 to N, and motion correction is performed to video images 2 to N based on the amount of movement. Create video images 2゛~N° with no time difference from video image 1, calculate the pixel data at the same position in N video images 1, 2゛~N゛, calculate the representative value, and calculate the representative value at that position. By converting it into pixel data, a single high-quality video image is created.

action The effect of this technical means is as follows.

If N consecutive video images with movement are subjected to motion correction so that they become images at the same time, N video images at the same time are created. When pixel data at the same position is averaged in these N video images, the signal level of the pixel data remains the same, but the noise level is reduced to 1/F because there is no correlation between the N video images. This results in a video image with a low noise level. Furthermore, since motion correction is also performed, noise in moving parts is similarly reduced.

EXAMPLES Hereinafter, examples of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an outline of image processing in an embodiment of the present invention. First, in N video images (video image 1 to video image N), motion amount detection 2 is performed between video image 1 and each of the other video images 2 to N. There are various methods for detecting the amount of motion, such as finding the deviation that minimizes the difference between consecutive pixels and finding the deviation that maximizes the cross-correlation function. Digital Signal Processing” 3 Nikkan Kogyo Shimbun, p.
(See p. 221-227 (1986)).

Based on the detected amount of motion, each video image is subjected to motion correction 3 to create video images 2° to N° with no time difference from video image 1. And video image 1,
A low-noise video image 5 can be obtained by calculating the additive average 4 of pixel data at the same position from 2' to N' and using the result as the pixel data at that position.

The outline of motion correction 3 and addition averaging 4 in this example is explained in the second section.
This will be explained in detail using figures. The waveform on the left side of the figure is a grayscale waveform that includes noise at the edges of three consecutive video images with movement.The 0 dotted line is the waveform before motion correction. The waveform after motion correction is shown as a solid line, as the edge portions are distorted, resulting in motion blur. The signal component excluding noise is 3
Since both shapes are the same, even if you average them, they remain the same. On the other hand, all three noise components have random shapes, and by taking an average, the noise amplitude can be reduced to 1/J. In this way, a video image with less noise can be obtained.

Effects of the Invention As described above, the present invention obtains a single high-quality video image by performing motion compensation on a plurality of consecutive video images and performing calculation processing on pixel data at the same position. Since this calculation process is performed using pixel data corresponding to the same position of the subject through motion correction, motion blur does not occur even in a moving video image, and a video image with less noise can be obtained.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an outline of image processing in an embodiment of the present invention, Fig. 2 is an explanatory diagram showing how noise is reduced in an embodiment of the invention, and Fig. 3 is a conventional image processing. FIG. 4, a block diagram showing an example of the method, is an explanatory diagram for explaining motion blur that occurs at the edge portion of an image due to a conventional image processing method. 1...N video images, 2...Motion amount detection, 3...Motion correction, 4...Additional average, 5... -Video images. Name of agent Patent attorney Toshio Nakao Haka18 Figure 1 Figure 2 Figure 143 Figure 4

Claims (3)

[Claims] (1) In an image processing method for obtaining one video image from a plurality of consecutive video images (video image 1 to video image N: N is an integer of 2 or more), video image 1 and other video images 2 to N are By determining the amount of movement of each part of the image between each image and performing motion correction on video images 2 to N based on the amount of movement, video images with no time difference from video image 1 (video images 2' to N': N ' is an integer of 2 or more), calculate the pixel data at the same position of N video images 1, 2' to N' to obtain a representative value, and use the pixel data at that position to create one image. An image processing method characterized by producing a video image. (2) A patent characterized in that in calculation processing for determining a representative value from pixel data at the same position of N video images 1, 2' to N', the average value of pixel data at the same position is used as the representative value. An image processing method according to claim (1). (3) A patent claim characterized in that in calculation processing for determining a representative value from pixel data at the same position of N video images 1, 2' to N', an intermediate value of pixel data at the same position is used as the representative value. The image processing method according to scope (1).