JP2009027556A - Image processing circuit - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the compressibility of image data by a simple hardware configuration without deteriorating image quality. <P>SOLUTION: An LCD source driver 2 includes an RGB data receiving circuit 4, an RGB/YCC converting circuit, a BTC coding circuit 6, a coded data memory 7, a BTC decoding circuit 8, a YCC/RGB converting circuit 9, and an LCD driving circuit 10. It is detected whether a change amount of pixels in a pixel block exceeds a threshold or not. When exceeding the threshold, it is determined that a change of an image is extremely large, and 3-level maximum/minimum exception BTC coding processing is performed in which pixels in the pixel block are converted into three representative values and coded. In the case that the change amount is smaller than or equal to the threshold, it is determined that the image is flat, and 2-level differential BTC coding processing is performed in which pixels in the pixel block are converted into two representative values and coded. In 3-level maximum/minimum exception BTC coding processing, the threshold is computed based on an average value of pixel values excepting a maximum value MAX and a minimum value MIN in the pixel block, thereby selecting three representative values meeting actual luminance and color tones. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing circuit that encodes image data.

  An artificial image such as an image generated by a PC tends to feel uncomfortable if artificial pictographic information such as characters and window frames cannot be displayed correctly. In particular, it is important to display character information without a sense of incongruity.

  In Patent Document 1, four levels of representative values in a pixel block are set to maintain image quality. However, with the technique of Patent Document 1, when an image originally has three levels, the image cannot be displayed correctly. In Patent Document 1, when selecting four levels of representative values, the maximum and minimum two representative values are stored, and the remaining two representative values are predicted by interpolation from the maximum and minimum representative values. The saving memory is reduced. For this reason, there is no problem when the prediction by the interpolation is in good agreement with the actual data, but when the image is essentially composed of three levels, an image different from the actual image is generated.

  To cope with such problems, it is important to correctly predict representative values. However, also in Patent Document 2, representative values are selected by interpolating from two representative colors at equal intervals. Here, although interpolation at three levels is included, the point of interpolation at equal intervals is the same.

  Non-Patent Document 1 achieves a high compression effect by using bitmap information in common for RGB and setting the representative value to 4-bit accuracy and 5-bit accuracy in the differential mode. However, there are the following problems.

  In order to share bitmap information in one RGB, it is necessary to optimally select a combination of a quantization table, a representative color, and a modifier, which has a very heavy search burden. In particular, generating the bit-compressed data stored in the built-in RAM of the LCD driver IC for portable devices has a very heavy hardware load.

In addition, since it is impossible to control the brightness and the color separately in the bitmap sharing, it is perceived as a large size artifact (color bounding: color fixed) of 4 × 2 blocks for a sudden change in color. There is a problem of end. Further, in order to realize a high gradation of 8 bits, only a 5-bit accuracy is ensured as a representative value even in the differential mode, which is a serious problem. This is because the accuracy of the quantization table cannot be ensured even in the case of a flat image, and therefore 8-bit accuracy cannot be guaranteed.
US Pat. No. 7,058,218 US Pat. No. 7,043,877 Jacob Stroem, Thomas Akenine-Moeller.iPACKMAN: High-Quality, Low-Complexity texture Compression for Mobile Phones, ACM Graphics Hardware 2005, pp. 63- 70.

  The present invention provides an image processing circuit capable of increasing the compression rate of image data with a simple hardware configuration without degrading the image quality.

  According to one aspect of the present invention, for each pixel block composed of a plurality of adjacent pixels, an activity amount detection that determines whether or not a difference between pixel values of a plurality of pixels in the pixel block exceeds a predetermined threshold value. Means, threshold setting means for setting the predetermined threshold for each of the pixel blocks, and each pixel value in the pixel block determined to have exceeded the predetermined threshold by the activity amount detecting means n ( n is an integer equal to or greater than 2) and is subjected to a first encoding process in which bit compression is performed, and each of the pixel blocks in the pixel block determined to have not exceeded the predetermined threshold by the activity amount detection unit Encoding means for performing a second encoding process for converting the pixel value into the n or less representative values and performing bit compression, and the threshold value setting means is a maximum value of each pixel value in the pixel block. And most Average value calculating means for calculating an average value of each pixel value in the pixel block based on the remaining pixel values excluding the value, a maximum value and a minimum value of each pixel value in the pixel block, and the average There is provided an image processing circuit comprising threshold value calculation means for calculating the threshold value based on an average value calculated by the value calculation means.

  According to the present invention, it is possible to increase the compression rate of image data without degrading the image quality.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of a liquid crystal display device incorporating an image processing circuit according to the present embodiment. 1 includes a liquid crystal panel (LCD) 1, an LCD source driver 2 that drives the liquid crystal panel 1, and an RGB data transmission circuit 3 that supplies RGB image data to the LCD source driver 2. It has. The RGB data transmission circuit is, for example, an application processor.

  The LCD source driver 2 includes an RGB data receiving circuit 4, an RGB / YCC conversion circuit 5, a BTC encoding circuit 6, an encoded data memory 7, a BTC decoding circuit 8, a YCC / RGB conversion circuit 9, And an LCD driving circuit 10.

  The RGB data receiving circuit 4 receives the image data transmitted from the RGB data transmitting circuit 3. The RGB / YCC conversion circuit 5 converts RGB image data into YCC data composed of luminance data Y and color difference data Cb and Cr. The BTC encoding circuit 6 generates encoded data corresponding to the luminance data Y and the color difference data Cb and Cr. The generated encoded data is stored in the encoded data memory 7.

  The BTC decoding circuit 8 reads the encoded data stored in the encoded data memory 7 and restores the luminance data Y and the color difference data Cb, Cr. The YCC / RGB conversion circuit 9 converts the restored luminance data Y and color difference data into RGB image data. The LCD drive circuit 10 converts the converted image data into an analog pixel voltage and supplies it to the liquid crystal panel 1.

