JPS6255347B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an image predictive restoration method for restoring an image sampled and read at a low pixel density to a high pixel density.

Generally, in digital copying machines, facsimile machines, etc., an electrical signal obtained by scanning a document image with a scanner is converted into a digital signal by an A/D converter, and the electrical signal is converted into a digital signal according to the density level. By sampling, images are read pixel by pixel.

At this time, in order to simplify the document image reading device and reduce the transmission line capacity required when digitally transmitting the image signal, coarse sampling is performed to reduce the number of bits per pixel and reduce the number of bits per pixel. Image processing is performed on a read image with a pixel density to convert it into an image with a high pixel density, thereby improving the quality of the reproduced image. Also,
Image processing means for this purpose uses arithmetic devices such as electronic computers to predict through simulation how the input image will change when it is reproduced, and to improve the resolution of the image due to the low pixel density. In order to improve this, predictive calculations are performed, but in this case, a major problem is how to faithfully reproduce the edge portions of the image.

In particular, based on each pixel information read at low pixel density by a scanner with a rectangular aperture, each pixel is divided into two small pixels.
When reproducing an image with twice the pixel density, conventionally, as shown in Figure 1a, when the pixel of interest X is divided into two small pixels x ₁ and x ₂ ,
Value quantization (black level is set to "2", middle gray level is set to "1", white level is set to "0"),
When _{the pixel} of interest When it is _“ 1”, _Q
When _B ≧ Q _G , the level of small pixel x ₁ is set to “2” and the level of small pixel x ₂ is set to “0”, and when Q _B < Q _G , the level of small pixel x ₁ is set to “0”, and the level of small pixel x Image reproduction is performed with the level of ₂ set to "2".

However, with such conventional predictive restoration methods,
For example, in Figure 1, pixels A, B, D, X, F, and G have a black level of "2", pixels C, E, and F have an intermediate level of "1", and the other parts have a white level of "0". In the case of an image, as shown in FIG. 2B, the image that has been subjected to the high pixel density processing will have noticeable irregularities on the left edge. Therefore, the prediction error of the reproduced image is relatively large, and although the apparent resolution of the image reproduced at high pixel density is improved, the image quality is poor due to conspicuous irregularities, especially at the edges of smooth figures. Along with becoming
The reproduced image is limited to a binary image consisting of a black level and a white level, and has the disadvantage that halftones cannot be expressed faithfully.

The present invention has been made with these points in mind. Based on pixel information read at low pixel density by a rectangular aperture, each pixel is divided into two small pixels to create an image with high pixel density. When reproducing images, especially when predicting and restoring edge parts that indicate intermediate tones, pattern parts that are prone to errors are excluded from the target so that prediction errors peculiar to image processing algorithms do not occur as in the past. The present invention provides an image predictive restoration method that can reproduce high-fidelity, high-quality images through processing.

Hereinafter, one embodiment of the present invention will be described in detail with reference to the flowchart shown in FIG.

In the image prediction restoration method according to the present invention, as shown in FIG. The density levels Q _X , Q _A to Q _H of each pixel in a specific pixel area are extracted (first step), and first the horizontal direction g1, right diagonal direction g2, vertical direction g3, and left diagonal direction g4 are extracted.
Each of the concentration gradients G1 to G4 in (see FIG. 3) is calculated according to the following equations (second step).

Next, in the present invention, the maximum value G.
MAX is extracted (third step), and its maximum value G・
By comparing MAX with G1 to G4 (fourth step) and sequentially performing the following judgments according to the results, each subpixel x ₁ and x ₂ when the pixel of interest X is divided into two. Determine concentration levels q _X1 and q _X2 .

(1) When G・MAX=G1, compare the preset threshold level T with the value of G1 (fifth step), and if G1<T, q _X1 = q _X2 = Q
Determine as _X (sixth step). At this time, if G1≧T, the process moves to the subsequent steps in FIG. 2, and predetermined determinations are sequentially performed. That is _, it is determined whether the density level Q _X of the pixel of interest _X is the black level "1", and when Q _X = 1, q _X1 = 1 and q It is determined whether the white level is "0", and when Q _X = 0, q _X1 =
Determine as q _X2 =0. In addition, if Q _X is not 0, a comparative judgment is made between Q _B and Q _G , paying particular attention to the density level states of pixels B and G that are vertically adjacent to the pixel of interest X, and Q _B < When Q _{G , a} comparison is made to _see if _Q 1, q _X2 = 1; otherwise, q _X1 = 0, q _X2 = 2Q _X. Also, if Q _B ≧Q _G, then Q _X ≦0.5
If so, determine q _X1 = 1 and _{q X2} = 2Q _X −1; otherwise, _{determine q}
Determine as.

(2) When G.MAX=G2 or G4, compare G3 and G1 (fifth step), and if G3>G1, determine q _X1 = q _X2 = Q _X (sixth step). At this time, if G3≦G1, the process moves to the following steps, and the same comparisons and judgments as described above are sequentially executed to determine the density levels q _X1 and q _X2 of the small pixels x ₁ and x ₂ , respectively. Have them do it as appropriate.

(3) When G・MAX=G3, determine as q _X1 = q _X2 = Q _X.

In the present invention, when the density levels q _X1 and q _X2 of each small pixel x ₁ and x ₂ are determined by dividing one pixel of interest This means that the same process is repeated sequentially by shifting. For example, when the image in Figure 4a is processed to increase pixel density using the image predictive restoration method of the present invention and then reproduced, the image in Figure 4b is
As shown in Figure 2, an image with good fidelity and little prediction error can be obtained. Note that the subsequent processing steps in FIG. 2 are those in which the conventional image predictive restoration method described above is executed as is.

