JPH0659087B2 - Image signal processing method - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for binarizing an image signal obtained by scanning an image by an image sensor and then performing binarization.

(Technical background of the invention) Various methods for reading an image by an image sensor such as a CCD line sensor or a CCD area sensor and binarizing the image signal have been conventionally proposed. For example, the applicant of the present application has proposed a method in which an image signal is defocused and then compared with a constant binarization level (Japanese Patent Laid-Open No. 62-130068).
No.).

The blur correction here emphasizes the edges of the image by emphasizing the high frequency components of the image, and a high-frequency emphasis filter using first-order differentiation or Laplacian (second-order differentiation) is used. However, since such a high-frequency emphasis filter essentially uses differentiation, it is vulnerable to noise, and especially noise in the background region of the image signal is also emphasized, which significantly improves the image quality after binarization processing. There was a problem of lowering it.

Therefore, it has been considered by the applicant of the present application that the image signal is sliced at a predetermined background noise cut level close to the background level to remove noise included in the background. However, in this case, since the background density level differs depending on the image, how to determine the background noise cut level becomes a problem. That is, if this setting is wrong, background noise cannot be sufficiently removed, or the fluctuation range (dynamic range) of the effective signal level of the image signal is narrowed, resulting in deterioration of image quality.

Further, the binarization level may be set as a fixed value, but in reality, the binarization level cannot be set constant because the images have different densities.
Therefore, it is considered that a histogram of an image is obtained before the binarization process and the binarization level is determined from this histogram. For example, the appearance frequency for each density level is sequentially added from the background side, and the density level at which the added value becomes a fixed ratio (for example, 80%) of the total number of pixels is set as a binarization level (special Kaisho 56-16
No. 60).

However, in this case, the binarization level is greatly affected by the shape of the background mountain, that is, the extent of the mountain extension, and the optimum binarization level may not be set depending on the image. Therefore, there is a problem in that the image quality is deteriorated due to background noise being included or the image being skipped.

(Object of the Invention) The present invention has been made in view of such a situation, and when the image signal is defocused by high-frequency emphasis prior to the binarization processing, the image is processed with the background noise cut level close to the background level. When slicing a signal to remove background noise, the background noise cut level can always be set appropriately, and even if the image changes, the optimum binarization level can always be set with high accuracy, and the image can be binarized. An object is to provide an image signal processing method capable of improving the image quality after processing.

(Structure of the Invention) According to the present invention, the object is to remove background noise by slicing an image signal obtained by scanning a film projection image with an image sensor at a background noise cut level close to the background level. After that, in the image signal processing method of sharpening the image signal by performing blur correction and further binarizing the image signal by a predetermined binarization level, a histogram for the density of the image signal is obtained, and a mountain corresponding to the background of the histogram is obtained. Of the density levels having a frequency of about 1/40 of the total number of pixels of the image sensor, the density levels sequentially deviated by a constant density level from the density level including the image to the density level including the image, respectively. An image signal processing method characterized by setting a background noise cut level and a binarization level, It is made.

In other words, by setting the density level on the side of the density level including the image as a standard among the density levels that are about 1/40 of the total number of pixels, even if the peak width of the histogram corresponding to the background changes, that effect It is hard to receive.

(Embodiment) FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a conceptual diagram of a part of a concrete example thereof, and FIGS. 3A to 3C are diagrams showing signal waveforms during processing, and FIG. Shows a 3 × 3 matrix, and FIG. 5 shows a histogram for a negative film.

In FIG. 1, reference numeral 10 is a light source. Light from the light source 10 is guided to an image sensor 20 via a condenser lens 12, a film 14, a projection lens 16 and a mirror 18, and a projected image of the film 14 is directed to the image sensor 20. Form an image. The image sensor 20 is a CCD line sensor or C
It is formed of a CD area sensor or the like and is driven by a pulse supplied from a pulse circuit (not shown) to scan an image and output a time-series image signal a. The image signal a has an output waveform as shown in FIG. 3A when the film 14 is negative. 3A to 3C, the horizontal axis represents time t or pixel order, and the vertical axis represents output (density) level.

The image signal a for one scanning line or one screen is temporarily stored in the semiconductor memory 22 such as a line memory or a frame memory. This image signal a is input to the histogram means 24 composed of a CPU (not shown), and the histogram A for the density is obtained here as shown in FIG. The CPU has a background noise cut level detecting means 26 and a binarization level detecting means 2 which the CPU itself has.
7, the background noise cut level b and the binarization level f are obtained from this histogram A.

In the histogram A, the horizontal axis represents the output (density) level V of the image signal a and the vertical axis represents the frequency H.
For No. 4, a mountain B corresponding to the background appears on the low output level side. CPU is mountain B of this histogram A
On the right side, that is, near the skirt on the high output (density) level side, the background noise cut level b and the binarization level f on the higher output level side than this are determined.

These levels b and f are determined as follows. That is, when the total number of pixels of the image sensor 20 is N, about N /
A constant density level β is added to the density level α of the point C, which is closer to the image signal, among the output levels having a frequency of 40, and α + β is determined as the background noise cut level b. Further, a constant density level γ is added to the density level α at the point C to obtain α +
Let γ be the binarization level f. For example, when a line sensor of 4000 pixels is used and the density level can be set in 64 steps (corresponding to 6 bits), the density level α at the point C at which the frequency is about 100 is obtained, and the density level α for two steps is obtained. The density levels can be added to make α + 2 the base cut level b, and the density levels for 8 levels can be added to α to make α + 8 the binarization level f. Here, the frequency N / 40 is determined based on experiments. A constant concentration level β,
The optimum value of γ should be determined based on experiments.

