JP6581469B2 - Image encoding apparatus, control method therefor, program, and storage medium - Google Patents

Description

  The present invention relates to an image data encoding technique.

  One of the image data encoding methods is JPEG-LS. This JPEG-LS can select lossless encoding or near lossless encoding. Near-lossless encoding is a method in which a difference between a pixel value before encoding and a pixel value obtained by decoding is within a predetermined range. JPEG-LS near-lossless coding switches between predictive coding and run-length coding based on the state of surrounding pixels of the pixel to be coded. In predictive coding, the pixel value of a pixel to be coded is predicted from surrounding pixels, and the prediction error is quantized and coded. Run-length encoding encodes the run length (number of consecutive values) of pixel values within a predetermined range. As another method related to quantization, in the case of a specific pixel value, a quantization method for assigning a specific quantization value is disclosed (for example, Patent Document 1).

Japanese Patent No. 4895567

  Consider a case where image data is near lossless coded by predictive coding. In this case, the image data obtained from the encoded data by the near lossless encoding tends to generate many pixels having a specific value defined by the set allowable error, and this appears as a pseudo contour. There is.

  The present invention has been made in view of such a point, and an object of the present invention is to provide a technique capable of suppressing generation of a pseudo contour more than ever while performing near-lossless coding.

In order to solve this problem, for example, an image encoding device of the present invention has the following configuration. That is,
An image encoding device that performs near-lossless encoding of each pixel of image data within a set allowable error range,
Storage means for storing surrounding pixels used for calculation of a prediction value when predictively encoding a pixel of interest in image data to be encoded;
Calculation means for calculating the prediction error of the focus from the pixel of interest and the predicted value of the pixel of interest obtained with reference to the storage means;
Quantizing means for obtaining a quantized value by quantizing the prediction error obtained by the calculating means using an index value representing a preset allowable error;
Encoding means for encoding the quantization value obtained by the quantization means and outputting the encoded value as the encoded data of the pixel of interest;
The quantization value obtained by the quantization means is inversely quantized using the index value, and the decoded pixel value at the position of the pixel of interest is obtained from the inverse quantization value obtained by the inverse quantization and the predicted value. And a local decoding means for storing the decoded pixel value in the storage means so that the decoded pixel value is one of the surrounding pixels when encoding subsequent to the pixel of interest.
The local decoding means includes
A dequantized value having a difference of 1 in value is generated depending on whether the sign of the quantized value is positive or negative.

  According to the present invention, it is possible to suppress the generation of a pseudo contour more than before while performing predictive coding by near lossless coding.

The figure which shows the structure of an image coding apparatus. The figure which shows the structure of an encoding part. The figure which shows the hardware constitutions of an information processing part. The figure which shows an example of a structure of the imaging part of a camera. The flowchart of the process performed in an encoding part. The flowchart of the process performed in an inverse quantization part. The figure which shows the positional relationship of a focused pixel and a surrounding pixel. The figure explaining Golomb-Rice encoding. The figure explaining inverse quantization. The figure which shows an example of a structure of an image decoding apparatus. The figure which shows an example of a structure of a decoding part. The flowchart of the process performed in a decoding part. The figure explaining inverse quantization.

  Hereinafter, embodiments according to the present invention will be described in detail with reference to the accompanying drawings. In addition, embodiment described below shows an example at the time of implementing this invention concretely, and is one of the concrete embodiments of the structure described in the claim.

[First Embodiment]
The first embodiment is applied to an image coding apparatus that performs near-lossless coding of image data by predictive coding. Hereinafter, the image coding apparatus shown in the present embodiment will be described. In the first embodiment, a case where the image encoding device is applied to a camera will be described.

  FIG. 1 is a block diagram of the camera 100. The camera 100 includes an operation unit 101, an imaging unit 102, an information processing unit 109, a storage unit 107, and an I / O interface 108. The information processing unit 109 includes an acquisition unit 103, an image processing unit 104, an encoding unit 105, and an output unit 106.

  FIG. 4A is a front view of the camera 100. As illustrated in FIG. 4A, the camera 100 includes an imaging unit 102. FIG. 4B is a diagram illustrating an internal configuration of the imaging unit 102. The imaging unit 102 includes a zoom lens 401, focus lenses 402 and 403, an aperture stop 404, and a shutter 405. The optical low-pass filter 406, iR cut filter 407, color filter 408, image sensor 409, and A / D conversion unit 410 are included. The user can adjust the amount of incident light entering the imaging unit 102 by adjusting the diaphragm 404. The image sensor 409 is a light receiving element such as a CMOS or a CCD. When the light amount of the subject is detected by the image sensor 409, the detected light amount is converted into a digital value by the A / D conversion unit 410 and output to the information processing unit 109 as digital data.

