JP3825825B2 - Filter device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a filter device for removing noise from blocked image data.
[0002]
[Prior art]
An orthogonal transform coding method is known as a technique for compressing and coding an image signal with high efficiency. In this method, image signals are grouped into blocks for each predetermined number of pixels, and then orthogonal transform such as discrete cosine transform (DCT) is performed, and the converted coefficients are quantized and entropy encoded. is there.
[0003]
FIG. 3 shows an encoding and decoding apparatus using the DCT as an example of block encoding.
In FIG. 3, digitized image data is input from an input terminal 701, and is blocked in units of m × n pixels by a blocking circuit 702. The image data blocked in m × n pixel units is output to the motion determination circuit 703 and the DCT circuit 704. The motion determination circuit 703 determines whether the corresponding block is a moving image region block or a still image region block, and outputs the determination result to the DCT circuit 704 as block motion information. The DCT circuit 704 orthogonally transforms the blocked image data into DCT coefficients. At this time, the DCT circuit 704 adaptively switches the DCT processing according to the motion information from the motion determination circuit 703.
[0004]
The data converted into the DCT coefficient and the motion information of the corresponding block are input to the quantization circuit 705, and the DCT coefficient is quantized in units of blocks. Quantization is performed by dividing the DCT coefficient by the quantization coefficient.
[0005]
The quantized data, the motion information, and the quantization parameter are encoded by a technique such as entropy encoding in the variable length encoding circuit 706 to obtain a desired data transfer rate. The encoded data is output from the output terminal 707, formatted in a transmission path or a recording medium in a subsequent circuit, and then recorded on the transmission or recording medium.
[0006]
At the time of decoding, the data transmitted or reproduced from the recording medium is input from the input terminal 708 to the ECC circuit 709 and the code error is corrected, and then decoded by the variable length code decoding circuit 710. The decoded data is inversely quantized by the inverse quantization circuit 711 based on the quantization parameter, and converted into DCT coefficients. At this time, the inverse DCT circuit 712 performs inverse DCT processing based on the motion information, converts DCT coefficients into image data, and writes the image data in the frame memory 713. The digital image data written in the frame memory 713 includes noise due to quantization error because the quantization circuit 704 and the inverse quantization circuit 710 perform quantization and inverse quantization processing. Therefore, after filtering processing for noise reduction is performed by the filter circuit 714 and the noise component is attenuated, it is output from the output terminal 715 and displayed on a monitor or the like.
Noise due to the quantization error is roughly classified into two types: mosquito noise appearing near the edge and block distortion appearing at the block boundary.
[0007]
FIG. 4 is a configuration example of a filter for reducing mosquito noise.
The reproduced digital image signal is input from the input terminal 301 in raster order. DL302, DL303, DL304, and DL305 are delay elements, and each delays one pixel. The pixel of interest to be filtered is stored in the DL 303, and the pixel 2 pixels to the right of the pixel of interest on the screen is input to the input terminal 301, the pixel right one pixel in the DL 302, and the pixel left one pixel in the DL 304 , DL305 stores two pixels to the left, and is a horizontal 5-tap filter. The difference absolute value calculation circuits 306 to 309 calculate the difference absolute value between the pixel of interest and its four left and right pixels, and the results are input to the comparators 311 to 314, respectively. A threshold for filter pixel selection is input to the input terminal 310, and the comparators 311 to 314 indicate “1” when the input absolute difference value is equal to or smaller than the threshold, and “0” otherwise. "Is output. Accordingly, the results of the comparators 311 to 314 are input to the counter 315, and the number of comparators that output “1” is counted. The counter 315 holds the number of pixels used for filtering (excluding the target pixel).
[0008]
The comparison results of the comparators 311 to 314 are input to the selection switches 316 to 319. When “1” is input, the # 1 side is selected, and when “0” is input, # Select the 0 side. Accordingly, the selection switches 316 to 319 output the value of each pixel when the # 1 side is selected, and output “0” when the # 0 side is selected.
[0009]
The output of the selection switches 316 to 319 and the pixel value of the target pixel from the DL 303 are input to the adder 320, and addition is performed. The number of pixels used for filtering (excluding the target pixel) is input from the counter 315 to the divider 321, and the addition result of the adder 320 is finally the number of pixels used for filtering (the above number + 1). Divide and output to output terminal 322.
With the above configuration, noise with a low amplitude of ± th or less is reduced, and the filter operates as a mosquito noise reduction filter.
[0010]
[Problems to be solved by the invention]
Although the filter configuration as described above has an effect of reducing mosquito noise, it is necessary to use a filter of another method for reducing block distortion, and there is a problem that the circuit scale increases.
[0011]
The present invention has been made to solve the above problems, and a filter device capable of reducing noise having different properties such as mosquito noise and block distortion without increasing the circuit scale, and the use thereof. An object of the present invention is to obtain a decoding device.
