JPH02294784A - Method for converting image scale factor - Google Patents

Abstract

PURPOSE:To calculate the gradation of reconstituted digitized image data considering the gradation of original image data by interpolating between plural quantized data by the prescribed number of interpolating data to obtain similar analog data, sampling the obtained data by a sampling interval corresponding to a conversion scale factor to obtain quantized data after scale factor conversion. CONSTITUTION:In the case of contracting or expanding multi-gradation digital image data read out by an image scanner or the like by converting its scale factor, many interpolating data are found out between respective data to approximate analog data assuming information cut out at the time of reading or digitizing data. The reproduced quantized data are sampled by the sampling interval corresponding to a contracting or expanding scale factor. Consequently, the thinning or interpolation of data can be uniformed over the whole areas and the feature of an original image can be sufficiently reflected to gradation.

Description

[Detailed Description of the Invention] [Industrial Application Field 1] This invention relates to digitalization (including quantization and Umemotoization).
The present invention relates to an image magnification conversion method for reducing or enlarging a ii7J image.

[Jubei's technique 1 To reduce or enlarge a digitized image, the first step is to convert the pixels arranged vertically and horizontally into the number of pixels corresponding to the conversion magnification of the image. The first step is to calculate the IW image information such as the gradation to be placed in each pixel. Among the above, calculation of image information such as the second gradation generally requires a complex algorithm and requires a large memory capacity. For this reason, in Jubei, various methods have been proposed, such as thinning with a weight α added to the conversion of the number of pixels, average value interpolation, etc., as described in Section 2.1.

Therefore, an example of the conventional thinning method will be explained with reference to PJG diagrams to Figures 8. The j7s6 diagram shows the original data for 10 pixels before magnification conversion, and the dot (pixel) positions are placed at equal intervals of 17g on the horizontal pongee.
The vertical pongee has m gradation data simulated.

Figure 7 shows the second step to reduce the original data to 70%.
liS4 shows the case where the 6th dot is thinned out.

In addition, in FIG. 8, the second, fifth, and eighth dots are similarly thinned out to reduce the size to 70%. Next, an example of an enlarged image obtained by conventional simple mean value interpolation will be shown with reference to FIGS. 9 and 10. FIG. 9 is a 130% enlargement of the original data in FIG. 6, with 13 dots compared to 10 dots in the original data. Interpolation is given between the second and third, fourth and fifth, and sixth and seventh original data as a simple average value of each data.

In addition, for the 10th younger brother figure, when expanding by 130%, the 2nd and 3rd, the 5th and 6th, and the 8th and 9th middle openings of the original data are given as a simple average value of each data. It is being

[Problem to be Solved by the Invention 1] However, in this conventional image magnification conversion method, even if the image magnification is reduced to the same value (70%), the gradation difference between FIG. 7 and FIG. 8 is different depending on the dot thinning position. The gradation of the original data of the f:PJG diagram, for example, up to the 4th dot which shows the maximum value, the gradation tends to increase gradually. This trend has completely disappeared.The same is true for the enlargements shown in Figures 9 and 10, where differences in gradation occur between the enlarged data or between the enlarged data and the original data. It can be clearly seen that

As described above, the conventional image magnification conversion method of opening and closing specific dots or interpolating data to a predetermined interval has the problem that the results change depending on where the processing is applied. When calculating the gradation of a pixel using such a conventional method, the gradation changes of the original Wi image are ignored, and the image quality deteriorates in terms of smoothness, preservation of thin lines, or blurring of outlines. Problem 7α
was there. [Objective of the Invention 1] This invention was made in view of the problems in conventional technology, and the thinning and interpolation of data is uniformly performed over the entire area, and when determining the gradation, the original image is used. The purpose of the present invention is to provide a method for converting the digitized image magnification that fully reflects the characteristics, requires no interpolation processing time, and requires less memo ')t-1. 1 Means for Solving the Problem 1 The purpose of this invention to achieve the above object is that, in the image magnification conversion method for reducing or enlarging a digitized image, The image information between each quantized data that was lost during the process is approximated and reproduced using a predetermined number of interpolated data, and the reproduced quantized data is sampled according to the conversion magnification.
By sampling at Vk, the quantized data of the digitized image after magnification conversion is obtained. Further, in an image magnification conversion method for reducing or enlarging a digitized image, each quantized data of the digitized image is divided at a predetermined number of division points, and the quantized data and the From the division and α, a sampling interval corresponding to the f rate is determined, and for each sampling interval, interpolation processing is performed by referring to the fostered data recorded in +'+ij, and the digitized image after magnification conversion is Quantized data is also a vf characteristic.