  FIG. 3 is a diagram illustrating an internal configuration of the information processing unit 109. The information processing unit 109 includes a CPU 301, a RAM 302, and a ROM 303, and each component is connected to each other via a system bus 304.

  The ROM 303 stores programs executed by the CPU 301 and various data. The RAM 302 is used as a work area and various buffers when the CPU 301 executes a program. The CPU 301 executes a program stored in the ROM 303 and uses the RAM 302 as a work area or a buffer. As a result, the CPU 301 functions as the acquisition unit 103, the image processing unit 104, the encoding unit 105, and the output unit 106 shown in FIG. 1 and controls the entire apparatus. Here, each component is realized by being executed by the CPU 301, but may be realized by a dedicated circuit or the like that performs equivalent processing.

  The operation unit 101 is an input device such as a button, dial, or touch panel provided in the camera body, and can be operated by the user to give instructions such as start / stop of shooting and setting of shooting conditions. A parameter NEAR, which will be described later, is also set from the operation unit 101 in accordance with a user instruction. Here, the parameter NEAR is an index value representing an allowable error when performing near-lossless encoding. When the user indicates the size of the allowable error, 1, 2, 4, 8,. One of the power numbers is assumed to be set as NEAR.

  The storage unit 107 is a non-volatile storage medium such as a memory card that can store the RAW image data acquired by the imaging unit 102 and the image data. The I / O interface 108 can use a serial bus connection implemented by a universal serial bus (USB), and has a corresponding USB connector. Of course, the I / O interface 108 may be used for LAN connection by optical fiber or wireless connection. A display unit 305 displays captured images and characters. A liquid crystal display is generally used for the display unit 305. Further, a touch panel function may be provided, and in that case, a user instruction using the touch panel can be handled as an input of the operation unit 101.

  Hereinafter, processing of the information processing unit 109 illustrated in the present embodiment will be described. Based on the imaging instruction to the user operation unit 101, the acquisition unit 103 acquires the RAW image data of the Bayer array output from the imaging unit 102 and supplies the RAW image data to the image processing unit 104. The image processing unit 104 performs demosaic processing on the RAW image data, generates image data in which one normal pixel is composed of three components of RGB, and supplies the image data to the encoding unit 105. The encoding unit 105 performs an encoding process on the supplied image data within a set allowable error range.

  Note that the pixel of the image data to be encoded in the present embodiment is composed of three RGB components as described above, and each component is represented by 8 bits each. The encoding unit 105 in the embodiment will be described by taking an example of encoding as a monochrome image in units of color components. That is, the R component monochrome multi-valued image, the G component monochrome multi-valued image, and the B component monochrome multi-valued image are encoded independently.

  Therefore, it is obvious that the image data to be encoded in the present embodiment may be a monochrome image encoding of only a single component. The image data to be encoded is not limited to the RGB format, and may be another color space or RAW image data. For example, when one pixel is composed of Y, Cb, and Cr and each is composed of 8 bits, each component may be encoded as a monochrome image. In addition, when applied to RAW image data, the imaging elements of the imaging unit 102 have a Bayer array (an array of 2 × 2 imaging elements corresponding to the colors R, G0, G1, and B). When the signal from each image sensor is expressed by 14 bits, a monochrome multivalued image composed of R components, a monochrome multivalued image composed of G0 components, a monochrome multivalued image composed of G1 components, B Each monochrome multi-valued image composed of components is encoded. Therefore, the present application is not limited by the type of color space of the image, the type of color components constituting the image, and the number of bits. It should be understood that the encoding target image has three components of R, G, and B, each having 8 bits (256 gradations), for the sake of easy understanding.

  Processing of the encoding unit 105 in the present embodiment will be described. FIG. 2 shows an internal configuration of the encoding unit 105. FIG. 5 shows encoding of monochrome multilevel image data of one color component (in the embodiment, any one of R, G, and B) performed by the encoding unit 105. It is a flowchart which shows a process. That is, the process of FIG. 5 is actually executed three times. Then, the code data obtained as a result of the three times is collected to form one file.

  First, in step S <b> 501, the color component input unit 201 inputs image data of one component in the image data supplied from the image processing unit 104 as a monochrome multilevel image.

  In step S502, the encoding unit 105 determines whether all the pixels in the monochrome multilevel image have been encoded. If all the pixels are encoded, the encoding process is terminated, and if there is a pixel that is not encoded, the process proceeds to step S503.

  In step S503, the encoding unit 105 inputs the pixel value of the monochrome multivalued image in units of lines. In step S504, the encoding unit 105 sets a pixel located at the left end of the input line as a target pixel to be encoded.