[0012]
[Means for Solving the Problems]
According to the first aspect of the present invention, in the image data including the mosquito noise and the block distortion obtained by decoding the image data that is divided into blocks and orthogonally transformed and encoded for each predetermined number of pixels in the vertical and horizontal directions, the pixel of the target pixel A difference value detecting means for detecting a difference value between each value of the pixel and a plurality of neighboring pixels, a comparing means for comparing each detected difference value with a predetermined threshold value, and Selecting means for selecting and outputting the pixel value when the comparison result of the comparing means is less than or equal to the threshold value, and outputting zero when the difference value is greater than the threshold value. And the pixel value of the pixel of interest are added together, and the result of the addition is divided by the value obtained by adding 1 to the number of pixel values selected by the selection means, and output by the calculation means as a filter output. A filter that reduces the noise, detects whether the pixel of interest is at the boundary of the block, and if it is detected that the pixel of interest is at the boundary of the block according to the detection result, It is characterized by a filter device that causes the arithmetic means to function as a filter for reducing block distortion by changing means for changing the threshold value to be large.
[0016]
[Action]
In the present invention, filter processing for reducing mosquito noise is normally performed. When the pixel of interest is at a block boundary, the threshold value is changed to perform filter processing for reducing block distortion.
[0018]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 1 is a block diagram showing an embodiment of the present invention. In FIG. 1, the parts denoted by the same reference numerals 301 to 322 as those in FIG. 4 are substantially the same as the parts in FIG. 101 is a pixel counter that counts the input pixels from the input terminal 301 to determine whether or not it is a block boundary, 102 is a shift register that doubles the threshold value input from the input terminal 310, and 103 is the pixel counter 101 It is a selection switch that selects a threshold value according to a value.
[0019]
Next, the operation will be described.
The reproduced digital image signal is input from the input terminal 301 in raster order. The delay elements DL302, DL303, DL304, and DL305 each delay one pixel. The pixel of interest to be filtered is stored in the DL 303, and the pixel 2 pixels to the right of the pixel of interest on the screen is input to the input terminal 301, the pixel right one pixel in the DL 302, and the pixel left one pixel in the DL 304 , DL305 stores two pixels to the left, and is a horizontal 5-tap filter. The pixel counter 101 counts the pixels input from the input terminal 301 and determines whether or not the pixel stored in the DL 303 is a pixel at the block boundary. For example, the horizontal position of the input pixel is counted, and it is determined whether it is a pixel on the horizontal boundary of the block. The difference absolute value calculation circuits 306 to 309 calculate the difference absolute value between the pixel of interest and its four left and right pixels, and the results are input to the comparators 311 to 314, respectively.
[0020]
A threshold for selecting a pixel used for filtering is input to the input terminal 310. The selection switch 103 receives threshold data from the input terminal 310 and threshold data that has been doubled through the shift register 102. The selection switch 103 is controlled by the pixel counter 101, and selects the # 2 side when the pixel of interest stored in the DL 303 is a pixel at the block boundary, and otherwise selects the # 1 side.
[0021]
The comparators 311 to 314 receive a threshold value from the selection switch 103 and output “1” if the input absolute difference value is equal to or smaller than the threshold value, and “0” otherwise. The counter 315 receives the results of the comparators 311 to 314 and counts the number of comparators that output “1”. Therefore, the counter 315 holds the number of pixels (excluding the target pixel) used for filtering.
[0022]
The comparison results of the comparators 311 to 314 are input to the selection switches 316 to 319. When “1” is input, the # 1 side is selected, and when “0” is input, # Select the 0 side. Accordingly, the selection switches 316 to 319 output the value of each pixel when the # 1 side is selected, and output “0” when the # 0 side is selected.
[0023]
The output of the selection switches 316 to 319 and the pixel value of the target pixel from the DL 303 are input to the adder 320, and addition is performed. The number of pixels used for filtering (excluding the target pixel) is input from the counter 315 to the divider 321, and the addition result of the adder 320 is finally the number of pixels used for filtering (the above number + 1). Divide and output to output terminal 322.
[0024]
With the above configuration, when the target pixel is a pixel at the block boundary, the threshold value used by the comparators 311 to 314 is twice that at the time of normal mosquito noise removal. Therefore, a stronger filter is used at the block boundary. Both mosquito noise and horizontal block distortion can be reduced simultaneously.
[0025]
FIG. 2 is a diagram showing a second embodiment of the present invention. This embodiment is an example of a 3 × 3 two-dimensional filter.
In the figure, 201 is an input terminal for reproduction data, 202 to 209 are delay elements, 210 to 217 are circuits for calculating the absolute difference between the target pixel to be filtered and adjacent pixels, and 218 is a block boundary by counting the input pixels. 219 is a threshold value table that stores threshold values used for pixel selection, 220 to 227 are comparators that compare the difference absolute values with threshold values, and 228 is a pixel value that is used for filtering. Counter for counting the number, selection switches 229 to 236 for selecting pixel values used for filtering, 237 for an adder for adding the selected pixel values, 238 for a divider for averaging, and 239 for an output terminal .
