JPH02284274A - Image processing method - Google Patents

Abstract

PURPOSE:To accurately segment picture data in the short time without requiring human intervention by converting color picture data to hue data, luminance data, and saturation data and using hue data or saturation data. CONSTITUTION:Color picture data is converted to hue data, luminance data, and saturation data of three attributes of color, and hue data or saturation data is used to segment picture data. For example, a square area abcd designated by a mouse is subjected to raster scanning, and luminance data RGB of a picture element G(i,j) is converted HLS (hue, luminance, and saturation), and the hue out of three attributes of color is defined as H(i,j). The picture element G(i,j) is displayed when the difference between H(i,j) and the hue Hk of a representative color of the segmented picture is smaller than a certain determined value H0, but it is not displayed when this difference is larger than H0, and the accuracy of picture segmentation is controlled by the value H0. Thus, the picture is accurately segmented even when the color of the picture to be segmented resembles that of the ground picture.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image processing method for cutting out arbitrary image data in an image processing system.

[Prior Art] Conventionally, when cutting out an arbitrary image from an image stored in an image memory, the following method has been considered.

Consider cutting out a T-shirt section 21 from the base 22 from the image shown in FIG.

For example, assume that the color of the T-shirt 21 is red and the color of the base 22 is green. Image data is generally R (Red), G
(Green) and B (Blue), it is necessary to distinguish between the two images based on the size and ratio of R, G, and B of the image data of the T-shirt part 21 and the image data of the base part 22. .

First, select one pixel or several pixels within the portion to be cut out, and obtain representative data Ro, o, and Bo. Then, all pixels are compared with R8, Go, and Bo, and if the data R, G, and B of the compared pixels are similar, it is determined that the pixel should be cut out.

[Problems to be Solved by the Invention] However, the above conventional example has the following drawbacks.

(1) When the image to be cut out and the background color are similar in color, for example, red and orange, purple and blue, etc. cannot be cut out well.

(2) Even if the image to be cut out and the background color are quite different, the colors in the image to be cut out have the same hue and brightness (
If only the brightness (brightness) or saturation is changing, a range that is impossible to extract may be created.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide an image processing method that eliminates the drawbacks of the conventional method and allows image data to be cut out accurately in a short period of time without any manual effort.

[Means for Solving the Problems] In order to solve the above problems, the image processing method of the present invention converts color image data into three color attributes: hue data, brightness data,
The color image data is converted into saturation data, and arbitrary image data is cut out from the color image data using hue data or saturation data.

[Example 1] In the first example of the present invention, in order to make it easier to identify the image, the R, G, and B luminance data of the image are converted into H (hue) and L
(brightness) and S (saturation) data to easily extract the features of the image to be cut out and the underlying image, thereby easily cutting out an arbitrary image data portion from the image data.

The image to be extracted is an image with a gradation similar to that used in boat fishing, and the hue and saturation are considered to be almost constant throughout the image, with only the brightness (luminance) changing. it is conceivable that. Therefore, we focused on this, thinking that it could be distinguished from other images by differences in hue and saturation.

FIG. 3 shows an image memory 31 for storing images, and the structure within one pixel includes a mask bit 32 in addition to an area 33 for storing RGB data. This mask bit 32 is a bit used to select whether or not to display the image on the CRT.
” is set, the image data will not be displayed on the CRT even if it is in memory. Conversely, if this bit is set to 0, the image data corresponding to that pixel will not be displayed on the CRT. Is displayed.

In FIG. 4, consider the case where a circle 41 is cut out.

The image here is not a monochrome solid image, but a natural image with gradation.

As shown in Figure 1, this method involves the step of specifying an image area (step 1) in which the image to be cropped is limited to a rectangular area using a pointing device such as a mouse, and the characteristic color of the image to be cropped is selected and determined. The process consists of an extraction pixel designation stage (step 2) for the extraction process, and an image clipping stage (step 3) for distinguishing the clipped image from the base image.

First, regarding image area designation, the flow of processing will be explained below with reference to the flowchart of FIG.

In steps 51 to 54, the image to be cut out is surrounded by a rectangle using the mouse. By specifying the start point and end point, the target image can be roughly separated as shown in FIG. Below, the separation of the base and cutout image is shown in Figure 4a.
Rusk scanning is performed within the area surrounded by b, c, and d. Here, the height and width of the rectangle is W, the coordinates of a are (0, O), and the coordinates of c are (w, h).

Next, in step 55, the four points a, bc, d of the rectangle are
RAVE, GA
Further, in step 56, VE and BAVE are determined.
RGB-HLS conversion is performed to convert RGB luminance data into hue (H), luminance (shi), and saturation (S) data.

The RGB-HLS conversion in step 56 will be explained below.

Here, H, L, and S are Hue (hue) and Lig, respectively.
htness (brightness, brightness)Saturation
These are the three attributes of color given by (saturation). The HLS color space has the same three attributes of color perception as the Munsell color system.
It is shown in a cylindrical coordinate system as shown in the figure. H92 represents the hue as a rotation angle in the color circle 91, and L9
4 is BLACK96 with the brightness as the position on the achromatic color line 95
S93 represents saturation as a distance from the achromatic color line 95 to the outside.

