JPH01168166A - Noise rejection method in multi-color picture - Google Patents

Abstract

PURPOSE:To eliminate noise in multi-color picture, especially uneven color noise by incorporating adjacent areas satisfying noise condition with respect to an area of a region and a color difference according to a predetermined noise rejection rule. CONSTITUTION:A CPU 102 applies noise rejection processing as to a multi-color picture stored in a picture memory 101 and a noise rejection processing program 105 for the processing is stored in a ROM 103. Rules I, II for noise rejection are incorporated in the noise rejection program 105 and the rule I applies mainly the rejection of uneven color noise and then the rule II rejects isolated point noise. Then the rule I incorporates the area based on the color difference and the area between adjacent regions. Thus, uneven color noise is sufficiently eliminated.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field] The present invention relates to a method for removing noise, mainly color unevenness noise, in multicolor images in devices such as copying machines and facsimiles that handle multicolor images.

[Prior art]

Conventionally, in devices that process this type of multicolor image,
Noise is removed by filtering a multicolor image. Although this can speed up processing, it cannot sufficiently remove noise such as color unevenness, and cannot flexibly deal with a variety of noises.

〔the purpose〕

SUMMARY OF THE INVENTION An object of the present invention is to provide a noise removal method that can effectively remove mainly color unevenness noise in a multicolor image and can flexibly deal with a variety of noises.

〔composition〕

The present invention removes noise in a multicolor image, particularly color unevenness noise, by integrating adjacent regions that satisfy noise conditions regarding area and color difference according to rules for noise removal. .

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows an example of a hardware configuration for implementing the present invention. 101 is an image memory, 102 is a central processing unit (CPU), 103 is a ROM, and 104 is a RAM
It is.

Multicolor image data is read by an external scanner and stored in image memory 101. This image memory 10
The CPU 102 executes noise removal processing according to the method of the present invention on the multicolor image stored in the RO.
It is stored in M103. As described later, there are two rules for noise removal, ■ and H, which are incorporated into the noise removal processing program 105, but may be made independent as a rule base. RAM 104 is used to store data in progress and processing result data.

The overall flow of the noise removal process will be explained with reference to the flowchart shown in FIG. First, a label is attached to each region of uniform color in the multicolor image in the image memory 101 (step 201). Next, the multicolor image after such labeling is recursively divided into four parts. , generates data describing the multicolor image in a quadtree structure, and stores it in the image description data area 107 in the RAM 104 (step 2o2). The leaves of the tree structure have the identification number of each area as attached information. Note that the processing up to this point is not new, so a detailed explanation will be omitted. Furthermore, a regular structure such as a subtree may be used instead of the quadtree structure.

When an image is described in a tree structure in this way, the amount of data can be reduced, and topological spatial searches of image regions can be performed at high speed in subsequent processing.

Next, table I shown in FIG. 3 and table 2 shown in FIG. 4 are created and stored in the table 1 area 108 and table 2 area 109 in the RAM 104, respectively (step 203).

Table I stores, for each labeled area, its identification number, color identification number, area (number of pixels), and area circumscribing rectangle information. FIG. 5 is an explanatory diagram of this circumscribed rectangle information. 5o1 is a labeled region of uniform color, and the x, y coordinates (min
x, min y) and (wax x, ma
x y ) as circumscribed rectangle information, the search can be performed efficiently by limiting the search range using this circumscribed rectangle information.

The table (2) stores R (red), G (green), and B (blue) data corresponding to each color identification number.

Also, based on this R, G, B data, the color corresponding to the color identification number is created in the uniform color space L * B +1)
) *Convert to space, and store the values of L*, a umbrella, b umbrella, hue (H), and saturation (C) data in table H. Note that the hue (H) is defined by the angle on the ao186, b* plane. In addition, the saturation (C) is defined as f7 - irises T2. □In this way, by integrating colors using the L I a* l) skewer color system, it becomes possible to determine noise that matches the visual characteristics of humans.

From step 204 to step 208, the above-mentioned quadtree structure description data, table! This is a processing part that performs noise removal by applying rules while referring to .

In step 204, a region of interest is set, in step 205, color unevenness noise is mainly removed according to rule (2), and in the next step 206, isolated point noise is removed according to rule (2). Then, in step 207, it is determined whether all areas have been set as attention areas, and if there are any unset areas remaining, the process returns to step 204 to set the next attention area and execute the process. If all the regions are set as regions of interest, it is checked in step 208 whether there are any regions that have been identified as noise, and if there are any noise regions, the process returns to step 204 and the process is restarted.

In this way, the process ends until there is no region determined to be noise, that is, when all noise regions are removed.

FIG. 6 is a flowchart of the noise removal process according to rule (2) in step 205. Rule I is a rule for mainly removing color unevenness noise in a multicolor image by integrating regions based on the area and color difference (color difference in the L*aIb umbrella color system) between adjacent regions. In the example, there are three types: A, B, and C.

Type A integrates small areas included in a large area. Type B integrates adjacent areas with small color differences. Type C is a small area integration. 6(a) is a flowchart when type A is applied, FIG. 6(b) is a flowchart when type B is applied, and FIG. 6(c) is a flowchart when type C is applied. You may selectively apply one of these three types to remove uneven color noise, or
Two or more types may be applied simultaneously. The processing details for each type will be explained below.

In type A, it is checked whether the area of the region of interest a is smaller than a preset threshold Th IA (step 60
1). If it is smaller than Th IA (in the case of a small area), a step 602 of searching for a group of regions = (bi) adjacent to the region of interest. Within this adjacent region group B, the region bi whose color difference with the region of interest is less than or equal to the threshold Th1E and which has the largest area is designated as b (step 6o3). Then, check whether the area of this region is larger than the threshold Th1B (step 604)
, if it is large (in the case of a large area), the attention area a is determined to be noise, and it is integrated into the area A (step 605).
.