  FIG. 2 is a block diagram showing an example of the internal configuration of the BTC encoding circuit 6 of FIG. The BTC encoding circuit 6 in FIG. 2 includes a 4H line memory 11, a 16-pixel storage register 12, a change amount calculation unit 13, a maximum / minimum calculation unit 14, a threshold value calculation unit 15, a representative value calculation unit 16, A bitmap calculation unit 17 and a compressed data storage register 18 are included.

  The 4H line memory 11 has a memory capacity enough to store image data for four horizontal lines. The 16-pixel storage register 12 stores image data for one pixel block (4 × 4 = 16 pixels) among the image data stored in the 4H line memory 11.

  The change amount calculation unit 13 calculates the change amount of the luminance data Y and the change amounts of the color difference data Cb and Cr in the pixel block. The maximum / minimum calculation unit 14 calculates the maximum value and the minimum value of the luminance data Y in the pixel block, and the maximum value and the minimum value of the color difference data Cb and Cr.

  The threshold value calculation unit 15 calculates the threshold value of the luminance data Y and the threshold values of the color difference data Cb and Cr in the pixel block. The representative value calculator 16 calculates the representative value of the luminance data Y and the representative values of the color difference data Cb and Cr in the pixel block.

  The bitmap calculation unit 17 calculates a bitmap of encoded data of luminance data Y and a bitmap of encoded data of color difference data Cb and Cr in the pixel block.

  The compressed data storage register 18 stores encoded data of luminance data Y and encoded data of color difference data Cb and Cr in the pixel block. The encoded data for one pixel block stored in the compressed data storage register 18 is transferred to the encoded data memory 7.

  FIG. 3 is a block diagram showing an example of the internal configuration of the BTC decoding circuit 8 of FIG. The BTC decoding circuit 8 of FIG. 3 includes a compressed data storage register 21, a flag extraction unit 22, a representative value extraction unit 23, a bitmap extraction unit 24, a 16 pixel storage register 25, and a 4H line memory 26. Have.

  The compressed data storage register 21 stores encoded data for one pixel block stored in the encoded data memory 7. As will be described later, the encoded data includes flag information indicating a compression format, a bit string indicating a representative value, and a bitmap indicating code information for 16 pixels. For such encoded data, the flag extraction unit 22 extracts flag information, the representative value extraction unit 23 extracts a bit string representing the representative value, and the bitmap extraction unit 24 extracts a bitmap.

  Each extracted bit string is converted into luminance data Y and color difference data Cb, Cr and then stored in the 16-pixel storage register 25. The luminance data Y and the color difference data Cb and Cr stored in the 16 pixel storage register 25 are stored in the 4H line memory 26. The encoded data stored in the 4H line memory 26 is converted into RGB image data by the YCC / RGB conversion circuit 9 and then transferred to the 1H line memory 27 in the LCD drive circuit 10.

  FIG. 4 is a block diagram showing a modification of the internal configuration of the BTC decoding circuit 8 of FIG. In the BTC decoding circuit 8 of FIG. 4, the connection order of the YCC / RGB conversion circuit 9 and the 4H line memory 26 is reverse to that of FIG.

  In the case of FIG. 4, the image data stored in the 4H line memory 26 can be transferred to the 1H line memory 27 in units of 1H (one horizontal line), and the transfer efficiency is improved. In the case of FIG. 4, the circuit configuration of the BTC decoding circuit 8 is random logic, and the 4H line memory 26 and the 1H line memory 27 having a regular structure are grouped outside the BTC decoding circuit 8. Can design.

  The image processing circuit according to the present embodiment includes at least one of the BTC encoding circuit 6 and the BTC decoding circuit 8 of FIG.

  FIG. 5 is a flowchart showing an example of a schematic processing procedure of the image processing circuit according to the present embodiment. This flowchart corresponds to the processing operations of the RGB / YCC conversion circuit 5 and the BTC encoding circuit 6 of FIG.

  First, RGB image data is converted into a YCC color space (step S1). Next, an image change amount in the pixel block is calculated (step S2).

  Next, for each of the luminance data Y and the color difference data Cb and Cr, it is determined whether or not the amount of change in the pixel block is less than a predetermined threshold (step S3).

  If the change amount is less than the threshold value, the pixel block is subjected to an encoding process using a two-level difference BTC described later (step S4). When the amount of change is less than the threshold value, it indicates that the pixel block is a flat image, and the luminance data and color difference data of each pixel in the pixel block may be expressed in two levels. Encoded data is generated by replacing the luminance data Y of the pixels and the color difference data Cb and Cr with two levels.

  On the other hand, if the amount of change is equal to or greater than the threshold value, the pixel block is subjected to encoding processing using a 3-level maximum / minimum exclusion BTC described later (step S5). That the amount of change is equal to or greater than the threshold value indicates that the change in the image in the pixel block is severe. Therefore, each pixel in the pixel block is expressed by three levels. In addition, since color information cannot be visually perceived as sensitively as luminance information, color difference data Cb and Cr may be expressed in two levels even if the amount of change is greater than or equal to a threshold value.

  In FIG. 3 described above, step S1 corresponds to color information conversion means, step S3 corresponds to activity amount detection means, and steps S4 and S5 correspond to encoding means.

  Hereinafter, the processing operation of FIG. 5 will be described in detail. FIG. 6 is a diagram showing an example of a pixel block composed of 4 × 4 pixels. The number of each pixel in FIG. 6 represents the value of the luminance data Y. FIG. 6 shows an example in which three types of luminance data values exist in the pixel block.

  In FIG. 6, the difference between the maximum value MAX and the minimum value MIN of the luminance data Y is 255-0 = 255. This difference is defined as the change amount Δ of the image data. If the amount of change Δ is small, it indicates that the image is flat, and if the amount of change is large, it indicates that the image changes drastically.

  If the change amount Δ is small, two representative values having different levels are provided in the pixel block. In this case, since the values of the representative values are similar, each representative value is not expressed as a bit string as it is, but is expressed by a median value of two representative values and a difference value of these representative values. That is, representative value 1 = median value + difference value, representative value 2 = median value−difference value.

  The median value is expressed by 8 bits, for example, but since it is assumed that the amount of change Δ is small, it is not necessary to provide 8 bits as the difference value, and for example, the difference is 5 bits. Even in this case, since two representative values can be calculated with 8-bit accuracy, 8-bit accuracy can be guaranteed.