  The line data holding unit 202 holds data necessary for the encoding process in a buffer area secured in advance in the RAM 302. It is assumed that the buffer area holds data for two lines, the line where the pixel of interest to be encoded is located and the immediately preceding line. FIG. 7 shows the positional relationship between the pixel of interest and its surrounding pixels. The x shown in the figure is the target pixel to be encoded. As shown in the figure, the reference range is the surrounding pixels (left pixel a, upper pixel b, upper left pixel c) of the target pixel in two lines, the line where the target pixel x is located and the line immediately preceding it. Further, in the embodiment, since the pixels are sequentially encoded in the raster scan order, the left pixel a, the upper pixel b, and the upper left pixel c of the target pixel x are already encoded pixels. In near lossless encoding, the difference between the pixel value before encoding and the pixel value after decoding can be set within a set allowable range. For this reason, the decoded pixel value in the buffer area is stored and held (local decoding).

  When the target pixel x is located at the left end of the first line, the surrounding pixels a, b, and c do not exist. When the pixel of interest x is located at the left end of the second and subsequent lines, the surrounding pixels b and c are actual pixels, but the surrounding pixel a does not exist. Such a nonexistent pixel is considered to have a predetermined value (for example, 0). The predetermined value may be common between the encoding device and the decoding device, and is not limited to “0”.

  In step S505, the encoding unit 105 determines whether all the pixels in the target line have been encoded. In other words, it is determined whether or not the encoding of the pixel at the line end (right end) of the line of interest has been completed. If it is determined that all pixels in the line have been encoded, the process proceeds to the next line in step S506. If encoding of all the pixels of the line of interest is incomplete, the encoding unit 105 proceeds with the process to step S507.

In step S507, the prediction conversion unit 203 calculates a prediction value p for the target pixel, and derives a prediction error e (= x−p) that is a difference between the target pixel x and the prediction value p. In the embodiment, the predicted value p is calculated by using MED (Media Edge Detection) prediction. The calculation formula of the predicted value p using the pixel values of the peripheral pixels a, b, and c in FIG. 7 is as follows.

  The prediction conversion unit 203 derives a prediction error from the difference between the predicted value p obtained as described above and the target pixel x.

In step S508, the quantization unit 204 quantizes the prediction error e to calculate a quantization value q. The quantization unit 204 performs quantization based on the parameter “NEAR” in near lossless coding. The parameter NEAR specifies the maximum value of the change in the pixel value accompanying encoding / decoding, and is referred to as an allowable error here. The calculation formula of the quantization value q using the prediction error e (= target pixel value x−prediction value p) and the parameter NEAR is as follows.

  When the parameter NEAR is limited to 1, 2, 4, 8,..., The value to be divided (2 × NEAR) is a power of 2, 4, 8, 16, and 2. That is, a simple bit shift operation is possible, and the operation cost can be reduced. Of course, even if the parameter NEAR is another numerical value, if the value to be divided (2 × NEAR) is a power of 2, an operation by bit shift is possible.

In step S509, the Golomb encoding unit 206 performs Golomb-Rice encoding on the quantized value q. The Golomb encoding unit 207 first converts the quantized value q into a non-negative integer value MV. The conversion formula is as follows. As is apparent from this equation, the sign of the quantized value q can be determined depending on whether the integer value MV is even or odd.

Next, the Golomb encoding unit 206 performs entropy encoding on the integer value MV by Golomb-Rice encoding using the parameter k. The procedure for Golomb-Rice coding is as follows.
(1) The integer value MV is expressed in binary, and the number 0 indicated by a value obtained by shifting the integer value MV to the right by k bits is arranged, and then 1 is added.
(2) After the step (1), the lower k bits of MV are taken out and added.

  FIG. 8 shows the relationship between the Golomb-Rice coding parameter k, the non-negative integer value MV, and the code word. The configuration of Golomb-Rice coding is not limited to this. For example, codes may be configured by reversing 0 and 1, and the order of (1) and (2) described in the above procedure may be changed. The codes may be configured by replacing them. Here, the method for determining the encoding parameter k is not particularly specified, but it is sufficient that the same parameter can be used on the encoding side and the decoding side. For example, a method of selecting a parameter k that seems to be appropriate for each color component in advance may be used, or it may be dynamically changed during the encoding process.

  In FIG. 8, for example, consider the case where the parameter k = 0. When the quantized value q is −2, the non-negative integer value MV is 3. When the quantized value q is +2, the non-negative integer value MV is 4. When the non-negative integer value MV is 3, the code word is “0001” and the code length is 4 bits. When the non-negative integer value MV is 4, the code word is “00001” and the code length is 5 bits. As described above, when the Golomb encoding unit 206 finishes outputting the code word for the pixel of interest x, the process proceeds to step S510.

