HDe 1.06 * Filter - Region It can control distortion of a spatially rectangular region. usage: - Call from [Region...] at Filter menu. - Rectangular setting window is displayed. In this state, non-grayed buttons and functions (Play, etc.) can be used. - Left-click the mouse botton on the image to specify starting point and select rectangular region. The current region is displayed in yellow rectangle. - Specify a value for the selected area. * When Fill is selected, specify a value directly. The value that can be specified is -24 through +24. The negative value indicates reducing distortion. -6 is to reduce distortion by 1/2, -12 by 1/4 and -24 by 1/16. The positive value indicates increasing distortion. +6 is to increase distortion by 2, +12 by 4 and +24 by 16. * When Add is selected, change the setting value. Increases or decreases value for the non-zero area. - If the current value is negative, it is displayed in green, and if positive, displayed in red. If the display is not clear, use Dark or Bright. - To fix the current select region, click Current region - Fix, or start selecting a new region. To cancel setting, click Cancel. Restrictions: - Up to 64 rectangular regions can be specified. If the displayed region is reduced, for example, an old region is covered by the new region, the hidden region is eliminated. Therefore it does not mean that the operating count is up to 64 times. - Setting is 16x16 (Macroblock) unit. If interlaced, 16x32 (2 Macroblocks) unit is used. - Specified rectangles are fixed for the specified time range. Motion tracking feature is not supported. - Specified values are not always used accordingly, for there is a restriction of AVC in which the distortion of consecutive macroblocks cannot be changed less than 1/4 or more than 4. So they can be changed to the acceptable range during encoding. - "Improving image quality on some parts implies deteriorating image quality on other parts" is a natural behavior. Because the bitrate is unchanged, consuming more bits to improve picture quality for some parts implies reducing bits from other parts, which causes deterioration of picture quality on that parts. * CINEMA CRAFT (R) Video Playback Kit for BlackMagic Design (OPTION) CINEMA CRAFT (R) Video Playback Kit for BlackMagic Design (abbr. VPK) is an option kit to display images on PC monitor on to external monitor through HDMI/HD SDI using BlackMagic Design card. - Please be noted that VPK is designed to work only with the BlackMagic Disign card included in this package. It is not supported to work with other cards. - Install VPK after HDe is installed. VPK checks whether HDe is installed or not. Once installed, images displayed on PC monitor are displayed on external monitor as they are. - HDe 1.06 is the first version that supports VPK, which supports VPK 1.00. However, in the future, version up of HDe or VPK alone, or both HDe and VPK may be required. When updating HDe, please check the supported version of VPK. - Please use our specified versions for Blackmagic and PC display drivers. Please be advised that using the different than our specified version of drivers is not supposed to get troubleshooted. - When Contrast, Brightness, or Blue only setting is changed, the same is applied to external monitor. However, magnifying and reducing images on PC are not applied to external monitor, so that it always displays 1:1. - When the power of the external monitor is turned on after HDe is started, images may not be displayed properly. Please avoid switching on/off external monitor or changing cables during HDe is running. - If both HD-SDI and HDMI cables are connected, images may not be displayed properly due to the frequency differences of two monitors. - If for some reason, HDe is exited abnormaly, images may not be displayed on external monitor after re-starting HDe. In such case, please reboot the PC on which HDe runs. - It is presumed that LCD connected via HDMI may not display images correctly due to the display settings for LCD, especially interlace/progressive conversion settings. For example, images that look as interlaced on PC monitor may look progressive on external monitor, or images seems stuttered by slow playback or consecutive frame-by-frame movement, etc. In those cases, changing the settings of the external LCD monitor may improve the behavior. - The following settings can be done on HDe side: Call it from Option - General. - Use BlackMagic if installed If unchecked, images are not displayed on external monitor. * Convert 24p to 60i Normally, 23.976p and 24p signals are output to external monitor as they are. In case the external monitor does not support these signals, they can be 2:3 pulled down and can be output as 59.94i or 60i. In this case, there may be a difference in still image state for some kind of external monitors. * Synchronize PC and SDI/HDMI timing PC monitor and external monitor do not in sync. Therefore timing of refreshing images do not match, which is not so noticeable in 24p mode, but if 60i signals are output, the timing mismatch of PC and external monitor may cause stuttering display on either of the monitors. So there are three options to control display preference for each monitor. - PC - Allow picture stuttering on ext. monitor PC monitor takes preference over external monitor and allows stuttering images on external monitor. - Ext - Allow picture tearing on PC monitor External monitor takes preference over PC monitor. It ignores vertical synchronization of PC monitor. The image display on external monitor is smoother, but picture tearing may occur on PC monitor. - Skip - Allow picture skip on PC monitor External monitor takes preference over PC monitor. The image display on