  On the other hand, when the amount of change Δ is large, human vision cannot be recognized with 8-bit accuracy, so even rough 4 bits will not cause any problem. Therefore, in this case, although three representative values having different levels are provided, the bit width of each representative value is reduced. When a high-definition image is displayed on a small-size screen used for a mobile phone or the like, the jagged feeling of the edge change of the image can be reduced and the high image quality can be maintained by expressing the pixel block at three levels.

  The amount of change is calculated for each of the luminance data Y and the color difference data Cb, Cr, but the maximum value MAX of the amount of change of each data may be the final amount of change. That is, if the change amount of the luminance data Y is ΔY, the change amount of the color difference data Cb is ΔCb, and the change amount of the color difference data Cr is ΔCr, the final change amount Δ = MAX (ΔY, ΔCb, ΔCr) may be obtained.

  It has been empirically known that the blue sky in the natural image has a change amount Δ = 15 or less. It is also empirically known that the change amount Δ = 31 or less at the boundary between the blue sky and the ground. Therefore, for example, the threshold value Δth in step S3 in FIG. 5 is set with reference to these empirical values. As will be described later, the threshold Δth may be calculated according to a predetermined algorithm. Hereinafter, an example in which the change amount and the threshold value are set separately for each of the luminance data Y and the color difference data Cb and Cr will be described.

  FIG. 7 is a diagram showing encoded data corresponding to the pixel block of FIG. The encoded data in FIG. 7 is generated by the BTC encoding circuit 6 in FIG.

  In the case of FIG. 6, the change amount Δ is 255-0 = 255, and it is determined that the change amount Δ is larger than the threshold value Δth, and the 3-level maximum / minimum exclusion BTC encoding process is performed. Each pixel in the pixel block is represented by three representative values. The three representative values are each represented by 4 bits as shown in FIG. Three representative values “0000”, “1000”, and “1111” are bit strings in which 4 bits on the LSB side are omitted. To these three representative values, 8 is actually added as an adjustment value by compressing to 4 bits. Thus, the representative value “0000” represents the value 8, the representative value “1000” represents 128 + 8 = 136, and the representative value “1111” represents 240 + 8 = 248.

  When 8 is not added, the brightness becomes dark as a whole, and thus the adjustment value described above is necessary. Each of the three representative values is stored as 4 bits, but actually, 4 bits “1000” is added to the LSB side and used as 8 bits.

  Since the pixel block in FIG. 6 is represented by three representative values, two bits are sufficient as information for identifying these representative values. Therefore, the information representing the pixel block bitmap is represented by 2 × 16 = 32 bits in total, with each pixel being 2 bits as shown in FIG.

  As described above, the encoded data of the pixel block in FIG. 6 includes a bit string of 4 × 3 = 12 bits in total representing three representative values, and a 32-bit bit string representing a bitmap of the pixel block, as shown in FIG. Is represented by a total of 44 bits.

  When the three representative values are compressed to 4 bits, it is not always necessary to add adjustment values if the compression method is changed. Therefore, the adjustment value may be arbitrarily changed.

  In the present embodiment, calculation is performed by a definitive algorithm described later without performing optimization by full search as in iPACKMAN, so that the hardware scale when realized by hardware can be reduced.

  FIG. 8 is a diagram illustrating another example of the pixel block, and FIG. 9 is a diagram illustrating encoded data corresponding to the pixel block of FIG. Each number in the pixel block in FIG. 8 indicates the value of the luminance data Y.

  The change amount Δ of the pixel block in FIG. 8 is 145−115 = 30. In this case, it is determined that the change amount Δ is smaller than the threshold value, and a two-level differential BTC encoding process is performed. Each pixel in the pixel block is represented by two representative values.

  The average value of the luminance data Y in the pixel block in FIG. The average value (upper average value) of pixels having a value equal to or higher than the average value (white dark area in FIG. 8) is 135, and the average value (lower average value) of pixels having a value less than the average value (shaded area in FIG. 8). Is 125. The difference value between the upper average value and the average value is 5, and the difference value between the lower average value and the average value is −5. Therefore, the value of each pixel can be roughly expressed by only the average value 130 and the difference value 5. Therefore, when encoding the pixel block of FIG. 8, two representative values in the pixel block are represented by a bit string representing the average value 130 and a bit string representing the difference value 5 from the average value.

  When the change in the image is small as in the pixel block of FIG. 8, it is easier to notice the change in the pixel unit than in the image with a rapid change as shown in FIG. 6. For this reason, the bit width of the bit string representing the representative value is increased to increase the bit accuracy in units of pixels. Specifically, as shown in FIG. 9, the bit string representing the average value is represented by 8 bits. Further, since the difference value from the average value is not so large, it is expressed by a 5-bit bit string as shown in FIG.

  Further, since the pixel block of FIG. 8 has only two representative values, these two representative values can be identified by one bit. Therefore, as shown in FIG. 9, the bitmap is expressed by a total of 16 bits.

  As a result, as shown in FIG. 9, the encoded data corresponding to the pixel block in FIG. 8 is a total of 29 bits including 8 bits representing the average value, 5 bits representing the difference value, and 16 bits representing the bitmap. Become.

  Next, the encoding process of the color difference data Cb and Cr will be described. FIG. 10 is a diagram illustrating an example of a pixel block, FIG. 11 is a diagram illustrating an example in which the pixel block of FIG. 10 is thinned out for each 2 × 2 pixel sub-block, and FIG. FIG.

  FIGS. 11 and 12 show an example in which a 4 × 4 pixel block is divided into 2 × 2 pixel sub-blocks, and a thinning process is performed to equalize the pixel values in each sub-block. Such decimation processing is generally known as 420 decimation.

  The value of the sub block is an average value of four pixels constituting the sub block. For example, the value of the upper left sub-block in the pixel block of FIG. 10 is (128 + 255 + 0 + 128) / 4 = 128.

  FIG. 12 represents the pixel block of FIG. 10 with only four pixel values. However, if 420 decimation is performed, the values of all the pixels in the pixel block can be represented with only four pixel values. It can be seen that it can be reduced.