FIG. 10 shows the conversion formula. This conversion formula is a general conversion formula that has been used for specifying colors in computer graphics. Further, in this embodiment, an HLS conversion formula is used, or any conversion formula may be used as long as it converts the RGB luminance level into three color attributes.

In addition, in Fig. 10, max and min are given R,
The three maximum and minimum values of G and B are shown.

H, S, and L are calculated from these values.

There is a relationship.

In FIG. 5, at step 56 RAVEGAVE B
By converting AVE into RGB-4HLS, the color is separated into three attributes of Hf, Lf, and Sf in step 57, and these are used as the representative color of the base.

Next, extraction pixel designation will be explained using the flowchart shown in FIG.

Specify the rough color of the part you want to cut out, and cut out the image based on that color. Using the mouse, click a pixel in a representative color part of the image you want to cut out, and display the luminance data R, G of that pixel.
, B. Furthermore, the values are changed to RGB-HLS) and the representative colors of the cut-out image are HR, I-,..., S.
Find R.

Next, image cutting will be explained using the flowchart shown in FIG.

As shown in FIG. 5, the area in which the image is to be cut out is a rectangle ab, c, and d surrounded by dotted lines specified by the mouse, and has a height and a width W. Set a to (0,0) for this rectangular area
and perform a mask scan. The luminance data RGB of pixel G (i, j) is converted from RGB to HLS, and among the three color attributes, hue (Hue) is focused, and it is set as H (i, j). The reason why we focused on hue is that the image you want to cut out is generally
This is because they often have a different hue and only the brightness (luminance) changes. For example, suppose you have a full-color image of an apple. Since the image is a photograph of a three-dimensional object, it naturally has shadows, and only the brightness (luminance) of that part changes, and the red "color" of the apple should hardly change. When cutting out an apple from a tree, it is best to use the change in hue as a parameter.

The difference between tl (i, j) and the hue H of the representative color of the cut out image, or if it is smaller than a certain determined value H, then the 1
ijj', '4 G (i, j) is considered to be a pixel constituting a cut-out image, and the mask pixel of G (i, j) is set to "θ" to display that pixel. Conversely, if the difference between It (i, j) and IIk is less than or equal to 2, G (i, l is considered not to be a pixel that constitutes the cropped image, and “1”
Set to not display that pixel.

Also, the value of ++o is 11. and Hk
It can be adaptively changed by the correlation with +1 if the hue of the background and cropped image are similar. Reduce the value of to control the accuracy of cropping one image. Image cutting is achieved as described above.

As explained above, according to the first embodiment of the present invention,
By converting the R, G, and B brightness data of an image into three color attributes and focusing on the change in hue, even when the image to be extracted and the remaining background image are similar in color, Images can be cropped accurately.

In addition, the underlying data is compared with data representative of the part to be cut out, and based on the results, the mask can be turned on or off.
By changing the parameter that determines F, the accuracy of extraction can be controlled.

Embodiment 2 FIG. 8 is a flowchart for explaining a second embodiment of the present invention. In the following description, only the points that are different from the first embodiment will be explained, and the explanation of the common parts will be omitted.

In the second example, consider a case where hue I+ of the representative color of the background and hue 1(, .

In this case as well, using the same concept as in the first embodiment, we focus on saturation among the three color attributes, and calculate S(i
, j) and the saturation Sk of the representative color of the cut-out image, or a certain determined value S. If it is smaller than that, the pixel G(i,
Considering that j) is the pixel that makes up the cut-out image,
The mask bit of G (i, j) is set to "0" and that pixel is displayed. Conversely, the difference between S (i, j) and S
. If this is the case, G (i, j) is considered not to be a pixel constituting the cut-out image, and the mask bit of G (i, j) is set to 1° so that the pixel is not displayed.

S again. The value of can control the accuracy of image cropping. Image cutting is achieved in the manner described above.

Note that although saturation was considered in this embodiment, both hue and saturation may be considered. In this case, more accurate cutting can be performed.

[Effects of the Invention] As explained above, according to the present invention, arbitrary image data can be cut out from a color image in a short time and accurately without any manual effort.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing the first embodiment of the present invention, FIG. 2 is a diagram explaining image cutting, FIG. 3 is a diagram showing the state of the image memory, and FIG. An explanatory diagram of cropping, Figure 5 is a flowchart of image area specification, and Figure 6 is:
Flowchart of extraction pixel designation, FIG. 7 is a flowchart of image cutting, FIG. 8 is a flowchart of image cutting and hanging of the second embodiment, FIG. 9 is a diagram showing HLS space, and FIG. 1O is a flowchart of image cutting. It is a figure explaining RGB-HLS conversion. 21... Image to be cut out 22... Base image 31... Image memory 32... Mask bits 33... Image data area 41... Image to be cut out 1... Color sample 2... Rotation angle 3 ...Saturation 4...Brightness 5...Achromatic color line Kuzu θ Diagram

Claims (1)

[Claims] An image processing method characterized by converting color image data into hue data, brightness data, and saturation data of three color attributes, and using the hue data or saturation data to cut out arbitrary image data from the color image data. .