  In step S510, the inverse quantization unit 205 performs local decoding processing including inverse quantization processing for the quantized value q. A flowchart of the processing performed in the inverse quantization unit 205 is shown in FIG.

  First, in step S601, the determination unit 208 acquires a quantized value q. In step S602, the determination unit 208 switches the inverse quantization method based on the sign of the quantization value. That is, when the quantized value q is 0 or more, the process proceeds to step S603 and the first inverse quantization unit 209 performs processing. When the quantized value q is smaller than 0, the process proceeds to step S604, where the second inverse quantization unit 210 performs the process.

In step S603, the first inverse quantization unit 209 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq. The calculation formula is as follows.
iq = q × 2 × NEAR
Here, when the parameter NEAR is limited to 1, 2, 4, and 8, the value to be multiplied (2 × NEAR) is a power of 2, 4, 8, 16, and 2. That is, iq can be obtained by a simple bit shift calculation, and the calculation cost can be reduced. Of course, even if the parameter NEAR is another numerical value, if the value to be multiplied (2 × NEAR) is a power of 2, an operation by bit shift is possible.

In step S604 and step S605, the second inverse quantization unit 210 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq. The calculation formula is as follows.
iq = q × 2 × NEAR + 1
First, in step S604, as in step S603, the quantized value q is multiplied by (2 × NEAR). In step S605, a level shift is performed so that the inverse quantized value iq falls within the allowable error. Here, “+1” is added. The level shift value is determined based on the quantization method in step S508 and the allowable error.

  The inverse quantization process will be described with reference to FIG. FIG. 9A shows an example in which the level shift in step 605 is not performed in the inverse quantization process. FIG. 9B shows an example in which level shift is performed by inverse quantization processing. Here, a case where the parameter NEAR is 2 will be described. Each field shown in the figure indicates a prediction error, a quantization value, and an inverse quantization value when the prediction error is quantized and inversely quantized. It can be seen that the difference between the prediction error and the dequantized value is all within the range of the parameter NEAR (= 2) in both FIGS. 9A and 9B.

  It can be seen that by performing quantization / inverse quantization processing by dividing / multiplying by a power of 2, a plurality of inverse quantization values can be selected in the inverse quantization. For example, when a prediction error is quantized from −2 to −5 to a quantized value−1, the dequantized values after dequantization are both −3 and −4 in the parameter NEAR (= 2). Within the range. In the present application, a plurality of inverse quantization methods are switched based on the sign of the quantized value using such a property. The effect of this process will be described later.

In step S606, the inverse quantization unit 205 calculates a decoded pixel value x ′ of the pixel of interest x using the inverse quantization value iq and the predicted value p. The calculation formula of the pixel value x ′ is as follows.
x ′ = iq + p
In near-lossless encoding, the absolute value of the difference between the pixel value x before encoding and the pixel value x ′ after decoding is equal to or less than the parameter NEAR (less than the allowable value).
| X−x ′ | ≦ NEAR
Returning to FIG. 5, in step S511, the encoding unit 105 updates the pixel value x of the pixel of interest. Specifically, the value of the pixel of interest x before encoding in the buffer area of the line data holding unit 202 is replaced with the pixel value x ′ after decoding. In step S512, the encoding unit 105 sets the pixel on the right side of the target pixel x as a new target pixel, and returns to step S505.

  Returning to FIG. 2, the code output unit 207 supplies the encoded data generated by the Golomb encoding unit 206 to the output unit 106. The output unit 106 generates a header including information necessary for decoding, and outputs the encoded data supplied from the code output unit 207 to the storage unit 107 following the generated header and stores it. The stored encoded data is output to a device outside the camera via the I / O interface 108.

  Next, the effect by the processing of the first inverse quantization unit 209 and the second inverse quantization unit 210 in the embodiment will be described. As described above, when the parameter NEAR is limited to 1, 2, 4, and 8, the multiplication process can be realized with a simple bit shift, and the calculation cost can be reduced. Further, the improvement of the image quality of the decoded image will be described with reference to FIG. Details of the decoding process will be described in the second embodiment. Here, a case where the parameter NEAR is 2 is taken as an example. FIG. 9A shows a case where the level shift of step 605 is not performed in the inverse quantization process. It can be seen that the inverse quantization value is a multiple of 4.

The decoded pixel value x ′ is calculated as follows:
Decoded pixel value x ′ = inverse quantized value iq + predicted value p
Since the inverse quantized value is a multiple of 4, if the predicted value p is a multiple of 4, the decoded pixel value x ′ is a multiple of 4. The predicted value p is calculated from the surrounding pixels by MED prediction. If MED prediction is shown again, it is as follows.