  FIG. 13 is a diagram showing an encoding process after performing 420 thinning. The pixel block in FIG. 10 shows an example in which the amount of change Δ is large. However, since the color difference data has a smaller spatial change than the luminance data Y, it is not necessary to provide three representative values unlike the luminance data Y. Therefore, in this embodiment, two representative values are provided, and the bit length of each representative value is set to 4 bits by omitting 4 bits on the LSB side. Further, 8 is added to the bit string representing each representative value as an adjustment value for data compression.

  The example of FIG. 13 shows an example in which the bit strings representing two representative values are “0000” and “1111”, but these bit strings actually have values obtained by adding “0100” to the LSB side. That is, “0000” represents the value 8 and “1111” represents the value 248.

  Further, as shown in FIG. 13, the bitmap representing the pixel block can be expressed by only 4 bits “0100” (in order from the upper left to the lower right). As a result, the pixel block of FIG. 10 can be expressed by 4 × 2 + 4 = 12 bits.

  On the other hand, FIG. 14 is a diagram showing an example of the encoding process when the image in the pixel block is flat (when the change amount is smaller than the threshold value). When the amount of change in the pixel block is small, similarly to FIG. 9, encoded data including a bit string representing an average value in the pixel block, a bit string representing a difference value from the average value, and a bit string representing a bitmap. Is generated.

  In the case of FIG. 14, the bit string representing the average value is represented by 8 bits, the bit string representing the average value is represented by 5 bits, and the bit string representing the bitmap is represented by 4 bits. Thereby, the pixel value in the pixel block is made to have 8-bit precision.

  Note that the difference value is not quantized (compressed). In order to reduce the bit width, it is desirable to quantize the difference value, and there is a known technique for quantizing the difference value. The reason that the difference value needs to be quantized is that the difference value can take a value in the same range as the pixel value.

  In the present embodiment, the difference value is encoded only when the amount of change, which is the difference between the maximum value MAX and the minimum value MIN, is smaller than the threshold value Δth. Therefore, the difference value takes a value much smaller than the pixel value. Thereby, even if the bit width of the difference value is made smaller than the bit width of the pixel value, the accuracy of the bit width of the pixel value can be ensured.

  In FIG. 9 and the like, the threshold value Δth = 30, the difference value is in the range of −15 to +15, and the difference value can be expressed by 4 bits. In FIG. 9, considering the case where the threshold value Δth exceeds 31, the difference value is expressed by 5 bits. However, if Δth is limited to 30 or less, the difference value can be expressed by 4 bits.

  FIG. 15 is a diagram showing an example of encoded data of the luminance data Y and the color difference data Cb and Cr when the change amount Δ of the pixel block is larger than the threshold value Δth. The encoded data of FIGS. 7 and 13 are combined. Is. In the case of FIG. 15, the encoded data of the luminance data Y is 44 bits, the encoded data of the color difference data Cb and Cr is 12 bits each, and the encoded data is 68 bits in total.

  FIG. 16 is a diagram illustrating an example of the encoded data of the luminance data Y and the color difference data Cb and Cr when the change amount Δ of the pixel block is smaller than the threshold value Δth. The encoded data of FIGS. 9 and 14 are combined. Is. In the case of FIG. 16, the encoded data of the luminance data Y is 29 bits, the encoded data of the color difference data Cb and Cr is 17 bits each, and the encoded data is 63 bits in total.

  FIG. 17 is a diagram illustrating an example of encoded data in which flag information indicating a compression format is added to the first bit of the encoded data in FIGS. 15 and 16. This flag information becomes 0 when three-level maximum / minimum exclusion BTC encoding processing for generating encoded data by providing three representative values as shown in FIG. 15, and two representative values as shown in FIG. When a two-level differential BTC encoding process for generating encoded data by providing a value is performed, the value is 1.

  As described above, if determination is made independently for each of Y, Cb, and Cr, three flags are required, and a representative value and a bitmap are selected for each of Y, Cb, and Cr. Here, a configuration example of a flag determined by Δ = MAX (ΔY, ΔCb, ΔCr) and one maximum Δ is shown. Although the three independent Δ determinations are not excluded, by configuring the data with the maximum Δ, there is an advantage that the data size can be easily adjusted as shown in FIG. If it is desired to handle three flags independently, there is a structure of encoded data related to a combination of these three flag values, and it is necessary to consider the maximum size.

  17A shows the encoded data corresponding to FIG. 15, and FIG. 17B shows the encoded data corresponding to FIG.

  FIG. 18 is a diagram for explaining how to determine the threshold value described in FIG. First, the maximum value MAX and the minimum value MIN in the pixel block are detected, and the average value AVE 'of the remaining pixel values excluding the maximum value MAX and the minimum value MIN is calculated. By calculating the average value AVE ′ excluding the maximum value MAX and the minimum value MIN, the average value AVE ′ is not affected by the maximum value MAX or the minimum value MIN, and an average value close to actual luminance and color can be detected. .

  In the case of an image generated by a PC, the ratio of displaying pictures and characters is high, and the maximum value MAX and the minimum value MIN may be different from other pixel values. In this case, even if the average value is calculated using the maximum value MAX and the minimum value MIN, there is a high possibility that the value is different from the actual luminance or color.

  In addition, picture information such as window frames displayed on an image generated by a PC often has the same color over a wide range. For this reason, when the average value is calculated using the maximum value MAX and the minimum value MIN, the average value varies greatly depending on the color and size of the window frame, and the actual brightness and color of the entire screen are different. become.

  On the other hand, in the case of a natural image, even if the average value is calculated in consideration of the maximum value MAX and the minimum value MIN, the difference from the actual brightness and color is not so much felt.

  After calculating the average value AVE ′ excluding the maximum value MAX and the minimum value MIN, as shown in FIG. 18, an intermediate value between the average value AVE ′ and the minimum value MIN is set as a low level threshold Th_low, and the average value AVE ′ An intermediate value between the maximum value MAX is set as a high level threshold Th_high. The obtained low level threshold Th_low and high level threshold Th_high are used when three representative values are calculated in the pixel block.