The image signal a and the background noise cut level b are input to the background noise cut circuit 28, where noise included in the background region is removed. That is, this circuit 28
Is, for example, a comparator 30 and a switch 32 as shown in FIG.
The comparator 30 compares the image signal a with the background cut level b. Also, the switch 32 is the comparator 3
When 0 is judged to be a <b, the background cut level b is selected, and when a ≧ b, the image signal a is selected. As a result, the output C of this circuit 28 is sliced at the background noise cut level b, and the noise in the background region is cut, as shown in FIG. 3B. In this way, the background cut level b is close to the background level d (Fig. 3A),
It is set to a level slightly higher than this.

Reference numeral 34 is a blur correction circuit, for example, the high-frequency emphasis filter 3
6 is used to emphasize the high frequency component of the signal c and to emphasize the edge of the image.

The high-frequency emphasis filter 36 functions to emphasize the center pixel when the image space is an odd matrix in which the center pixel appears. As the high-frequency emphasis filter 36, for example, when each pixel data of a 3 × 3 matrix is set to a to i as shown in FIG. 4, the data e for the central pixel is obtained by using the data of the four pixels around it. = 5e- (b + d + h + f), and E is used as a new image signal. In this case, the high-frequency emphasis filter 36 is set as shown in FIG. 2, and each element of this matrix has a perimeter 4 centered on the central pixel.
The image data of one pixel is integrated and the sum E of the integrated values is obtained. The image signal e thus emphasized is compared with the binarization level f in the comparator 38 to obtain the binarization signal g.

As described above, since the background noise is removed by the background noise cut level b and the blur correction processing by high-frequency emphasis is performed, fluctuations due to small noise in the image signal, especially in the background area, are excessive during blur correction processing. It is never expanded. Therefore, it is not affected by background noise.

Further, the present invention does not add the appearance frequency of each density level from the background side of the histogram and set the density level at which the added value becomes a fixed ratio of the total number of pixels as a binarization level, but instead of the total number of pixels. Concentration level C with a frequency of about 1/40
Based on, the concentration level β is a constant concentration level β,
Since it is determined by adding or subtracting γ, it is possible to always determine the appropriate background level b and binarization level f even if the width or height of the peaks of the histogram with respect to the background changes depending on the image. Especially the binarization level f
The background noise does not become an obstacle even if it is set sufficiently close to the lower limit noise cut level b, and high-accuracy image reading is possible for images such as stamps and blue prints with low density.

As the smoothing circuit 50 of this embodiment, it is possible to use, for example, a median filter that adopts an intermediate value (median) of a 3 × 3 matrix centered on the center pixel as image data of the center pixel.

Further, the background noise cut level b and the binarization level f are set to a higher density level side than the background d as in the above embodiment when an image formed by a negative film is used, but when a positive film is used. On the contrary, it is, of course, set to the lower density level side than the background, and in this case, N located on the left side of the histogram peak is
Density levels α to β, γ
Is subtracted to determine the background noise cut level b = α-β and the binarization level f = α-γ, and the present invention also includes such a case.

(Effect of the Invention) As described above, the present invention uses the background noise cut level close to the background level, slices the image signal prior to the blur correction processing, and removes the noise of the background level. In the binarization, the background noise cut level and the binarization level are converted from the density level on the signal side of the image among the density levels having a frequency of about 1/40 of the total number N of pixels of the image sensor to include the density including the image. Since the density levels that are deviated by a constant density level to the level side are set as the background noise cut level and the binarization level respectively, the background noise is based on the position near the foot of the histogram corresponding to the background. The cut level and the binarization level can be determined. Therefore, the background noise cut level and the binarization level can be determined without being greatly affected even if the width of the peaks of this histogram changes, and the background noise cut level and the binarization level can always be set to appropriate levels even if the image changes. The value level can be set. Therefore, in the binarization process after blur correction,
There is no inconvenience that the background noise cut level is too close to the background to sufficiently remove noise, or conversely it is too far from the background level to sacrifice the dynamic range of the image signal. Since the binarization level can always be set to an appropriate level, background noise is not emphasized and the image is not skipped, and high-quality binarization processing is possible.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a conceptual diagram of a specific example thereof, FIGS. 3A to 3C are diagrams showing signal waveforms in a processing process, and FIG. 4 is a 3 × 3 matrix. FIG. 5 is a diagram showing a histogram. 20 ... Image sensor, 24 ... Histogram means, 27 ... Binary level detection circuit, 28 ... Underground noise cut circuit, a ... Image signal, b ... Underground noise cut level, d ... Background level , F …… Binarization level, A …… Histogram.

Claims (4)

[Claims] 1. An image signal obtained by scanning a film projection image with an image sensor by slicing a background noise cut level close to the background level to remove background noise, and then correcting the blur. In the image signal processing method of sharpening, and further binarizing by a predetermined binarization level, a histogram for the density of the image signal is obtained, and the mountain corresponding to the background of the histogram is the total number of pixels of the image sensor. Of the density levels with a frequency of about 1/40, the density levels that are deviated from the density level that includes the image to the density level that includes the image are sequentially shifted to the background density cut level and the binarization level. An image signal processing method comprising: 2. The image signal processing method according to claim 1, wherein the blur correction is performed by using a high-frequency emphasis filter. 3. The film projection image is a projection image obtained by a negative film, and the background noise cut level is
The image signal processing method according to claim 1, wherein the image signal processing method is set in the vicinity of the skirt on the high density level side of the mountain of the histogram with respect to the background. 4. A film projection image is a projection image obtained by a positive film, and the background noise cut level is set in the vicinity of the base of the low density level side of the mountain of the histogram with respect to the background. The image signal processing method according to claim 1.