[0026]
Next, the operation will be described.
Playback data is input to the input terminal 201 in raster order. The input pixel values are sequentially delayed by DL 202 to 209. When the pixel of interest to be filtered is stored in the DL 205, the pixel value at the lower right of the pixel of interest on the screen is input to the input terminal 201, the pixel value one pixel below is input to the DL 202, The lower left pixel value is DL204, the right pixel value is one pixel, DL206 is the left pixel value, DL207 is the upper right pixel value, DL208 is the upper pixel value, and DL209 is the upper pixel value. The delay time of each delay element is set so that the upper left pixel value is stored. In the case of the NTSC system, pixel values on one line in the same field with respect to the target pixel are input or stored in DL207 to 209, and pixels on one line in the same field are input or stored in input terminals 201 and DL202 and DL203. .
[0027]
The difference absolute value calculation circuits 210 to 217 calculate the value of the DL 205 that is the target pixel and the difference absolute value of each of the eight adjacent pixels.
The pixel counter 218 counts the pixels input from the input terminal 201 and determines whether or not the target pixel stored in the DL 205 is a pixel at the block boundary. For example, the horizontal and vertical positions of the input pixel are counted to determine whether it is a pixel on the horizontal or vertical boundary of the block.
[0028]
The threshold value table 219 stores a threshold value used in the case of a block boundary pixel and a normally used threshold value, and selects a threshold value to be output based on the determination result of the pixel counter 218. It has become. The content of the table is, for example, a table having a value larger than a normally used threshold value for block boundary pixels.
[0029]
The comparators 220 to 227 are input with threshold values from the threshold value table 219, and are input with absolute difference values from the absolute difference calculation circuits 210 to 217, respectively. If the input difference absolute value is less than or equal to the threshold value, “1” is output, otherwise “0” is output. The counter 228 receives the results of the comparators 220 to 227 and counts the number of comparators that output “1”. Therefore, the counter 228 holds the number of pixels (excluding the target pixel) used for filtering.
[0030]
The comparison results of the comparators 220 to 227 are input to the selection switches 229 to 236. When “1” is input, the # 1 side is selected, and when “0” is input, # Select the 0 side. Accordingly, the selection switches 229 to 236 output the value of each pixel when the # 1 side is selected, and output “0” when the # 0 side is selected.
[0031]
The adder 237 receives the outputs of the selection switches 229 to 236 and the pixel value of the target pixel from the DL 205, and performs addition. The number of pixels used for filtering (excluding the target pixel) is input from the counter 228 to the divider 238, and the addition result of the adder 237 is finally the number of pixels used for filtering (the above number + 1). Divide and output to output terminal 239.
[0032]
With the above configuration, when the target pixel is a pixel at the block boundary, the threshold for the block boundary pixel is selected from the threshold table, so that a stronger filter is applied at the block boundary, and mosquito noise, horizontal and horizontal Both noises in vertical block distortion can be reduced simultaneously.
[0034]
【The invention's effect】
According to the present invention, since the circuit for reducing mosquito noise can be shared with the circuit for reducing block distortion, both noises can be reduced simultaneously without increasing the circuit scale.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of the present invention.
FIG. 2 is a block diagram showing a second embodiment of the present invention.
FIG. 3 is a block diagram of a conventional encoding and decoding apparatus.
FIG. 4 is a block diagram of a conventional filtering circuit.
[Explanation of symbols]
301, 201 Playback data input terminals 302 to 305, 202 to 209 Delay elements 306 to 309, 210 to 217 Difference absolute value calculation circuit 310 Pixel selection threshold input terminals 311 to 314, 220 to 227 Comparators 315, 228 Counter circuits 316 to 319, 229 to 236 Select switches 320, 237 Adders 321, 238 Dividers 101, 218 Pixel counter 102 Shift register 103 Select switches 322, 239 Output terminals

Claims (1)

In image data including mosquito noise and block distortion obtained by decoding image data that has been divided into blocks and orthogonally transformed and encoded for each predetermined number of pixels in the vertical and horizontal directions. Difference value detecting means for detecting a difference value between each pixel value and each pixel value;
Comparison means for comparing each detected difference value with a predetermined threshold value;
Selection means for selecting and outputting the pixel value when the comparison result of the comparison means is less than or equal to the threshold value, and for outputting zero when the difference value is greater than the threshold value;
An arithmetic means for adding the output of each selection means and the pixel value of the pixel of interest, dividing the addition result by the value obtained by adding 1 to the number of pixel values selected by the selection means, and outputting the result as a filter output To configure a filter that reduces mosquito noise,
Whether or not the pixel of interest is at the boundary of the block is detected, and when it is detected that the pixel of interest is at the block boundary according to the detection result, the threshold value is increased. A filter device characterized by causing the arithmetic means to function as a filter for reducing block distortion by changing means for changing.

Images