  If the surrounding pixels a, b, and c are multiples of 4, the predicted value p is a multiple of 4, and the decoded pixel value x ′ is a multiple of 4. Furthermore, since the decoded pixel value x ′ is used as a surrounding pixel, a large multiple of 4 is generated in the decoded pixel value, and a pseudo contour is generated in the decoded image. Although the case where the parameter NEAR is 2 is taken as an example here, the same applies to the case where the parameter NEAR is 1, 4, and 8.

  In order to reduce the pseudo contour of the decoded image, as shown in FIG. 9B, level shift is performed by inverse quantization processing. When the quantized value is 0 or more, the inverse quantized value is a multiple of 4. However, when the quantized value is smaller than 0, the inverse quantized value is not a multiple of 4. By performing such inverse quantization processing, the pseudo contour of the decoded image can be reduced and the image quality can be improved.

  Here, an example in which the prediction value p is calculated by MED prediction is shown, but other prediction methods may be used as long as the prediction is calculated only by addition and subtraction from surrounding pixels. For example, the case where the left adjacent pixel a is used as a predicted value or the case where the upper adjacent pixel b is used as a predicted value can be used.

  As described above, according to the present embodiment, when image data is subjected to near-lossless encoding, the amount of computation in predictive encoding is reduced by switching between a plurality of inverse quantization methods based on positive and negative codes of quantization values. And the image quality of the decoded image can be improved.

  Next, a decoding apparatus that decodes encoded data generated by the camera 100 of the first embodiment will be described. FIG. 10 is a diagram illustrating an example of a configuration of a device 1000 that performs image decoding according to the first embodiment. The device 1000 includes an I / O interface 1001, a storage unit 1002, an operation unit 1003, an information processing unit 1008, and a display unit 1007.

  The type of device 1000 is not particularly limited, but is typically a computer or a camera. The device 1000 in the first embodiment will be described as a computer. Similar to the I / O interface 108, the I / O interface 1001 has a USB connector. The device 1000 acquires compressed image data output from the I / O interface 108 of the camera 100 via the I / O interface 1001.

  The recording unit 1002 is a nonvolatile storage medium such as a memory card that can store compressed image data acquired from the I / O interface 1001. The operation unit 1003 is an input device such as a keyboard and a mouse provided in the computer, and can be operated by a user to issue an instruction such as a decoding process. The display unit 1007 can display a decoded image, and a liquid crystal display is generally used. The internal configuration of the information processing unit 1008 is the same as that of the information processing unit 109 shown in FIG.

  Hereinafter, processing of the information processing unit 1008 in the first embodiment will be described. Based on the decoding instruction to the user operation unit 1003, the acquisition unit 1004 acquires compressed image data (data file encoded by the apparatus of the first embodiment) from the storage unit 1002, and stores the compressed image data in the decoding unit 1005. Supply.

  The decoding unit 1005 performs a decoding process on the compressed image data. In the first embodiment, an example will be described in which one pixel is represented by 8 bits for each of RGB and is decoded as a monochrome image in units of color components. That is, the R component monochrome multivalued image, the G component monochrome multivalued image, and the B component monochrome multivalued image are decoded independently.

  Processing of the decoding unit 1005 shown in the first embodiment will be described. FIG. 11 is an internal configuration of the decoding unit 1005, and FIG. 12 is a flowchart showing a decoding process for code data of one color component (in the embodiment, any one of R, G, and B) performed by the decoding unit 1005. . That is, the process of FIG. 12 is actually executed three times. Then, color image data is generated by combining the monochrome multi-valued images of the R component, G component, and B component obtained as a result of performing three times. The line data holding unit 1104 has a capacity for two lines and is cleared to 0 prior to decoding.

  First, in step S <b> 1201, the code input unit 1101 inputs code data of one component in the compressed image data output from the acquisition unit 1004. In step S1202, the decoding unit 1005 determines whether all the pixels in the monochrome multi-valued image have been decoded. If all the pixels are decoded, the decoding process is terminated. If there is a pixel that has not been decoded, the process proceeds to step S1203. In step S1203, the decoding unit 1005 sets a pixel located at the left end of the line to be decoded as a target pixel to be decoded.

  The line data holding unit 1104 holds data necessary for the decoding process. Here, a buffer memory is prepared for two lines of the line where the target pixel to be decoded is located and the line immediately before the line, and the decoded pixel value is held. FIG. 7 shows the positional relationship between the pixel of interest and its surrounding pixels. x is a target pixel to be decoded. As shown in the figure, the reference range is the surrounding pixels (left pixel a, upper pixel b, upper left pixel c) of the target pixel in two lines, the line where the target pixel x is located and the line immediately preceding it. Since the pixels are sequentially decoded in the raster scan order, the left pixel a, the upper pixel b, and the upper left pixel c are decoded pixels. Note that when the pixel of interest x is located at the left end of the first line, a, b, and c do not exist. In addition, when the pixel of interest x is located at the left end of the line after the second line, a does not exist. Such a nonexistent pixel is considered to have a predetermined value (for example, 0). This is why the line data holding unit 1104 is cleared to zero prior to encoding.

