KR20030037007A - Method for removal image noise using selective median filter - Google Patents

Abstract

PURPOSE: A method for filtering video noise using a selective median filter is provided to selectively remove noise using a pre-processing filter that discriminates an original image from noise to filter the noise so as to improve picture quality. CONSTITUTION: An input pixel value is compared with pixel values of pixels in proximity to the input pixel, to judge whether or not filtering is carried out for the input pixel. According to the judged result, filtering is performed using the current pixel value and a predetermined number of pixel values of pixels adjacent to the current pixel. When the input pixel value becomes a maximum or minimum value for the neighboring pixel values in the judgement step, the filtering is carried out. The filtering step outputs median of the input pixel value and the predetermined number of neighboring pixel values(S30).

Description

Method for removal image noise using selective median filter

The present invention relates to a method for removing image noise using an optional median filter. In particular, image quality is improved and computational throughput is improved by using a preprocessing filter that distinguishes the original image from noise and reduces the median computation amount according to an arbitrary pixel selection. The present invention relates to a method for removing image noise using an optional median filter which can achieve a reduction.

The noise contained in the image data not only degrades the visual quality, but also greatly reduces the compression efficiency of the still image or video compression system. In general, compared to an image with no noise, the compression efficiency under the same compression condition is worse than several to several tens of times depending on the type or intensity of the noise. Because of the high frequency components included in the noise, many high frequency component coefficients remain even after quantization after transform coding such as a DCT (Descrete Cosine Transform) transform. This is because it is relatively longer than when it is not mixed. Therefore, preprocessing filter processing to remove noise included in the image before compression coding is very important for maintaining compression efficiency. However, the noise filter used in the past has a problem that it is accompanied with the disadvantage that the sharpness of the screen is lowered by lowering the sharpness of the original image. This is a problem caused by indiscriminately filtering without distinguishing the pixel value included in the original image and the noise included in the image.

Now, let's look at the conventional filter process.

1A and 1B are a corresponding pixel and control block diagram illustrating a conventional median filtering concept. As shown in FIGS. 1A and 1B, as an example of a pre-processing filter according to the prior art, the 3X3 median filter 10 may include an input current pixel value a and eight pixel values adjacent to the position (a). A total of nine pixel values including 1,2, ... 7, 8) are sorted in size order, and then the pixel values a 'newly replaced by the values corresponding to the middle are outputted. In other words, the current pixel value a is replaced with the median value among the pixel values in the 3X3 window shown in FIG. 1A. The noise reduction method of the conventional median filter is known to show excellent performance, especially when impulsive noise is added to the image. When noise is mixed at a pixel position, the pixel value is greatly distinguished from the pixel value around it, and there is a lot of room for a significant difference. If it is changed to the median of 3X3, the neighboring pixels (1, 2) This is because noise can be eliminated while maintaining the sharpness with. For this purpose, it is possible to eliminate noticeable noise by substituting the middle value compared with the surrounding pixel value.

However, such a conventional median filter has the following problems.

If the current pixel value a has a value which is not the maximum or minimum among the eight arranged pixels (1, 2, ..., 7, 8), that is, the probability that it is not noise is high, Instead of the pixel value a, an intermediate value a 'is assigned. In this case, since the pixel value at the corresponding position of the filtered image has lost the original pixel value, the image using the noise reduction filter has a subjective feeling of crushing of the screen compared to the image that is not. Cause. In case of losing the original pixel value as described above, even if the inputted current pixel contains noise, the pixel value with noisy noise does not have a significant pixel value such as being maximum or minimum compared with the surrounding pixel values. If not, it may be better to keep the screen clear without filtering in terms of maintaining the sharpness of the screen. This is because the filtering to remove this noise basically has the characteristics of low pass filtering.

On the other hand, another problem of the conventional median filter is a problem of the amount of computation required to sort the neighboring pixel values in size order. That is, in order to perform the intermediate value filter processing, the size of the pixel value (a) of the current position and the eight pixel values (1, 2, ..., 7, 8) adjacent to each other should be arranged in order. To do this, this process should be performed for 24,708 pixels (if QCIF = 176 X 144), and if applied to 10 frames per second, 247,080 alignments should be performed per second. It is not possible to apply the conventional median filter shown in FIG. There was a problem.

Accordingly, an object of the present invention is to provide an image noise removing method using a selective median filter that selectively removes noise by using a preprocessing filter for discriminating and processing the original image and noise to improve the clarity of the image quality.

