JPH0463064A - Hybrid picture encoding and decoding device - Google Patents

Abstract

PURPOSE:To improve the easiness of read of character narrow lines or the like by providing a means which separates a picture signal including natural pictures, characters, graphics, etc., into character components and the background, a means which encodes character components separated by this separating means as a binary picture, and a means which encodes the background as a natural picture. CONSTITUTION:An inputted picture signal 1 including natural pictures, characters, and graphics is separated to character components (binary picture) 3 and a background (natural picture) 4 by a character component extracting part 2. Since being a binary picture, character components 3 are encoded to an MMR code by a binary picture encoding part 5 to obtain a binary picture coded data 6. Since being a natural picture, the background 4 is encoded to a DCT code by a natural picture encoding part 7 to obtain natural picture coded data 8. At the time of decoding, binary picture coded data 6 is decoded by a binary picture decoding part 9 to obtain decoded character components 10, and natural picture coded data 8 is decoded by a natural picture decoding part 11 to obtain a decoded background 12. The decoded character components 10 are multiplied in a (k) multiplying part 13 by (k) because being the binary picture, and they are added to the decoded by an adding part 14 to obtain a reproduced picture 15.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a hybrid image encoding/decoding device for efficiently compressing image signals including natural images, characters, figures, etc. to obtain reproduced images. .

(Prior art) Image encoding techniques include encoding methods (MH, MR, etc.) that assume that each pixel is binary black and white, and image encoding methods that assume that each pixel has a gradation (hereinafter the present invention). (This is called a natural painting)
Coding methods targeting [predictive coding (DPCM), orthogonal transform coding (DCT).

DFT, KLT, etc.). However, if the encoded image contains characters and figures in addition to natural images, if the same encoding method is used for the entire screen, noise will occur around the characters, making the characters difficult to read. For this reason, a hybrid method has been proposed in which one image is separated into a character component and a background, and block coding is applied to the character component and DCT is applied to the background.

(Problem to be Solved by the Invention) However, since character components can originally be binarized, if block coding for natural images is used, it is difficult to increase compression efficiency. There is. Also,
If possible, MH that is already widely used for facsimile,
It is desirable that it be a natural extension of a binary image encoding method such as MR.

(Object of the Invention) The present invention has been made to solve these problems, and it treats character components as binary images, and
The present invention aims to provide a hybrid encoding/decoding device that increases compression efficiency by applying a value image encoding method.

(Means for Solving the Problems) In order to solve the above problems and achieve the objects, the present invention provides a means for separating an image signal including a natural image, a single character figure, etc. into a character component and a background; A means for encoding the separated character components as a binary image, a means for encoding the background as a natural image, and a means for encoding the binary image encoded data and the natural image encoded data. It is characterized in that it includes means for decoding each, and means for multiplying the decoded character component by a constant and adding it to the decoded background.

(Function) In the hybrid image encoding/decoding device of the present invention, when an image signal containing a natural image and characters/figures is input, the image signal is separated into a character component and a background, and one character component is divided into two. The image is encoded as a value image, and the background is encoded as a natural image. Further, at the time of decoding, a binary image is decoded, a value multiplied by a constant, and a decoded natural image are added to restore the original image signal.

(Example) Hereinafter, embodiments of the present invention will be described in detail based on the drawings.

FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention, and FIG. 2 is an explanatory diagram of a method for extracting character components in a character component extraction section of the apparatus.

In FIG. 1, 1 is an image signal including natural images, characters, figures, etc., 2 is a character component extraction section, and 3 and 4 are character components (binary images) and background separated and extracted by the character component extraction section. (
(natural image), 5 and 7 are encoding units for binary image 3 and natural image 4, respectively; 6 and 8 are each encoded data of the binary image and natural image encoded by the encoding units, respectively; 9, 11 is a decoding unit for each encoded data of the binary image and natural image, 10.12 is a character component (binary image) and background (natural image) decoded by each decoding unit, and 13 is a character An addition section 14 adds the character component multiplied by a constant in the component section 13 to the decoded background, and 15 is a reproduced image signal.