  In step S1204, the decoding unit 1005 determines whether or not decoding of all the pixels in the line of interest has been completed. In other words, it is determined whether or not the decoding of the pixel at the line end (right end) of the line of interest has been completed. If it is determined that all the pixels in the line of interest have been decoded, the process proceeds to the next line in step S1205. When decoding of all the pixels of the line of interest is incomplete, the decoding unit 1005 advances the process to step S1206.

In step S1206, the prediction conversion unit 1102 calculates a prediction value p using MED (Media Edge Detection) prediction for prediction conversion. The formula for calculating the predicted value p using the decoded pixel values of the peripheral pixels a, b, and c in FIG. 7 has already been shown.

In step S1207, Golomb decoding section 1103 performs entropy decoding on quantized value q based on Golomb-Rice decoding. The procedure for Golomb-Rice decoding using the parameter k is as follows.
(1) One bit is extracted from the code data, and the count is incremented as long as the extracted bit is “0”. When the extracted bit is “1”, the process proceeds to the next step (2). Here, the number whose bits are “0” is referred to as “ZEROcnt”.
(2) Shift Zerocnt to the left by k bits. Further, a value obtained by extracting k bits from the code data is added to the shifted ZEROcnt to obtain a symbol MV.

Next, Golomb decoding section 1102 calculates quantized value q from symbol MV. The conversion formula is as follows.

In step S1208, the inverse quantization unit 205 performs inverse quantization on the quantized value q. A process performed by the inverse quantization unit 205 is shown in a flowchart in FIG. Since the processing of the inverse quantization unit 205 is the same as the processing in the encoding unit, description thereof is omitted here. By the process of the inverse quantization unit 205, a quantized value iq is calculated, and a pixel value x ′ after decoding of the pixel of interest x is calculated using the predicted value p.
x ′ = iq + p
In step S1209, the decoding unit 1005 sets a decoded pixel value x ′ of the pixel of interest x. That is, the decoding unit 1005 stores the decoded pixel value x ′ at the position of the pixel of interest x in the buffer memory for two lines held by the line data holding unit 1104. In step S1210, the decoding unit 1005 sets the pixel immediately adjacent to the target pixel x as a new target pixel, and returns to step S1204.

  Returning to FIG. 11, the color component output unit 1105 outputs the decoded image data to the output unit 1006. Returning to FIG. 10, the output unit 1006 converts the image data of R, G, and B components into a preset format, and outputs the data to the storage unit 1002 for storage. Or it outputs to the display part 1007 and displays.

  As described above, in the decoding apparatus, when performing near-lossless decoding of image data, a plurality of inverse quantization methods are switched based on the sign of the quantization value, thereby reducing the amount of computation in predictive encoding and decoding. Expected to improve image quality.

[Second Embodiment]
The first embodiment has described an example in which image data is near-lossless encoded / decoded by predictive encoding. In the second embodiment, a modification of the quantization and inverse quantization methods of the image encoding device will be described.

  Since the configuration of the image encoding apparatus in the second embodiment is the same as that of FIG. 1 of the first embodiment and the configuration of the encoding unit 105 is also the same as that of FIG. 2, description thereof will be omitted. Therefore, the quantization unit 204 and the inverse quantization unit 205 different from the first embodiment will be described together with the processing of the encoding unit 105.

  FIG. 5 is a flowchart showing an encoding process for image data of one color component (in the embodiment, any one of R, G, and B) performed by the encoding unit 105.

Steps S501 to S507 are the same as those in the first embodiment, and a description thereof will be omitted here. In step S508, the quantization unit 204 quantizes the prediction error e to calculate a quantization value q. This quantization processing is performed based on the parameter NEAR in the near lossless encoding. The calculation formula of the quantization value q using the prediction error e (= the pixel value of the target pixel x−the prediction value p) and the parameter NEAR in the second embodiment is as follows.

  Since step S509 is common to the first embodiment, description thereof is omitted here. In step S510, the inverse quantization unit 205 performs inverse quantization on the quantized value q. The processing performed in the inverse quantization unit 205 is as shown in the flowchart of FIG.

  First, in step S601, the determination unit 208 acquires a quantized value q. In step S602, the determination unit 208 switches the inverse quantization method based on the sign of the quantization value. That is, when the quantized value q is 0 or more, the process proceeds to step S603 and the first inverse quantization unit 209 performs processing. When the quantized value q is smaller than 0, the process proceeds to step S604, where the second inverse quantization unit 210 performs the process.