external monitor is smoother, but pictures may be skipped on PC monitor. If the reading speed from HDD is not fast enough, images on external monitor may not displayed smoothly even if Ext or Skip is selected. Especially, since 4:2:2 10-bit source such as QuickTime V210 has more data per time than that of 4:2:0 8-bit decoded file, it may takes time to read frames so that playback may not be very smooth. Please remember such case may happen. ------------------------------------------------------------------------------- HDe 1.04 * NTSC 720x480, PAL 720x576 output From version 1.04, HDe can encode NTSC 720x480 and PAL 720x576. In the following description, SD represents the above resolutions, and HD represents the resolution of 1920x1080. Please pay attention to the following things. - In Blu-ray standard, there are differences in acceptable resolutions of Primary and Secondary video. (For the details, please refer to the document of Blu-ray disc.) * Bitrates For Primary video, 40 Mbps is the maximum. For Secondary video, 8 Mbps is the maximum. * Combination of Primary video and Secondary video The acceptable resolution combinations of Primary and Secondary are listed below: Primary video Acceptable Secondary video 1920x1080 59.94i 720x480 59.94i, 1920x1080 59.94i 1920x1080 50i 720x576 50i, 1920x1080 50i 1920x1080 23.976p 720x480 23.976p, 1920x1080 23.976p 1920x1080 24p 720x480 24p, 1920x1080 24p 720x480 59.94i 720x480 59.94i 720x576 50i 720x576 50i Note: 720x480 24p, 720x480 23.976p, 720x480 29.97p, 720x576 25p can be used for Secondary video, but not for Primary video. These SD progressive format can only be used for Secondary video. - The restriction about GOP for Blu-ray is rather complicated. If the max bitrate (not average bitrate) is less than 15 Mbps, approx 2 seconds is the maximum GOP length in the standard. And if the max bitrate is 15 Mbps or more, approx 1 second is the max GOP length. In multipass encoding mode, GOP structure is determined at the time when video information file is created. If max bitrate is changed afterwards, HDe behaves as follows: * First, suppose encoded with max bitrate 8 Mbps (a) Later, suppose encoded with max bitrate 18 Mbps (b) ===> Encoding error may occur. In case (a), GOP length is allowed to approx 2 seconds, but in case (b), GOP length is limited to approx 1 second. * First, suppose encoded with max bitrate 18 Mbps (a) Later, suppose encoded with max bitrate 8 Mbps (b) ===> Capable of encoding, but GOP length is limited to approx 1 second. - In general, the resolution of the source file of NTSC is 720x486 (bottom first), but the resolution of output stream will be 720x480 (top first). The encoding area out of 486 lines is specified at "Offset line" of "File specification." The initial value is 5, meaning 5 to 484 lines are targeted out of 0 to 485 lines. In the standard of Blu-ray disc, 5 is specified. When other than 5 is used, generated stream comforms to Blu-ray standard, but please note that display position at stream playback time will be different from 720x486. - The resolution of NTSC Planar 420 file (Filter, Decoder output) is 720x480, not 720x486. - Blu-ray requires multiple slices for HD, but single slice is also allowed for SD. However, encoding will be slowed down when single slice is selected. - Since Accom HVD, HVU, HVL, Planar 420 files don't have file header, HDe determines the resolution by the file size. However, for instance, if the file size is a multiple of both HD and PAL, HDe cannot determine the resolution and explicit indication must be done. In such cases, a display will be shown to prompt users to specify the resolution. ------------------------------------------------------------------------------- HDe 1.03 * Encoding with multiple files concatenated. HDe 1.03 can use multiple source files to produce one encoded file. Restrictions: - The maximum number of source files for each output file is 32. - The format of source files for each output file must exactly match with each other. - You cannot mix AVI, QuickTime V210 and so on. - Width, height, bits per pixel, frame rate must coincide with each other. Especially with AVI, 10-bit and 8-bit must not be intermingled. Usage: - In the "Main window", open "File specification window" by clicking "Add", "Insert" or "Edit" button. Then click "Add" or "Insert" button to open "Source files" dialog box and select source files. The added files are shown in the list in the "File specification window". Note: In "Source files" dialog box, you can select multiple files at a time. - If necessary, you can specify reference area for each source file. This function is called "Trim." Select a file from the list, then specify Origin (Org.), Trim (in), Trim (out), and Duration (Dur.) on the left side. On the right side displayed the location relative to the top of the referenced concatenated area. The timecode (TC) in the Trim settings is referenced only for this setting. It is not referenced from others. The TC on the right side is used in "Control window" that is shown when "Ok" button is clicked in "File specification window" and that shows images and so on. - The file numbers of Encode (Enc.) in, out, etc. that are shown in the "Control window" are counted from the "Trim in" of the first source file. Example: File A (1000 frames, Trim in 200, Trim dur 100 frames) File B (1000 frames, Trim in 300, Trim dur 200 frames) File C (1000 frames, Trim in 400, Trim dur 300 frames) With above settings, seekable and displayable area in "Control window" contains 600 frames. The breakdown of it is, File A (#200...#299) File B (#300...#400) File C (#400...