  If the average value is calculated by removing the maximum value MAX and the minimum value MIN as in the present embodiment, the effect is particularly great when encoding an image generated by a PC. In the case of an image generated by a PC, the number of luminances and colors appearing on the entire screen is not so large. Therefore, by performing the three-level encoding as described above, the compression rate can be increased without degrading the image quality. .

  On the other hand, in the case of a natural image, since the probability that the same value exists in the image is low, an average value of each region divided by the low level threshold and the high level threshold may be used as the representative value. In the present embodiment, for simplification of processing, an average value is calculated by excluding the maximum value MAX and the minimum value MIN for both an image generated by a PC and a natural image. As a result, it is not necessary to change the processing method between the image generated by the PC and the natural image, the hardware configuration can be simplified, the amount of calculation required for the encoding process can be reduced, and the encoding process can be performed at high speed. . Therefore, the present embodiment is more effective when applied to an electronic device that does not have high hardware performance such as a mobile phone and has a limited memory capacity.

  FIG. 19 is a flowchart illustrating an example of a processing procedure in which each pixel of a pixel block is represented by three representative values. First, the maximum value MAX and the minimum value MIN in the image block are detected (step S11). Next, a high level threshold Th_high and a low level threshold Th_low, and an intermediate value MIN between the maximum value MAX and the minimum value MIN are calculated (step S12). The detailed processing procedure of step S12 will be described later.

Next, three representative values V_plus, V_mid, and V_minus in the pixel block are calculated (step S13). These three representative values are represented by the following formulas (1) to (3).
V_plus = Round1 (MAX, bit) (1)
V_mid = Round1 (MID, bit) (2)
V_plus = Round1 (MIN, bit) (3)

  The function Round1 is a function for reducing the bit width, and the bit of the argument indicates the bit width after the reduction.

  For example, 128 is “10000000” in 8 bits, but is represented by 4 bits “1000”. Actually, since 8 is added as an adjustment value at the time of bit compression, V_mid = Round1 (10000000,4) = 128 + 8 = 136. Similarly, V_plus = 248 and V_minus = 8. These V_plus, V_mid, and V_minus are representative values.

  Next, a process of distributing each pixel in the pixel block into three representative values is performed (steps S14 to S7). First, in step S14, the pixel value is compared with a high level threshold Th_high and a low level threshold Th_low. If the pixel value is equal to or higher than the high level threshold Th_high, the representative value V_plus is selected (step S15). If the pixel value is greater than the low level threshold Th_low and less than the high level threshold Th_high, the representative value V_mid is selected (step S16). If the pixel value is below the low level, the representative value V_min is selected (step S17).

  FIG. 20 is a flowchart showing an example of a detailed processing procedure of step S18 of FIG. First, a variable SUM for calculating the total pixel value and a variable CNT for measuring the number of pixel values are both initially set to zero (step S21).

  Next, it is determined whether the pixel value is the maximum value MAX or the minimum value MIN (step S22). If the determination in step S22 is NO, the pixel value is added to the variable SUM, and the variable CNT is increased by 1 (step S23).

  If the determination in step S22 is YES, or if the process in step S23 is completed, it is determined whether or not the processing for all the pixels in the pixel block has been completed (step S24). When determination of step S24 is NO, the process after step S22 is repeated.

  Through the processing in steps S22 to S14, the cumulative addition value of the pixel values excluding the maximum value MAX and the minimum value MIN is held in the variable SUM, and the total number of pixel values excluding the maximum value MAX and the minimum value MIN is stored in the variable CNT. Retained.

  If the determination in step S24 is yes, it is determined whether the variable CNT is zero (step S25). When the variable CNT is zero, all the pixels in the pixel block have the maximum value MAX or the minimum value MIN, which indicates that there is no intermediate value. In this case, it is assumed that two levels (or one level) are handled. As a process of (2), an intermediate value between the maximum value MAX and the minimum value MIN is calculated, and this is set as the intermediate value MID (step S26).

  If the determination in step S25 is no, it is determined whether the maximum value MAX and the minimum value MIN are equal (step S27). If they are equal, there is no intermediate value, but the maximum value MAX is set as the intermediate value as a temporary process (step S28).

  If the determination in step S27 is NO, it indicates that an intermediate value exists, and a value obtained by dividing the variable SUM by the variable CNT is set as the intermediate value MID (step S29). This intermediate value MID is nothing but an average value of pixel values excluding the maximum value MAX and the minimum value MIN. This step S29 corresponds to an average value calculation means.

When the process of step S26, S28 or S29 is completed, a threshold value is calculated based on the following equations (4) and (5) (step S30). This step S30 corresponds to threshold setting means.
Th_high = ROUND ((MAX + MID) / 2) (4)
Th_low = ROUND ((MIN + MID) / 2) (5)

  The feature of the flowchart of FIG. 20 is that the search process is not performed for optimization. Since the process is performed with a definitive algorithm, a small amount of calculation is required and the calculation time can be shortened. Therefore, according to the flowchart of FIG. 20, the hardware configuration of the BTC encoding circuit 6 shown in FIG. 2 can be simplified.

  As described above, in the first embodiment, it is detected whether or not the amount of change in the image in the pixel block exceeds the threshold. If the amount of change exceeds the threshold, it is determined that the change in the image is severe, A three-level maximum / minimum exclusion BTC encoding process is performed in which each pixel is converted into three representative values (two representative values in the case of color difference data) and encoded. Therefore, a two-level differential BTC encoding process is performed in which each pixel in the pixel block is converted into two representative values and encoded. In the 3-level maximum / minimum exclusion BTC encoding process, the threshold value is calculated based on the average value of the pixel values excluding the maximum value MAX and the minimum value MIN in the pixel block, and therefore, three representatives that match the actual luminance and hue. A value can be selected. Since the three representative values are each expressed by 4 bits, the bit width can be reduced. In the two-level differential BTC encoding process, two representative values are selected using the 420 decimation process, and each representative value is represented by 8 bits, so that 8-bit accuracy is obtained.