In step S603, the first inverse quantization unit 209 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq.
iq = q × 2 × NEAR
Here, when the parameter NEAR is limited to 1, 2, 4, and 8, the value to be multiplied (2 × NEAR) is a power of 2, 4, 8, 16, and 2. That is, the inverse quantization value iq can be calculated by a simple bit shift calculation, and the calculation cost can be reduced. Of course, even if the parameter NEAR is another numerical value, if the value to be multiplied (2 × NEAR) is a power of 2, an operation by bit shift is possible. Even if the parameter NEAR is 3, 5 or the like, and the value to be multiplied is not a power of 2, an operation by bit shift is not possible, but quantization and inverse quantization can be performed.

In step S604 and step S605, the second inverse quantization unit 210 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq. The calculation formula is as follows.
iq = q × 2 × NEAR-1
First, in step S604, as in step S603, the quantized value q is multiplied by (2 × NEAR). In step S605, a level shift is performed so that the inverse quantized value iq falls within the allowable error. Here, 1 is subtracted (-1 is added). The level shift value is determined based on the quantization method in step S508 and the allowable error.

  The inverse quantization process will be described with reference to FIG. FIG. 13 shows an example in which the above quantization and inverse quantization processes are performed. Here, when the parameter NEAR is 2, the prediction error, quantization value, and inverse quantization value when the prediction error is quantized and inversely quantized are shown. When comparing the prediction error and the inverse quantized value, it can be seen that all of FIG. 13 falls within the difference within the range of the parameter NEAR (= 2).

  In the quantization and inverse quantization processing of the second embodiment, the quantization value 0 is set so as to allow an error of +2 and -2, and the coding efficiency near the quantization value 0 is improved. be able to.

  Similarly to the first embodiment, when the quantized value is 0 or more, the inverse quantized value is a multiple of 4, but when the quantized value is smaller than 0, the inverse quantized value is 4 It turns out that it does not become a multiple. By performing such inverse quantization processing, the pseudo contour of the decoded image can be reduced and the image quality can be improved.

  Note that step S511 and step S512 are the same as those in the first embodiment, and a description thereof will be omitted here.

  Next, a decoding process in the decoding device will be described. The configuration of the decoding device is the same as that of FIG. 10 of the first embodiment, and the configuration of the sub rear portion 1005 is also the same as that of FIG. Therefore, hereinafter, the process of the decoding unit 1005 in the decoding apparatus will be described with reference to FIG. 12 and a modification of the quantization and inverse quantization methods in the decoding process.

  Steps S1201 to S1207 are the same as those in the first embodiment, and a description thereof will be omitted here. In step S1208, the inverse quantization unit 205 performs inverse quantization on the quantized value q. A flowchart of the processing performed in the inverse quantization unit 205 is shown in FIG.

  First, in step S601, the determination unit 208 acquires a quantized value q. In step S602, the determination unit 208 switches the inverse quantization method based on the sign of the quantization value. That is, when the quantized value q is 0 or more, the process proceeds to step S603 and the first inverse quantization unit 209 performs processing. When the quantized value q is smaller than 0, the process proceeds to step S604, where the second inverse quantization unit 210 performs the process.

In step S603, the first inverse quantization unit 209 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq. The calculation formula is as follows.
iq = q × 2 × NEAR
Here, when the parameter NEAR is limited to 1, 2, 4, and 8, the value to be multiplied (2 × NEAR) is a power of 2, 4, 8, 16, and 2. That is, the inverse quantization value iq can be calculated by a simple bit shift calculation, and the calculation cost can be reduced. Of course, even if the parameter NEAR is another numerical value, if the value to be multiplied (2 × NEAR) is a power of 2, an operation by bit shift is possible. Also, even when the value to be multiplied is not a power of 2, such as a parameter NEAR of 3 or 5, and is an even number, calculation by bit shift cannot be performed, but quantization and inverse quantization can be performed.

In step S604 and step S605, the second inverse quantization unit 210 performs inverse quantization using the quantization value q and the parameter NEAR, and calculates an inverse quantization value iq. The calculation formula is as follows.
iq = q × 2 × NEAR-1
First, in step S604, as in step S603, the quantized value q is multiplied by (2 × NEAR). In step S605, a level shift is performed so that the inverse quantized value iq falls within the allowable error. Here, 1 is subtracted (-1 is added).

  The inverse quantization process will be described with reference to FIG. FIG. 13 shows an example in which the above quantization and inverse quantization processes are performed. Here, when the parameter NEAR is 2, the prediction error, quantization value, and inverse quantization value when the prediction error is quantized and inversely quantized are shown. When comparing the prediction error and the inverse quantized value, it can be seen that all of FIG. 13 falls within the difference within the range of the parameter NEAR (= 2).