[Effect] When reading an image to obtain digitized data, there was information between each piece of data that was truncated during reading and digitization. Therefore, when converting the magnification to reduce or enlarge an image, a large number of interpolated data are obtained between each data, and Kintetsu analog data is created that virtualizes the information that was discarded during reading or digitization. Then, this reproduced quantized data is converted to Umemoto at a sampling interval corresponding to the reduction or enlargement magnification. Digitized LI! data obtained in this way is reduced or enlarged! This is the quantized data of the i image. If the approximate analog data is re-sampled and the quantized data of the image after magnification conversion is converted, much of the image information data such as interpolated gradations will be discarded, reducing processing time and memory. It is not preferable from the point of view of effective use. Therefore, the quantized data is divided into a predetermined number of divisions, and during interpolation processing, the sampling position is first determined in accordance with the conversion magnification, and this determined r:. Calculates interpolated tone data only at the sampling position. [Example 1] A simple example of the present invention will be explained for a dog based on FIGS. 1 to 5. Note that FIG. 6 will be referred to in the following explanation as it is as original data before magnification conversion.

The image to which the present invention is applied is uppercase Wi-tone digital image data when read by an image scanner or the like. Therefore, in order to convert the image magnification, it is necessary to convert not only the number of pixels in the main scanning direction or the sub-scanning direction, but also the gradation information. The PpJl diagram shows the flow of the image magnification conversion method. In step (2), the interpolation accuracy, that is, how many divisions between two points of the original data is to be divided is determined. For example, if the number of divisions is 10, the area between two points is divided into 10 areas i! It will be i1+,
Interpolation in step 2 can be performed assuming that pixels exist at a total of 11 locations, which correspond to the start and end points of each region. Also, minutes? !
Assuming that 4 losses is 100, it is sufficient to divide the 2-point question into 100 regions and perform interpolation (step ■) assuming that there are pixels in a total of 101 locations. However, Oi! In the case of 10 losses, the start point of the first area (first data) and the end point of the 10th area (11th data), and in the case of a division loss of 100, the start point of the first area! The end of the @10011'l area is the two points of the original data, and the numbers of interpolated data actually calculated are 9 and 9La, respectively. The number of divisions described above is closely related to the degree of analog approximation when calculating approximate analog data from digitized image data at the interpolation port, and the accuracy of the conversion magnification in terms of processing quality. Once the number of divisions and conversion magnification for the correction are set,
Next, in step ■, the sampling interval is determined from these two values using the following formula. Sampling interval divided into two rl00/Conversion magnification %... (Summer) For example, if the original data is divided into 10 between two points and the conversion magnification is 100%, the sampling interval will be 10, and the data of m 1 @ Ll is adopted by sampling, the interval 10#l/'L
The data of the 11th fish selected for sampling is the same as the original data. In the above two steps, firstly, uniform thinning or darkening effect over the entire image is achieved, and secondly, the calculation of the WI tone value of the reconstructed image data taking into consideration the gradation of the original image data. This will be explained with reference to Figure 3, which is the original data in Figure 6 reduced to 70%. It is approximated by a straight line, and the points on the straight line are used as interpolated values. In other words, in Figure 2, the original data is 10 dots, and each data interval is divided into 10 to measure the approximate analog data.The degree of analog approximation of this approximate analog data increases as the number of divisions increases. However, the precision of the conversion magnification of the re-vt image with respect to the original image will also increase. Therefore, step
By substituting % into the formula 0(), the sampling interval is calculated as 14. Therefore, the result of sampling the approximate analog data in Fig. 2 obtained earlier at 14 intervals is the data reduced to 70% of the data in Fig. 3. Fig. 3 improves the gradation of the original data in Fig. 6. It can be seen that this has been reproduced better than the conventional model, such as Figure 7. Similarly, FIG. 5 shows an example in which the original data in FIG. 6 is divided into 10 parts and expanded to 130%. According to the above formula (X), the sampling interval becomes 7 as shown in Fig. 4, and the result is shown in Fig. rjS5. Even in this fifth fAI::, the gradation of the original data is well reproduced except for the maximum value.