Furthermore, FIG. 5 shows an example of a configuration for concretely implementing the image predictive restoration method according to the present invention. Shift register SR・A with pixel information for one pixel,
The input pixel information is sequentially sent to SR・B and SR・C, and the input pixel information is sent to shift register L・SR1 for one line.
The output of shift register L/SR1 is delayed by one line and then sent sequentially to shift registers SR・D, SR・X, and SR・E for one pixel, and then the output of shift register L・SR1 is sent to shift register L・SR2 for one line. After being delayed by one line, it is sequentially sent to shift registers SR・F, SR・G, and SR・H for one pixel, where each pixel information in a specific pixel area in a 3×3 configuration is extracted. . Next, the output of the shift register SR for each pixel is sent to the first arithmetic unit OP.
1, where the concentration gradients G1 to G4 in each direction g1 to g4 are calculated based on the above equation (1), and the results are sent to the first comparator CMP1, where the concentration gradients G1 to G4 of G1 to G4 are calculated. The maximum value G.MAX is extracted, and a second comparator CMP2 compares G.MAX with G1 to G4 to determine in which direction the concentration gradient having the maximum value lies. Then, the result is sent to the second arithmetic unit
It is sent to OP2, where the algorithms (1) to (3) are executed as appropriate to determine the density levels _qX1 and _qX2 of each of the subpixels _x1 and _x2 obtained by dividing the pixel of interest X into two. Each information of the determined small pixels x ₁ and x ₂ is 2
After the lines are collectively stored in the output buffer memory BM, each piece of information is sequentially sent to an external plotter in synchronization with the writing speed.

As described above, in the image predictive restoration method according to the present invention, each pixel adjacent to the pixel of interest in the longitudinal direction is When reproducing an image with high pixel density by determining the density level of each subpixel obtained by dividing the pixel of interest into two using an algorithm that takes into account the distribution state of the density level of the pixel of interest, The density gradients in the vertical, horizontal, and diagonal directions within a specific pixel area of n configuration are calculated, and the pattern that causes a prediction error specific to the algorithm is processed according to the direction with the maximum density gradient. This has the excellent advantage of being able to perform pre-processing to remove the image from the image, and that it can reproduce high-quality images with good fidelity in which unevenness is not so noticeable at the edge portions of images showing intermediate tones. have.

[Brief explanation of the drawing]

Figure 1a is a diagram showing a specific pixel area in a 3x3 configuration centered on the pixel of interest when an image was read at low pixel density, and Figure 1b is an image reconstructed from that image using a conventional predictive restoration method. FIG. 2 is a flowchart showing an embodiment of the image predictive restoration method according to the present invention, FIG. 3 is a diagram showing the direction for determining the density gradient in the same embodiment, and FIG. 4a is a diagram showing the original image. FIG. 5 is a diagram showing an example of the image, and FIG. It is a diagram. SR・A~SR・H, SR・X...Shift register for 1 pixel, L・SR1, L・SR2...Shift register for 1 line, OP1, OP2...Arithmetic unit,
CMP1, CMP2...Comparator, BM...Buffer memory.

Claims (1)

[Claims] 1. Based on image information read by scanning and sampling an image at low pixel density with a rectangular aperture, the density level Qx of the pixel of interest and each of the pixels adjacent to the pixel of interest in the longitudinal direction A method of predicting and restoring an image with high pixel density by determining _the density levels _q In advance, the left and right directions, right diagonal directions, and
Concentration gradients in the vertical direction and left diagonal direction
G1, G2, G3, and G4 are calculated, and the density levels q _X1 and q _X2 of each of the small pixels are determined in the following manner according to the maximum density gradient G. An image predictive restoration method. 1 When G・MAX=G1 Compare with the preset threshold level T, and when G1<T, set q _X1 = q _X2 = Q _X. When G1≧T (i) If Q _X = 1, set q _X1 = q _X2 = 1. (ii) If Q _X = 0, set q _X1 = q _X2 = 0. (iii) If Q _X ≠ 0, according to the density levels Q _B and Q _G of each pixel adjacent above and below the pixel of interest. When Q _B < Q _G and Q _X > 0.5, q _X1 =
2Q _X −1, q _X2 =1. When Q _B <Q _G and Q _X ≦0.5, q _X1 =
0, q _X2 = 2Q _X. When Q _B ≧Q _G and Q _X ≦0.5, q _X1 =
1, q _X2 = 2Q _X -1. When Q _B ≧ Q _G and Q _X > 0.5, q _X1 =
2Q _X , q _X2 = 0. 2 When G・MAX=G2 or G4 When G3>G1 q _X1 = q _X2 = Q _X. When G3≦G1 (i) If Q _X = 1, set q _X1 = q _X2 = 1. (ii) If Q _X = 0, set q _X1 = q _X2 = 0. (iii) If Q _X ≠0, then Q _B <Q _G and Q _X >0.5, then q _X1 =
2Q _X −1, q _X2 =1. When Q _B <Q _G and Q _X ≦0.5, q _X1 =
0, q _X2 = 2Q _X. When Q _B ≧Q _G and Q _X ≦0.5, q _X1 =
1, q _X2 = 2Q _X -1. When Q _B ≧ Q _G and Q _X > 0.5, q _X1 =
2Q _X , q _X2 = 0. 3 When G・MAX=G3, let q _X1 = q _X2 = Q _X.