  With such a process, the bit width of the encoded data can be further reduced without degrading the image quality in any of the 3-level maximum / minimum exclusion BTC encoding process and the 2-level differential BTC encoding process.

(Second Embodiment)
The second embodiment is characterized in that the three-level maximum / minimum exclusion BTC encoding process is different from that of the first embodiment. Since the rest is the same as that of the first embodiment, the following description will focus on the differences.

  FIG. 21 is a diagram for explaining the outline of the 3-level maximum / minimum exclusion BTC encoding process according to the second embodiment. FIG. 21A is a diagram showing an example of a bitmap when each pixel in the pixel block is expressed by three representative values.

  Considering a group of three pixels in a pixel block, there are 3 × 3 × 3 = 27 combinations of representative values in this group. Therefore, 5 bits (2 5 = 32> 27) are sufficient to represent each group. Accordingly, each group can be reduced by 1 bit, rather than expressed by 2 × 3 = 6 bits.

  In FIG. 21A, three adjacent pixels are shown as a group by a broken line. In the pixel block, five groups and one remaining pixel are formed. Each group is expressed by 5 bits as shown in FIG. Therefore, the bit map can be expressed by 5 × 5 + 2 = 27 bits as shown in FIG. 21C, and is reduced by 5 bits compared to the 32-bit bitmap as shown in FIG.

  In addition, although the example which makes 3 pixels one group was demonstrated in FIG. 21, you may comprise a group by multiple pixels other than 3 pixels. However, grouping that reduces the bit width of the bitmap as much as possible is desirable. From the viewpoint of reducing the bit width, it is desirable to increase the number of groups. Therefore, it is desirable to perform grouping in consideration of the bit width of the bitmap.

  FIG. 22 is a diagram illustrating an example of encoded data generated in the second embodiment. The encoded data in FIG. 22 differs from that in FIG. 17 in the case where the flag information indicating the compression format is 0. More specifically, the bit strings constituting the bitmap are different, and the total bit width when the flag information is 0 is 64 bits, which is reduced by 5 bits compared to FIG.

  In FIG. 21, the method of converting 6 bits to 5 bits has been described, but other methods of reducing bits may be employed. Specifically, there is a method of converting a value in ternary display into a value in binary display. For example, since 00101 is 0 × 9 + 1 × 3 + 1 = 4, it becomes 011 in the ternary display. Therefore, when converted to binary display, it becomes 00100, and it can be seen that 1 bit is reduced to 5 bits. This method is only an example, and other methods may be adopted.

  As described above, in the second embodiment, when the 3-level maximum / minimum exclusion BTC encoding process is performed, the bit width representing the bitmap can be further reduced, so that the bit width of the encoded data can be further reduced.

(Third embodiment)
The third embodiment is characterized in that the contents of the two-level differential BTC encoding process are different from those in the first and second embodiments.

  As described above, in the two-level difference BTC encoding process, the average value of the pixel block and the difference value from this average value are included in the encoded data. As the difference value is smaller, the appearance frequency is higher. Therefore, as many 0s as possible are included in the bit string representing the difference value having a high appearance frequency so that the difference value having the higher appearance frequency has a lower bit transition probability. As a result, power consumption when the encoded data is transferred from the BTC encoding circuit 6 to the encoded data memory 7 is reduced.

  For example, a code 00000 is assigned to the difference value 0, and a code 11111 is assigned to the difference value 31 having the lowest occurrence frequency. Thereby, the data stored in the encoded data memory 7 is biased to zero. When the bit transitions from 0 to 1 or from 1 to 0, a current flows through the encoded data memory 7 and the power consumption increases. Therefore, the power consumption of the encoded data memory 7 can be reduced by reducing the number of bit transitions.

  FIG. 23 is a diagram illustrating an example of encoded data corresponding to a difference value. Here, transition means that the bit value changes between frames in the same storage area of the encoded data memory 7, and does not mean that 0 and 1 of the bit string representing the difference value change. Therefore, a bit string with many 0s in the bit string is preferentially selected over a bit string with few bit transitions.

  In the above description, encoding of the difference value has been described. However, the same principle can be applied to a bit string representing a bitmap. Adjacent pixels in the bitmap often take the same pixel value. Therefore, instead of the conversion from 6 bits to 5 bits described in the second embodiment, conversion to 5 bits may be performed in consideration of the occurrence frequency of the bit string.

  FIG. 24 is a diagram illustrating an example of encoded data corresponding to a bitmap of a pixel block. In the “combination” column of FIG. 24, the bitmap value 00 is a value 0, the bitmap value 01 is a value 1, and the bitmap value 10 is a value 2. For example, the combination of bitmaps “00”, “01”, and “01” is 011.

  There are 27 types of groups in FIG. 21A in total, and each group is represented by a 3-bit width. In FIG. 21A, the three pixels constituting the group need not be adjacent to each other, but in the present embodiment, one group is constituted by the adjacent three pixels.

  Those having a small change amount among the three pixels constituting the group are combinations “000”, “111”, and “222” that take the same value for all three pixels. Therefore, for these combinations, encoded data including at most one 1 is allocated.

  Conversely, the combination “020” has the largest amount of change. Since the value changes from 0 to 2 and then returns to 0, the amount of change is 2 + 2 = 4. Therefore, all “1” encoded data is assigned to the combination “020”.

  FIG. 25 is a block diagram showing the internal configuration of the BTC encoding circuit 6 according to the third embodiment, and FIG. 26 is a block diagram showing the internal configuration of the BTC decoding circuit 8 according to the third embodiment. In FIG. 25 and FIG. 26, the same reference numerals are given to the components common to FIG. 2 and FIG. 3, and the differences will be mainly described below.

  The BTC encoding circuit 6 in FIG. 25 includes a first transition minimum encoding circuit 31 and a second transition minimum encoding circuit 32 as a configuration not shown in FIG. As shown in FIG. 23, the first transition minimum encoding circuit 31 generates encoded data corresponding to the difference value. The second transition minimum encoding circuit 32 generates encoded data corresponding to the bitmap as shown in FIG.