In addition, another object of the present invention is to minimize the amount of calculation of the pre-processing filter by the median calculation according to the selection of any pixel in the adjacent pixels to remove the image noise using the low-cost system and the selective median filter that can utilize the existing system To provide a method.

1A and 1B are corresponding pixel and control block diagrams illustrating a conventional median filtering concept;

2 is a control block diagram illustrating a preprocessor for selective noise cancellation of the present invention;

3 is a diagram illustrating an operation process of the filter processing determination unit of the present invention in one dimension;

4 is a flowchart illustrating an optional noise reduction process of the present invention;

5A and 5B are processing block diagrams of a pixel and filter processing unit showing a cross-shaped median filter processing process according to a first embodiment of the present invention;

6A and 6B are processing block diagrams of a pixel and a filter processing unit showing a diagonal value filter processing process according to a second embodiment of the present invention;

7 is a process block diagram of a filter processing unit showing another filter processing process according to the third embodiment of the present invention.

Explanation of symbols on the main parts of the drawings

10: 3X3 median filter 20: filter processing determination unit

21: filter processing unit

In order to achieve the object of the present invention, an image noise removing method using an optional median filter of the present invention comprises comparing the input pixel value with a predetermined number of pixel values adjacent to an input pixel position. A filter processing determining step of determining whether or not to filter the pixel; And a filter processing step of filtering using the current pixel value and a predetermined number of pixel values adjacent thereto depending on whether the filter processing is determined.

At this time, in the filter processing determination step, the filter processing step is performed when the input pixel value becomes the maximum or minimum value for the adjacent pixel value compared to four adjacent pixel values vertically, horizontally, and adjacent to the input pixel. One is preferable.

The filter processing may output an intermediate value of the input pixel value and a total of five pixel values adjacent to the top, bottom, left, and right sides, or the input pixel value and the total of five pixel values adjacent to the top, right, left, bottom, and bottom of the input pixel value. The median value may be output, or the median value of the input pixel value and a predetermined number of pixel values adjacent thereto may be output.

Meanwhile, in the filter processing step, it may be preferable to output an average value of weights selectively added to input pixel values and a predetermined number of pixel values adjacent thereto as pixel values.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the present invention.

2 is a control block diagram illustrating a preprocessor for selective noise cancellation of the present invention. As shown in FIG. 2, the preprocessor according to the present invention includes a filter processing determining unit 20 that determines whether noise is included, and a filter processing unit in which actual filter processing is performed according to whether the filter processing determining unit 20 determines. It consists of 21.

The preprocessor operation of the present invention described above will be described.

The filter processing determination unit 20 determines whether or not the current pixel P, which has been previously inputted for the selective filter application, contains noise. That is, when the filter processing determination unit 20 determines that the pixel P at the current position contains noise, the filter processing unit 21 filters the output P 'and outputs the current pixel P. If it is determined that does not contain noise, the filter processing by the filter processing section 21 is omitted so that the current pixel value P is maintained as it is. Accordingly, it is possible to prevent the blurring of an image caused by the filter process that loses the original pixel value P, which was a problem of the conventional method, and to prevent the original pixel value P from being noisy. By outputting the result, the operation of the filter processor 21 may be omitted, thereby obtaining a calculation amount reduction effect.

At this time, the filter processing determiner 20 determines the four pixel values and the current pixel values (up, down, left, right, or diagonally) of the current pixel in order to determine whether the pixel value P of the current position includes noise. By comparing P), it is determined whether the current pixel value P is the maximum or minimum value by comparing with the surrounding four pixel values. This is to determine whether there is a significant difference between the current pixel and the surrounding pixels. The reason for determining the maximum and minimum is that in general, when the current pixel includes noise, the pixel value P has a value that is clearly different from the neighboring pixel value.

3 is a diagram schematically illustrating an operation process of the filter processing determining unit of the present invention in one dimension. As illustrated in FIG. 3, the filter processing determiner 20 may include P _left , P _right , and P when the current pixel value P _center is between two neighboring pixel values, that is, in the left and right directions. _If it is within the range determined by _up and P _bottom , it is determined that it is not noise.

4 is a flowchart illustrating an optional noise reduction process of the present invention. As shown in FIG. 4, it is first determined whether the current pixel value P _center is located between the pixel values P _left and P _right positioned to the _left and _right of the current pixel (S10). If not, it is determined by considering the relationship between the pixel values (P _up , P _bottom ) located above and below (S20). As a result of the determination, when the current pixel value P _center is not located between the pixel values P _up , P _bottom , P _left , and P _right positioned _up , down, _left , and _right , the filter processor performs an intermediate value filter process ( S30), if at least one position therebetween is thereby output as the current pixel value (P _center) is the current pixel value (P _center) the input to bypass the filter processing (S40).

