JP2848904B2 - Electronic still camera - Google Patents

Description

The present invention relates to noise reduction of an electronic still camera.

(Prior Art) As a noise removal method for a television image composed of two dimensions in the horizontal direction and the vertical direction, a noise removal method only in the horizontal direction using an analog filter, and a vertical direction using a delay element for one horizontal scan period There is known a line noise elimination method using the correlation between the images.

(Problems to be Solved by the Invention) However, in the removal of fixed pattern noise of the image sensor CCD, a vertical stripe pattern can be removed by a horizontal noise (one-dimensional) analog noise removal method, but a horizontal stripe pattern can be removed. Cannot be removed. Further, since the line noise removing method removes only uncorrelated signals, fixed pattern noise having vertical stripe correlation cannot be theoretically removed. A video camera using a CCD area sensor is used when the subject is dark (the amount of input light is small).
The fixed pattern noise peculiar to the CCD appears due to the variation of the dark current of each pixel of the CCD and the like, which is one of the factors deteriorating the image quality. The appearance pattern of the fixed pattern noise differs depending on each CCD, and also on the predetermined single CCD depending on conditions such as temperature.

Conventionally, in order to eliminate this fixed pattern noise, a frame memory was used to increase the exposure time by increasing the exposure time to reduce the effect of noise, but this is expensive and the frame memory is expensive. I was

Further, the conventional long-time exposure method causes image blurring and is not suitable as a still image.

Therefore, the present invention relates to an electronic still camera using a frequency transform such as a discrete cosine transform (DCT) and an image sensor CCD.
Focusing on the two-dimensional periodicity of the fixed pattern noise, performs two-dimensional frequency conversion on the image, extracts the frequency of the fixed pattern noise, and extracts noise data from the data obtained by frequency-converting the image using the data. It is an object of the present invention to remove fixed pattern noise in a restored image to improve restored image quality, and to remove only noise components without using a frame memory and regardless of long-time exposure.

(Means for Solving the Problems) In order to achieve this object, the present invention relates to an electronic still camera, comprising: conversion data extraction means 202 and 206 for extracting conversion data subjected to two-dimensional frequency conversion in a pre-exposure state;
Storage means 204, 207 for storing the conversion data extracted by the extraction means, gain control means 105, 125, 205 for controlling the gain of the noise component stored in the storage means, from the image data in the main exposure state Noise component removing means 208 for removing a noise component, further restricting the AC component data extracted by the extracting means 202 to a noise band, and averaging the data restricted to the noise band. Band-limited averaging means for storing the converted data in the storage means 204, and storing the DC component data extracted by the extraction means 206 in the storage means 207, and The noise component removing means is constituted by the subtractor 208.

(Example) An example of the present invention will be described with reference to FIG. Exposure for photographing to remove noise of CCD 103 used as an image element in an electronic still camera with an image data compression device using a frequency decomposition method such as discrete cosine transform (DCT) with the configuration shown in the figure Pre-exposure for noise component extraction before
A device 126 that performs processing 111, 112, and 113 using frequency decomposition techniques such as the above, extracts frequency-resolved data, a filter that limits the data to a band for noise removal, and averages it over the previous screen. And a memory device 128 for holding the averaged data. The noise component is held in the memory device 128, and the exposure level of the subject and the AGC in the process circuit 105 during exposure for photographing. And a gain control device 129 for controlling the gain of the noise component data held in the memory device 128 by using the data of (1) and (2) to remove the noise of the CCD by subtracting the noise component data from the image data.

When recording on the memory card 119: Light corresponding to the subject entered through the lens 101 is reduced to an appropriate amount by the aperture 102, received by the image sensor CCD103, and converted into an electric signal. The shutter speed of the aperture 102 and the CD 103 is controlled by the output of the automatic exposure mechanism 125 in response to a signal from the exposure sensor 126, and the state of proper exposure is determined. This CC
The reset noise included in the output of D103 is removed by a correlated double sampling (CDS: Correlated Double Sampling) circuit 104. In this circuit, the signal level is taken out by a sample-and-hold circuit after the feedthrough level of the CCD output signal is clamped to a constant potential every pixel period. According to this operation, the potential difference between the feedthrough level and the signal level is extracted as an effective signal voltage, and reset noise superimposed on both levels is almost eliminated. The reset noise is thermal noise generated when the floating diffusion layer is charged to a constant potential by the reset transistor. This is because the floating diffusion layer is set to a different potential for each reset due to a noise voltage at the timing when the reset transistor is turned off. This noise is superimposed on both the feedthrough level and the signal level by the same amount every reset.