To explain this operation, an input image signal 1 including natural images, characters, and figures is separated into a character component (binary image) 3 and a background (natural image) 4 in a character component extraction section 2. In the character component extraction unit 2, first, a certain pixel X11. around 9
Find the average value AI of X9 pixels. X, , is the density of the pixel, and the larger the value, the higher the density, character component 3 is assumed to be CI4, and background 4 is assumed to be B1.1.

Xl, I A1. , compared with a given threshold T.

X+, r A+, if r<T then CI, I = O
, B l , 1 = X1j, so that Figure 2 a), X 1
, (If A 1 , 4≧T then C, , = 1, B, ,
,=A,9. By setting +T (FIG. 2b), character component 3 and background 4 are separated.

Since the character component 3 is a binary image, the binary image encoding unit 5 converts M
MR encoding is performed to obtain binary image encoded data 6. Also, since the background 4 is a natural image, the natural image encoding unit 7 performs DCT
Encoding is performed to obtain encoded natural image data 8.

At the time of decoding, the binary image encoded data 6 is decoded by the binary image decoding unit 9 to become a decoded character component 10, and the natural image encoded data 8 is decoded by the natural image decoding unit 11. , becomes the decoded background 12.

Since the decoded character component 10 is a binary image, it is multiplied by a k multiplier 13 and added to the decoded background 12 in an adder 14 to obtain a reproduced image signal 15.

Here, k is determined by the character component extraction unit 2 as the average value of x, , A, , for all pixels whose CI4 is 1, and is separately sent to the decoding side.

As image signal 1 containing natural images, characters, and figures, a part of the color chart (896 pixels horizontally x 880 pixels vertically) of the Institute of Image Electronics Engineers Color Test Chart No. 11 was scanned with a drum scanner at 10 lines/mm in main scanning and 10 lines in sub-scanning. When encoding what is read in book/mm, cyan (C), magenta (M), yellow (Y), each color with 8 bits.

When T=30, the bit rate of the binary image encoded data 6 was 0.146 bit/pixel.

In addition, the background is DCT encoded with a quantization step size of 16, and the bit rate of natural image encoded data 8 is 0.456.
bit/pixel, total 0.60 bit
/ρ1xel. Note that DCT encoding is C, M.

For each Y color, DCT encoding, quantization,
It was performed in the order of variable length encoding.

In addition, the color text portion of the same color test chart (896 pixels horizontally x
When T=30, the bit rate of binary image encoded data 6 was 0.390 bit/pixel.

In addition, the background is DCT encoded with a quantization step size of 16, and the bit rate of natural image encoded data 8 is 0.41.
Since it is 0bit/pixel, 1 total is 0.80bi
It became t/pixel.

As shown in item 2, a good restored image could be obtained with a high compression rate of less than 1 bit/pixel. Subjective evaluations of image quality also showed that it improved the readability of characters, thin lines, etc.

(Effects of the Invention) As described above, according to the present invention, high compression efficiency can be achieved by treating character components as binary images. This also had the effect of improving the readability of thin lines of characters.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention. FIG. 2 is an explanatory diagram of a method for extracting character components in the character component extraction section of the apparatus. 1... Image signal including natural images, characters, and figures, 2...
Character component extraction unit, 3...Character component (binary image), 4
...Background (natural image), 5...Binary image encoding unit,
6... Binary image encoded data, 7... Natural image encoder, 8... Natural image encoded data, 9... Binary image encoder, 10... Decoded Character component (binary image), 11... Natural image decoding unit, 12... Decoded background (natural image), 13... k times part, 14
. . . addition unit, 15 . . . playback image signal. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] means for separating an image signal including natural images, characters, figures, etc. into character components and background; means for encoding the character components separated by the separation means as a binary image; For this purpose, there is a means for encoding it as a natural image,
It is characterized by comprising means for respectively decoding the binary image encoded data and the natural image encoded data, and means for multiplying the decoded character component by a constant and adding it to the decoded background. Hybrid image encoding/decoding device.