Although the description of the present invention shows an example of judging in the left and right direction first and then in the up and down direction, it may be implemented in a form of judging the left and right after judging the up and down direction by changing the order.

5A and 5B are processing block diagrams of a pixel and filter processing unit showing a cross-shaped intermediate value filter processing according to a first embodiment of the present invention. As shown in FIGS. 5A and 5B, in order to reduce the computational throughput, the pixel of the current position P and the adjacent top, bottom, left, and right sides except for the four pixels 1, 2, 3, and 4 in the diagonal direction instead of the window of 3 × 3 are used. A total of five pixel values are arranged in size order using four pixel values (50), and then an intermediate value of five pixels is obtained (51). This method not only reduces the amount of calculation compared to the conventional method of obtaining nine median values, but also has an improvement effect in that the result of crushing is reduced when the filter process is performed.

6A and 6B are process block diagrams of a pixel and filter processing unit illustrating a diagonal median filter processing process according to a second embodiment of the present invention. As shown in FIGS. 6A and 6B, four pixels 1, 2, 3, and 4 in the diagonal direction are identical to the current pixel value P, instead of the pixel values of the current pixel up, down, left, and right as compared to FIG. 5. It is a form realized to apply (60) and calculate (61) an intermediate value.

By applying the intermediate value calculation process shown in FIGS. 5 and 6, it can be assumed that a predetermined number of pixels may be arbitrarily selected from the surrounding eight pixels to calculate the current pixel and the median value. As a result, when the calculation amount is reduced and the filtering process is performed, the same improvement effect as the result is reduced compared to the existing method can be obtained.

7 is a process block diagram of a filter processing unit showing another filter processing process according to the third embodiment of the present invention. As shown in FIG. 7, another embodiment in which a weighted average value with surrounding pixels is replaced with a current pixel value instead of a median value (70), and the weighted average calculated value is replaced with a pixel value at the current position (71). An example is shown. It is well known that this can be applied to all of the filter processing units shown in FIGS. 5 and 6.

As described above, the video noise elimination method using the selective median filter according to the present invention provides a real-time interactive application for video telephony, video communication in mobile networks by reducing pixel value calculation throughput and improving the sharpness of image quality. As a pre-processing filter for multimedia video services such as video application service on the Internet, it can be applied to various applications.

The present invention is not limited to the above-described embodiment, and it will be apparent that many modifications are possible by those skilled in the art within the technical spirit of the present invention.

Claims (8)

A filter processing determining step of determining whether or not to filter the input pixel by comparing an input pixel value with a magnitude between a predetermined number of pixel values adjacent to an input pixel position; And A filter processing step of selectively filtering using the current pixel value and a predetermined number of pixel values adjacent thereto according to the determination of the filter processing; Image noise removal method using a selective median filter, characterized in that made. The method of claim 1, wherein in the filter processing determination step, Image noise using an optional median filter characterized in that the filter processing step is performed when the input pixel value becomes the maximum or minimum value for the adjacent pixel value compared to a predetermined number of adjacent pixel values for the input pixel. How to remove. The method of claim 1 or 2, wherein in the filter processing determination step, An optional intermediate value filter is set so that the filter processing step is performed when the input pixel value becomes the maximum or minimum value of the adjacent pixel value compared with four adjacent pixel values vertically, horizontally, and horizontally adjacent to the input pixel. Image noise reduction method using. The method of claim 1 or 2, wherein in the filter processing determination step, Characterized in that the filter processing step is performed when the input pixel value becomes the maximum or minimum value for the adjacent pixel value compared to four adjacent pixel values adjacent to the upper left, upper right, lower left and right with respect to the input pixel. Image noise reduction method using selective median filter. 2. The method of claim 1, wherein the filter processing step outputs a median value of an input pixel value and a total of five pixel values adjacent to each other up, down, left, and right. The method of claim 1, wherein the filter processing step removes image noise using an optional median filter, wherein the median value of the input pixel value and a total of five pixel values adjacent to the upper left, upper right, lower left, and lower right is output. Way. 2. The method of claim 1, wherein the filtering process outputs an intermediate value between an input pixel value and a predetermined number of pixel values adjacent thereto. 8. The method as claimed in any one of claims 1 to 5, wherein the filter processing includes outputting, as a pixel value, an average value of weights selectively added to an input pixel value and a predetermined number of pixel values adjacent thereto. A video noise reduction method using an optional median filter.