The output signal of the correlated double sampling circuit is input to the camera process circuit 105, and the luminance signal Y and the color difference signals RY, B-
Y, and is output. The luminance signal Y is converted into digital data by the A / D converter 106 at a specified sampling frequency. The recording / playback switch 121 is a recording image (a)
Is connected. Since the discrete cosine transform (DCT), which is a compression process, is processed as image data in units of two-dimensional blocks such as 8 pixels × 8 pixels, it is necessary to convert a one-dimensional image into blocks. In this conversion control, two sets of eight line buffers 108 and 109 are provided, and alternately switched by switches 107 and 110 to read serial data in units of write blocks.

Next, two-dimensional discrete cosine transform is performed on a block of 8 pixels × 8 pixels or the like. First, a one-dimensional discrete cosine transform circuit 11
According to 1, one-dimensional discrete cosine transform is executed, and the output data is temporarily stored in a random memory (RAM) 112. Next, the data stored earlier is read, and the one-dimensional discrete cosine transform circuit 113 executes the one-dimensional discrete cosine transform to perform the two-dimensional discrete cosine transform as a whole. Next, so-called zigzag scanning is performed by a random memory (RAM) 115 in order to arrange two-dimensional data into one-dimensional data.

Next, the output of the zigzag-scanned random memory (RAM) 115 is quantized by the quantization circuit 114 with reference to a quantization table 120 having quantization data defined by the standard. It is quantized by dividing the output by the quantized data. Although the quantized output is Huffman-coded, the DC component of the discrete cosine transform is subjected to differential pulse code modulation, and the AC component of the discrete cosine transform is subjected to zigzag scanning.
According to 116, it is coded by a Huffman coding circuit 117. This output data (serial data) is stored in a temporary buffer memory (RAM) 118 and written to a memory card 119. The recording ends here.

When reproducing data from the memory card as a video signal: The recording / reproducing switch 121 is switched to the reproducing side. Memory card 1
The Huffman encoding circuit 117 stores the 19 data in the buffer RAM 118 and performs the Huffman encoding.
Is subjected to Huffman decoding (demodulation) in accordance with a look-up table 116, and the demodulated output is inversely quantized by multiplying the data of a
Stored in AM115. The one-dimensional data is rearranged into two-dimensional data by the RAM 115, and the one-dimensional discrete cosine transform (DC
T) Inverse discrete cosine transform (D
CT) to restore image data. Next, the image data of 8 pixels × 8 pixels is converted into data of one scanning line by the line buffers 108 and 109 of the line, output as an analog signal by the D / A converter 122, and encoded into an NTSC composite video signal by the encoder 123. Video output terminal
Output from 124.

Next, the main part of the present invention in the above-described camera will be described. One-dimensional discrete cosine transform circuit 113 and quantization circuit 114
And a noise elimination changeover switch 201 is additionally provided. The function of the changeover switch 201 is such that the one-dimensional discrete cosine transform circuit 223 and the quantization circuit 114 are connected via a subtracter 208 connected between the e and f terminals during recording, and the terminals c and d are connected during reproduction. The connection operates such that the one-dimensional discrete cosine transform circuit 113 and the quantization circuit 114 are directly connected. Output signals of the camera process circuit 105 and the automatic exposure mechanism are input to a subtracter 208 via a gain control circuit 205. The subtracter 208 has a function of removing a noise component.

An AC component extraction circuit 202 for extracting an AC component after performing a two-dimensional discrete cosine transform as a whole is connected to an output side of the one-dimensional discrete cosine transform circuit 113 at the time of recording and to an input side of the noise elimination switch 201. Then, the output is connected to the averaging filter circuit 203. The output of the averaging filter circuit 203 is connected to a memory 204 for storing the output, and is connected to a gain control circuit 205 via this memory.

The operation of the digital still video camera configured as described above to remove noise will be described with reference to the block diagram of FIG. 1 and the flowchart of FIG.

 Pre-exposure is performed prior to main exposure.

First, the camera is set to the recording state. The state of the recording / reproducing switch 121 is set to the a-side connection, and the state of the changeover switch 201 is set to the e-side and f-side connection.

Next, the aperture 102 is completely closed (step 1), and the first shutter is released (step 2). The CCD 103 is exposed in this state without incident light (step 3).

The luminance signal of the output signal of the CCD 103 is processed through the above-described path, and is subjected to a two-dimensional discrete cosine transform as a whole (step 4). The data appears at the e terminal of the changeover switch 201. The AC component of the data appearing at the e terminal is extracted by the AC component extraction circuit 202, and only the AC component effective for noise removal is filtered from the output data by the band limiting filter of the averaging filter circuit 203 (step 5). The noise-removal effective AC component is averaged by the averaging circuit of the same circuit 203 in the block in which the discrete cosine exchange has been executed, and the average data is stored in the memory 204.
(Step 6).