  Final encoded data is generated using the encoded data generated by the first and second transition minimum encoding circuits 31 and 32 and stored in the compressed data storage register 18.

  The BTC decoding circuit 8 of FIG. 26 includes a first transition minimum decoding circuit 33 and a second transition minimum decoding circuit 34 as configurations not shown in FIG. The first transition minimum decoding circuit 33 decodes the encoded data of FIG. 23 to the original difference value. The second transition minimum decoding circuit 34 decodes the encoded data in FIG. 24 into the original bitmap data.

  The difference value and bitmap data decoded by the first and second transition minimum decoding circuits 33 and 34 are stored in the 16-pixel storage register 12, and luminance data Y and color difference data for 16 pixels are generated. .

  As described above, in the third embodiment, the channel coding technique is applied to the 2-level differential BTC encoding process and the 3-level maximum / minimum exclusion BTC encoding process by utilizing the fact that the correlation of images is high. The power consumption (current) of the encoded data memory 7 is reduced. More specifically, since as many 0s as possible are assigned to a bit string representing a difference value having a high appearance frequency, the bit transition probability in the encoded data memory 7 can be reduced, and the power consumption of the encoded data memory 7 can be further reduced. . Also, as for the bit strings constituting the bitmap, as many 0s as possible are assigned to the bit strings representing the pixels with little change, so that the power consumption of the encoded data memory 7 can be similarly reduced.

  When encoding processing is performed using a variable length code, random access to the encoded data memory 7 becomes difficult. Therefore, in the present embodiment, the code length (bit width) is shared, and random access to the encoded data memory 7 is possible. The transition minimum encoding process of FIG. 24 is a reversible conversion, and information is not lost by this encoding, so there is no influence on the image quality.

(Fourth embodiment)
In the above-described 2-level differential BTC encoding process, 420 decimation processing is adopted, but at least one of the 3-level maximum / minimum exclusion BTC encoding process and 2-level differential BTC encoding process performs 422 decimation processing described below. May be.

  FIG. 27 is a diagram illustrating an example of a pixel block. FIG. 28 is a diagram showing an example in which the pixel block in FIG. 27 is thinned 422 by two pixels adjacent in the vertical direction. In FIG. 28, the average value of two pixels adjacent in the vertical direction is set as the value of the two pixels. For example, the average value of two pixels in the vertical direction from the upper left corner is (128 + 0) / 2 = 64, and the values of these two pixels are both set to 64.

  FIG. 29 is a diagram in which the processing result of FIG. 28 is expressed by 8 pixels. Note that 422 thinning may be performed in units of two pixels adjacent in the horizontal direction instead of two pixels adjacent in the vertical direction.

  FIG. 30 is a diagram illustrating an example in which 422 decimation processing is performed in the two-level differential BTC encoding processing. In the case of FIG. 30, the bitmap is increased from 4 bits to 8 bits as shown in FIG. 13, so that the spatial resolution can be improved, and an image closer to the original image can be obtained than 420 thinning.

  In the encoded data in FIG. 30, two representative values are each 4 bits wide, and 8 is added to each representative value as an adjustment value by data compression.

  FIG. 31 is a diagram for explaining the horizontal change amount and the vertical change amount when performing vertical and horizontal 422 decimation. The total change amount between the pixels in the horizontal direction is the horizontal change amount, and the total change amount between the pixels in the vertical direction is the vertical change amount.

  In the case of FIG. 31, the horizontal change amount = 0 and the vertical change amount = 4 × 5 + 4 × 20 + 4 × 5 = 120. The horizontal change amount is compared with the vertical change amount, the horizontal direction with a small change amount is selected, and 422 thinning is performed.

  FIG. 32 is a diagram showing an example of encoded data generated by performing 422 decimation in the horizontal direction for the example of FIG. FIG. 32A shows the case where the BTC mode selection flag is 0 (when the 3-level maximum / minimum exclusion BTC encoding process is performed on the luminance data Y), and only one of the color difference data Cb is thinned out 422 to obtain the color difference. Data Cr shows an example in which 420 thinning is performed as in FIG. For the color difference data Cr, an HV selection flag indicating that the change amount is calculated in the horizontal or vertical direction is provided. Since this flag is 0, it indicates that the amount of change is calculated in the horizontal direction.

  The reason why the 422 thinning is performed for the color difference data Cb instead of the color difference data Cr is because Cr has a stronger correlation between pixels than Cb, and even if the 422 thinning is performed for the color difference data Cr, the image quality is not improved so much. is there.

  On the other hand, FIG. 32B shows a case where the BTC mode selection flag is 1 (when the 2-level difference BTC encoding process is performed on the luminance data Y), and an example in which 422 decimation is performed for both the color difference data Cb and Cr. Show.

  In FIG. 32A, the reason why only the color difference data Cb is subjected to 422 thinning is to make the total number of bits of encoded data the same when the BTC mode selection flag is 0 and 1.

  Note that this fourth embodiment may be combined with the third embodiment.

  Thus, in the fourth embodiment, since the 422 thinning process is performed on the color difference data, the image quality can be improved as compared with the 420 thinning process.

  In each of the above-described embodiments, when the change amount is larger than the threshold value, three representative values are provided for the luminance data Y, and two representative values are provided for the color difference data Cb and Cr, and the change amount is equal to or less than the threshold value. In the above, an example in which two representative values are provided for both the luminance data Y and the color difference data Cb and Cr has been described, but the number of representative values is not particularly limited. When the number of representative values is generalized, n (n is an integer of 2 or more) representative values are provided when the change amount is larger than the threshold value, and n or less representative values are provided when the change amount is less than or equal to the threshold value. Will be provided.