However, the feature near the maximum value, which is one of the biggest features of the original data in Figure 6, is not reproduced. This is due to the accuracy of the sampling interval, and is the result of actually performing sampling with the calculated value of the exact sampling interval, 7.69, as 7. In order to avoid such inaccuracies, the number of divisions is increased by one order of magnitude to 100, which results in a sampling interval of 76, which increases the accuracy by one order of magnitude.

However, performing interpolation first and then sampling as described above generates a lot of data that is not actually used, which is inefficient in terms of processing time and memory capacity. Therefore, this problem was solved by executing steps ■ and ■.

In other words, in step 2, we determine in which interval and at which position of the original data the data to be adopted for sampling are located before performing the interpolation process, and in step 2, we identify only the data to be adopted at the time of sampling. calculate. If we look at this method with reference to Figure 2, we can see that the first dot of the first dot is Ichiba, and the second data to be adopted is the second one of the original data. This data is located between the dot and the third dot, and is located 4 points away from the second dot of the original data. Therefore, if the gradation data at this position is calculated as the data to be adopted for sampling and is used as the second data of the re-death data, there is no need to correct the other unnecessary data and process it. This can contribute to time reduction and effective use of memory. In particular, if the number of divisions in step ■ is increased by one digit, for example, in order to improve the accuracy of Ji calculation, the processing time and memory capacity will increase approximately 10 times using only the steps ■ and ■. ■
The addition of and ■ will result in almost no change in processing time and memory capacity. [Effects of the Invention] This invention obtains approximate analog data by interpolating single child data with a predetermined number of interpolation data, samples it at a sampling interval corresponding to the conversion magnification, and obtains quantized data after the magnification conversion. Therefore, the spacing or interpolation effect is applied uniformly to the entire image, and it is possible to calculate the gradation of the reconstructed digitized image data taking into account the floor Iv4 of the original image data. Furthermore, the position of the data that should be adopted through sampling is calculated in which interval of the original data is calculated before the interpolation process, and the interpolation process is performed only on the data that should be adopted through sampling. , processing time is shortened and memory capacity does not increase.

[Brief explanation of drawings]

Figures 1 to 5 show various examples of the method of the present invention, Figure 1 is a 7-flow chart of the present invention, Figure tjS2 is a process explanatory diagram for reducing original data to 70%, Figure 3 is a reduced data diagram that reduces the original data to 70%, and Figure 4 is a processing explanatory diagram for expanding the original data to 130%.
Figure S5 is an enlarged data diagram that enlarges the original data to 130%.6 Figure 6 is a diagram of the original data, Figure 7 is a diagram of 70% reduced data of the original data using the conventional force method, and the rjSS diagram is also the thinned out position. Figure 9 is a 70% reduced data diagram with different 1 3 0 q. Figure 10 shows 130% enlarged data with different interpolation positions. Figure l Figure 5 Figure λ Figure 3 Figure 6 Figure 7 Figure 8 Dot position Dot position

Claims (2)

[Claims] (1) In an image magnification conversion method for reducing or enlarging a digitized image, image information between each quantized data lost during digitization is approximated and reproduced by a predetermined number of interpolated data. At the same time, the reproduced quantized data is sampled at a sampling interval corresponding to the conversion magnification, thereby obtaining the quantized data of the digitized image after the magnification conversion. (2) In an image magnification conversion method for reducing or enlarging a digitized image, each quantized data of the digitized image is divided at a predetermined number of division points, and the quantized data and From this dividing point, a sampling interval corresponding to the magnification is determined, and the interpolation process is performed by referring to the quantized data for each sampling interval, and the quantized data of the digitized image after magnification conversion is obtained. How to convert image magnification.