For the noise component belonging to the low range that cannot be extracted by the AC component of the discrete cosine transform (DCT), the DC component of the two-dimensional discrete cosine transform data appearing at the e terminal of the changeover switch 201 is extracted by the DC component extraction circuit 206, and the extracted noise data is extracted. Is stored in the memory 207. In this case, as the DC component as noise data, all DC components at this stage of pre-exposure are stored and prepared for subtraction data in main exposure.

Next, main exposure is performed. The aperture 102 of the light passing through the lens 101 is determined by the control of the automatic exposure mechanism 125 responding to the output signal of the exposure sensor 126 (step 7), and the CCD 10
The shutter speed of 3 is determined, and the main exposure is executed (step 8). Subsequent signal processing is performed in the same manner as in the case of the pre-exposure, and the output data of the two-dimensional discrete cosine transform (step 9) appears at the e terminal of the switching switch 201. The data appearing at the e terminal contains a noise component to be removed. Noise components to be removed are stored in memory
The level of the noise component is read out from the 204 and converted by the gain control circuit 205, and the converted output is supplied to the subtractor 205. Here, noise components are removed (subtracted) from the output data appearing at the e terminal (step 10). Since the fixed pattern noise level appears differently depending on the brightness of the subject, the level gain control circuit 20 of the noise component to be removed is used.
The gain of 5 is converted by controlling with a signal from the automatic exposure mechanism 125 and the camera process circuit 105.

The data from which the desired noise has been removed becomes the input signal of the quantization circuit 114 (the signal appearing at the terminal of the noise changeover switch 201f), is compressed (step 11), and is recorded on the recording medium 119 (step 12).

Noise represented by a DC component can be removed by the subtracter 208 in the same manner as described above by reading the component from the memory 207 and applying the component to the gain control circuit 205.

When the data recorded on the recording medium 119 is reproduced, the state of the changeover switch 201 is changed from the connection state with the subtracter 208 to the cutoff state, directly connected to both the c and d terminals, and the recording / reproduction changeover switch 121 is Connected to terminal b. In this state, a video signal free of fixed pattern noise is obtained from the video output terminal 124 by performing the reverse process starting from the recording medium 119 as described above.

FIG. 4 shows a state of an image from which noise is removed.
(A) and (d) show the noise of the CCD in the pre-exposure state without incident light. (B) and (e) are output images captured in the main exposure state. Images (c) and (f) are obtained by removing the noise of the present invention from the output images (b) and (e). (A) and (d) are images from which noise components have been removed.

Next, another embodiment will be described with reference to FIG. The components common to the embodiment of FIG. 1 are shown with the same figure numbers.
In another embodiment, in addition to the configuration of the embodiment of FIG. 1, two-dimensional discrete cosine-transformed DC component is used for extracting CCD fixed pattern noise that cannot be extracted from the AC component of discrete cosine transformed data. And a memory 207 for storing the extracted data, and the extracted data is output to the subtracter 208 as a noise component.

(Effect of the Invention) It is possible to remove the adverse effect on the image caused by the fixed pattern noise of the CCD at the time of photographing by the camera, and to improve the image quality.

[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 another embodiment of the present invention, FIG. 3 is a flowchart showing a processing procedure of the present invention, and FIG. It is a state diagram of the screen which shows the state of FIG. 201 ... changeover switch, 202 ... AC component extraction circuit, 203
...... Averaging filter circuit, 204, 207 ... Memory, 205 ...
... gain control circuit, 206 ... DC component extraction circuit, 208 ... subtractor.

──────────────────────────────────────────────────の Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) H04N 5/14-5/217 H04N 5/30-5/335

Claims (4)

(57) [Claims] A conversion data extracting means for extracting conversion data subjected to two-dimensional frequency conversion in a pre-exposure state;
(206), storage means (204) and (207) for storing the conversion data extracted by the extraction means, and gain control means (105) for controlling the gain of the noise component stored in the storage means. 125) and (205) and noise component removing means (20) for removing noise components from image data in the main exposure state.
8) An electronic still camera comprising: 2. The data of the AC component extracted by the extracting means (202) is limited to a noise band, and the data obtained by averaging the data limited to the noise band is converted to the data (20).
2. The electronic still camera according to claim 1, further comprising a band-limited average storing means stored in 4). 3. The electronic still camera according to claim 2, wherein the DC component data extracted by said extraction means (206) is stored in said storage means (207). 4. An electronic still camera according to claim 2, wherein said noise component removing means is constituted by a subtractor (208).