1 is a block diagram showing a schematic configuration of a liquid crystal display device incorporating an image processing circuit according to an embodiment. The block diagram which shows an example of an internal structure of the BTC encoding circuit 6 of FIG. The block diagram which shows an example of an internal structure of the BTC decoding circuit 8 of FIG. The block diagram which shows the modification of the internal structure of the BTC decoding circuit 8 of FIG. 5 is a flowchart illustrating an example of a schematic processing procedure of the image processing circuit according to the present embodiment. The figure which shows an example of the pixel block which consists of 4x4 pixels. The figure which shows the encoding data corresponding to the pixel block of FIG. The figure which shows the other example of a pixel block. The figure which shows the encoding data corresponding to the pixel block of FIG. The figure which shows an example of a pixel block. The figure which shows the example which thinned out the pixel block of FIG. 10 for every 2 * 2 pixel subblock. The figure which represented the pixel block only with the value of four subblocks. The figure which shows the encoding process after performing 420 thinning. The figure which shows an example of an encoding process in case the image in a pixel block is flat. The figure which shows an example of the coding data of luminance data Y and color difference data Cb, Cr when the variation | change_quantity (DELTA) of a pixel block is larger than threshold value (DELTA) th. The figure which shows an example of the encoding data of luminance data Y and color difference data Cb and Cr when the variation | change_quantity (DELTA) of a pixel block is smaller than threshold value (DELTA) th. (A) and (b) are figures which show an example of the coding data which added the flag information which shows a compression format to the head bit of the coding data of FIG.15 and FIG.16. The figure explaining how to determine the threshold value explained in FIG. The flowchart which shows an example of the process sequence which represents each pixel of a pixel block with three representative values. 20 is a flowchart showing an example of a detailed processing procedure of step S18 of FIG. (A), (b), and (c) are diagrams for explaining an overview of a 3-level maximum / minimum exclusion BTC encoding process according to the second embodiment. (A) and (b) are figures which show an example of the coding data produced | generated in 2nd Embodiment. The figure which shows an example of the coding data corresponding to a difference value. The figure which shows an example of the encoding data corresponding to the bitmap of a pixel block. The block diagram which shows the internal structure of the BTC encoding circuit 6 by 3rd Embodiment. The block diagram which shows the internal structure of the BTC decoding circuit 8 by 3rd Embodiment. The figure which shows an example of a pixel block. FIG. 28 is a diagram showing an example in which the pixel block in FIG. 27 is subjected to 422 thinning by two pixels adjacent in the vertical direction. The figure which expressed the processing result of FIG. 28 by 8 pixels. The figure which shows the example which performs 422 thinning-out processing in 2 level difference BTC encoding processing. The figure explaining the horizontal variation | change_quantity and vertical variation | change_quantity in the case of performing 422 thinning | decimation in the vertical direction and a horizontal direction. The figure which shows an example of the coding data produced | generated by performing 422 thinning | decimation of the horizontal direction about the example of FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Liquid crystal panel 2 LCD source driver 3 RGB data transmission circuit 4 RGB data reception circuit 5 RGB / YCC conversion circuit 6 BTC encoding circuit 7 Encoded data memory 8 BTC decoding circuit 9 YCC / RGB conversion circuit 10 LCD drive circuit 11 4H Line memory 12 16 pixel storage register 13 change amount calculation unit 14 maximum / minimum calculation unit 15 threshold calculation unit 16 representative value calculation unit 17 bitmap calculation unit 18 compressed data storage register 21 compressed data storage register 22 flag extraction unit 23 representative value extraction unit 24 Bitmap extraction unit 25 16 pixel storage register 26 4H line memory 27 1H line memory

Claims (5)

An activity amount detecting means for determining whether or not a difference between pixel values of a plurality of pixels in the pixel block exceeds a predetermined threshold for each of the pixel blocks composed of a plurality of adjacent pixels;
Threshold setting means for setting the predetermined threshold for each of the pixel blocks;
First encoding that converts each pixel value in the pixel block determined to have exceeded the predetermined threshold by the activity amount detection means into n (n is an integer of 2 or more) representative values and performs bit compression. A second encoding process for performing bit compression by converting each pixel value in the pixel block determined to have not exceeded the predetermined threshold by the activity amount detection means to the n or less representative values Encoding means for performing
The threshold setting means includes
Average value calculating means for calculating an average value of the pixel values in the pixel block based on the remaining pixel values excluding the maximum value and the minimum value of the pixel values in the pixel block;
Threshold value calculating means for calculating the threshold value based on a maximum value and a minimum value of each pixel value in the pixel block and an average value calculated by the average value calculating means; Image processing circuit. The first encoding process corresponds to flag information indicating that the first encoding process has been performed, a bit string obtained by bit compression processing of the n representative values, and each pixel in the pixel block. Generating first encoded data including bitmap information indicating assignment of representative values;
The second encoding process corresponds to flag information indicating that the second encoding process has been performed, a bit string obtained by bit compression processing of the n or less representative values, and each pixel in the pixel block. The image processing circuit according to claim 1, wherein second encoded data including bitmap information indicating assignment of representative values to be generated is generated.   The image processing circuit according to claim 2, wherein the first encoded data is a value obtained by adding a common value to a value corresponding to a bit string obtained by bit-compressing the n representative values. The second encoding process includes:
After dividing the pixel block into a plurality of vertically or horizontally long sub-blocks, assigning one of the n or less representative values to each of the plurality of sub-blocks to generate the second encoded data The image processing circuit according to claim 2, wherein: Color information conversion means for converting image data consisting of a plurality of color information into luminance data and color difference data,
The activity amount detecting means determines, for each of the pixel blocks, whether or not each of a difference in luminance data and a difference in color difference data of a plurality of pixels in the pixel block exceeds a predetermined threshold value,
The first encoding process includes:
A third encoding process for converting the luminance data of a plurality of pixels in the pixel block determined to have exceeded the predetermined threshold by the activity amount detection means into three representative values and performing bit compression;
A fourth encoding process that converts the color difference data of a plurality of pixels in the pixel block determined to have exceeded the predetermined threshold by the activity amount detection means into two representative values and performs bit compression, and
The second encoding process includes:
A fifth encoding process for converting the luminance data of a plurality of pixels in the pixel block determined not to exceed the predetermined threshold by the activity amount detection means into a median value and a difference value and bit-compressing the data;
A sixth encoding process for converting the color difference data of a plurality of pixels in the pixel block determined not to exceed the predetermined threshold by the activity amount detection means into a median value and a difference value and performing bit compression; The image processing circuit according to claim 1, further comprising:

Images