  In the quantization and inverse quantization processing in the decoding apparatus of the second embodiment, the quantization value 0 is set so as to allow an error of +2 and -2, and the coding efficiency near the quantization value 0 is set. Can be improved.

  Similarly to the first embodiment, when the quantized value is 0 or more, the inverse quantized value is a multiple of 4, but when the quantized value is smaller than 0, the inverse quantized value is 4 It turns out that it does not become a multiple. By performing such inverse quantization processing, the pseudo contour of the decoded image can be reduced and the image quality can be improved.

  Steps S1209 and S1210 are common to those in the second embodiment, and a description thereof will be omitted here.

(Other examples)
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

DESCRIPTION OF SYMBOLS 100 ... Camera 101 ... Operation part 102 ... Imaging part 103 ... Acquisition part 104 ... Image processing part 105 ... Encoding part 106 ... Output part 107 ... Storage part 108 ... I / O, 109 ... Information Processing unit 201 ... Color component input unit 202 202 Line data holding unit 203 ... Prediction conversion unit 204 ... Quantization unit 205 ... Inverse quantization unit 206 ... Golomb encoding unit 207 ... Code output unit 208 ... determination unit, 209 ... first inverse quantization unit, 210 ... second inverse quantization unit

Claims (9)

An image encoding device that performs near-lossless encoding of each pixel of image data within a set allowable error range,
Storage means for storing surrounding pixels used for calculation of a prediction value when predictively encoding a pixel of interest in image data to be encoded;
Calculating means for calculating a prediction error of the target pixel from the target pixel and the predicted value of the target pixel obtained by referring to the storage unit;
Quantizing means for obtaining a quantized value by quantizing the prediction error obtained by the calculating means using an index value representing a preset allowable error;
Encoding means for encoding the quantization value obtained by the quantization means and outputting the encoded value as the encoded data of the pixel of interest;
The quantization value obtained by the quantization means is inversely quantized using the index value, and the decoded pixel value at the position of the pixel of interest is obtained from the inverse quantization value obtained by the inverse quantization and the predicted value. And a local decoding means for storing the decoded pixel value in the storage means so that the decoded pixel value is one of the surrounding pixels when encoding subsequent to the pixel of interest.
The local decoding means includes
An image coding apparatus, wherein a dequantized value having a difference of 1 is generated between a positive value and a negative value of the sign of the quantized value. When the pixel of interest to be encoded is x, the predicted value is p, and the index value is NEAR,
The quantization means calculates a quantized value q according to the following equation:

The image coding apparatus according to claim 1. The local decoding means calculates the inverse quantization value iq according to the following equation: iq = q × 2 × NEAR (when q ≧ 0)
iq = q × 2 × NEAR + 1 (when q <0)
The image coding apparatus according to claim 2, wherein: When the pixel of interest to be encoded is x, the predicted value is p, and the index value is NEAR,
The quantization means calculates a quantized value q according to the following equation:

The image coding apparatus according to claim 1. The local decoding means calculates the inverse quantization value iq according to the following equation: iq = q × 2 × NEAR (when q ≧ 0)
iq = q × 2 × NEAR-1 (when q <0)
The image coding apparatus according to claim 4, wherein:   The image coding apparatus according to claim 1, further comprising a generation unit configured to generate a power of 2 as the index value in response to a user operation.   A program for causing a computer to function as each unit included in the image encoding device according to any one of claims 1 to 6 when the computer reads and executes the computer.   A computer-readable storage medium storing the program according to claim 7. Storage means for storing surrounding pixels used for calculation of a prediction value when predictive encoding of a pixel of interest in image data to be encoded is performed, and each pixel of the image data is within a set allowable error range A control method for an image encoding device that performs near-lossless encoding in
A calculating step of calculating a prediction error of the target pixel from the target pixel and a predicted value of the target pixel obtained by referring to the storage unit;
A quantization step in which the quantization means obtains a quantized value by quantizing the prediction error obtained in the calculation step using an index value representing a preset allowable error;
An encoding step, wherein the encoding means encodes the quantized value obtained in the quantization step and outputs the encoded value as encoded data of the pixel of interest;
A local decoding unit inversely quantizes the quantization value obtained in the quantization step using the index value, and calculates the pixel of interest from the inverse quantization value obtained by the inverse quantization and the predicted value. A local decoding step of obtaining a decoded pixel value of a position, and storing the decoded pixel value in the storage means so as to be one of the surrounding pixels when encoding subsequent to the pixel of interest;
The local decoding step includes
A method for controlling an image coding apparatus, comprising: generating an inverse quantized value having a difference of 1 in a case where the sign of the quantized value is positive and negative.

Images