Note that the search for adjacent regions is performed on a quadtree structure.

The color difference is shown in the table ■ corresponding to the color identification number in Table I.
Obtained from the data. In addition, areas determined to be noise are specifically integrated by updating the quadtree structure and updating the table ■ (updating area and circumscribed rectangle information).
The color difference used for the determination is 7=*a*b color difference in the color system,
That is, ΔE (Llamb Shin, 1976). This is the same for types B and C as well as for the process at step 206.

In type B, it is determined whether the area of the region of interest a is larger than a threshold Th1B (step 611), and if it is larger than Th1B, a group of adjacent regions B=(bi) of the region of interest a is searched for (step 612).

The color difference with the region of interest a in the adjacent region group B is the threshold Th1B.
The region bi below E and the region of interest a are integrated. That is, here, the noise determination condition is actually only the color difference, and one area is arbitrary.

Note that Th1E>Th1BE.

In type C, the area of the region of interest a is the threshold Th1G
Check whether it is smaller than (step 621), and if it is smaller (
(when the area is small), search for the adjacent region group B = (bi) (
step 622). Within adjacent region group B, the color difference with the region of interest a is less than or equal to the threshold Th ICE, and the area is equal to or less than the threshold Th ICE
The region bi below the ICB (small area) and the region of interest a are integrated (step 623), where Th1CE>Th1
It is BE.

FIG. 7 is a flowchart of the noise removal process according to rule (2) in step 206. Rule (2) is a rule for removing isolated point noise by region integration based on the color and degree of isolation of the region of interest. Yellow (Y), magenta (M), cyan (AC), and black (B) caused by malfunctions of the printing mechanism, etc.
K) Noise caused by adhesion of ink to a small area and white noise are removed by this process.

First, it is determined whether the area of the region of interest a is smaller than the threshold Th2 (step 701). If the area is smaller than Th2, L* (lightness), H (hue),
Based on C (chroma) data.

It is determined whether it is ink adhesion or white spot isolated point noise (step 702).

That is, if the color of the attention area a is yellow, magenta, or cyan, C (saturation)) Thl and L umbrella ≦ Th2 and Hl - ΔH < H < H1 + ΔH (where Hl is the hue value of yellow, magenta, or cyan). )
In addition, if the color of attention area a is black or white.

It is checked whether the following conditions are satisfied: C (saturation)<Th3 and L>Th4 or C (saturation) Th3 and L<Th5. If any of the conditions is met, the region of interest is determined to be isolated point noise.

If it is determined in step 702 that it is noise, a group of adjacent regions B=(bi) of the region of interest a is searched for (step 703). Region b with the largest area within this adjacent region group B
Let i be b, the other area be bi', and the set be B'
= (bi') (step 704).

Determine whether the area of the region is larger than the threshold Th2b (
Step 705). If the area is larger than Th2b, it is determined whether the areas of all regions bi' of B' are smaller than the threshold Th2b' and are of the same color system as the region of interest a (
Step 706). If this determination condition is met, the process advances to step 707.

In step 707, a group of regions BIT within a rectangular range as shown in FIG. 8 around the circumscribed rectangle of the region of interest a is
= '(b i IT ) (step 707
). Next, it is determined that B=Bb (however, Bb=B"-B') (step 708). If this condition is satisfied,
The region of interest a is integrated into region A (step 709).

Through the above processing, the quadtree-structured image description data and table I are updated, and these are obtained in the image description data area 107 and table I area 108 as processing results.

Note that the rules and processing algorithms for removing one-color unevenness noise and isolated point noise are not limited to those in the above embodiments, and may be modified as necessary. Furthermore, the hardware configuration for processing can be changed in the same way.

〔effect〕

As is clear from the above explanation, according to the present invention, by using predetermined noise removal rules, it is possible to sufficiently remove color unevenness noise, etc., which cannot be removed by - type of filtering processing. In addition, it is possible to flexibly respond to various noises simply by changing the rules.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing an example of a hardware configuration for implementing the present invention, Figure 2 is a flowchart of the entire noise removal process, Figure 3 is an explanatory diagram of Table I, and Figure 4 is an explanation of Table ■. 5 is an explanatory diagram of circumscribed rectangle information, FIG. 6 is a flowchart of noise removal processing steps according to Rule I, and FIG.
The figure is a flowchart of the noise removal processing steps according to rule ■.
FIG. 8 is an explanatory diagram of a rectangular range to be searched around the region of interest. 101... Image memory, 102... Central processing unit, 103... ROM, 104... RAM, 1
05... Noise removal processing program, 107... Image description data area, 108... Table ■ area, 109... Table ■ area. Fig. 2 Fig. 3 Fig. 4 (maxχ2m χt) Fig. 6 (Q) (shi) (C)

Claims (4)

[Claims] (1) A method for removing noise in a multicolor image, which is characterized by integrating adjacent regions that satisfy noise conditions regarding area and color difference according to rules for noise removal. (2) In the multicolor image according to claim 1, wherein an area having an area below a certain threshold value is integrated into an adjacent area having a color difference below a certain threshold value and an area above a certain threshold value. How to remove noise. (3) A method for removing noise in a multicolor image according to claim 1, characterized in that an area having an area larger than a certain threshold value and an adjacent area having a color difference from the area smaller than a threshold value are integrated. (4) A multicolor image according to claim 1, characterized in that an area whose area is less than a certain area and an adjacent area whose color difference is less than a certain threshold value and whose area is less than a certain threshold value are integrated. How to remove noise inside.