#699) File# 0 in "Control window" corresponds to A(#200) and file# 250 to B(#450). Attention: - Changing Trim in of the first source file afterwards causes shifting of Enc. in, Enc. out and segment (chapter) settings specified at "Control window", for these settings are counted from the Trim in of the first source file. - Ecl files for 1.03 and later and 1.02 and earlier have a limited compatibility. Ecl files saved by 1.02 and earlier can be read by 1.03 and later. Ecl files saved by 1.03 and later can by read by 1.02 and earlier with the following restrictions: - All but the first file of multiple source files are ignored. - All Trim in and out settings are ignored. Especially if Trim in of the first source file is set, Enc. in, Enc. out and segment (chapter) settings will be shifted, for 1.02 does not have Trim in feature. - There are some cases in which some capture system, editing system or the like produce QuickTime file or AVI file containing more data than actually specified even though the specified duration is properly written in the header of the file. In this case, the initial value of "Trim out" may be set before the last frame because it is calculated from the duration in the header. Moreover, with QuickTime file, since the product of duration written in header and frame rate is not always an integral number, there may be an error less than one frame between the initial value of "Trim out" and the duration. So please beware of it, especially when concatenating multiple files. * Segment re-encoding function HDe 1.03 can update file by partially encoding it in segment (closed GOP) unit. This feature is called "Segment re-encoding." Segment boundary is not bounded to the locations specified at segment setting. The scene change point where the encoder automatically interted I frame is also a segment boundary, based on which point, partial encoding can be done. Restrictions: - It cannot be applied to CBR streams. - You cannot use segment re-encoding function to do the following things: - Replacing pictures to be encoded. (Source file must be identical to that used at the first encoding time.) - Inserting new segment division point. - Removing existing segment division point. - The maximum bitrate of global bitrate setting must not be changed at re-encoding time. Since the maximum bitrate of local bitrate setting must be less than or equal to that of global bitrate setting, you cannot specify higher than the maximum bitrate that is set at first encoding time. - In some conditions of original stream, HDe automatically extends re-encoding area in order to comply with CPB for H.264/AVC (Coded Picture Buffer, which is similar to VBV (Video Buffer Verifier) for MPEG) constraints of H.264/AVC standard. In this case, extension is done by segment unit. Usage of Segment re-encoding: - Specify encoding range. (In the setting window, it is mentioned as "Part.") - Specify average, minimum and maximum bitrates and V/C for the encoding range. - Press "Encode" button to start encoding. - When encoding for the specified area is finished, updating process for output files is invoked. (When encoding was interrupted or some errors occurred during encoding, updating process will not invoke and the files won't be updated.) The file update process is performed in the order of H.264/AVC file, Video information file, decoded file, and filtered file. Please note here the following things: - The original of H.264/AVC file and Video information file are replaced only after file update process is finished. That is, original files are not modified during the update process. Therefore an interruption of this update won't affect the original files. - The update of decoded file and filtered file is performed by directly rewriting the original file. The original file is partially updated when the update is interrupted. In this case, decoded file does not match with decoded result of H.264/AVC file. Also filtered file does not match with filtered result of encoding time. The update of these files is not necessarily done by frame unit. The original file is partially updated within a frame when the update is interrupted. Caution or tips when using segment re-encoding: - Segment re-encoding is a function that partially updates the original file. Since the original file is generated in order to keep the average bitrate specified at encoding time, changing the average bitrate of encoding range will change the overall file size or average bitrate. And that is a natural usage. When partially increasing the average bitrate at segment re-encoding, overall file size will be increased. If you consider using segment re-encoding afterwards, please specify average bitrate with some margin in the first time. * Filter The window shown by clicking "Filter..." in the "Control window" is called "Filter" window. Summary of filters: HDe performs filtering process as the following sequence. In {} is shown the internal data format after each process is done. (a) Source input { 4:2:2 8/10 bits or 4:2:0 8 bits } (b) Extend to 4:4:4 10 bits { 4:4:4 10 bits } (c) "Texture" and "Spike" { 4:4:4 10 bits } (d) Horizontal linear filter { 4:2:2 12 bits } (e) Vertical linear filter { 4:2:0 14 bits } (f) Add dither noise { 4:2:0 14 bits } (g) Convert to 4:2:0 8 bits { 4:2:0 8 bits } (h) Feeding the result to encoder core (*) "Q.characteristics", "Grain" is performed during encoding process. Threshold settings: During the filtering process, HDe uses threasholds (abbr TH) in several parts, which can be changed by user. When a filter has threshold setting fields, the filter is applied only when the difference between the original value and the filtered value is less than or equal to the specified thresholds. Since thresholds are specified separately for Y, Cb, and Cr, the filter is not applied unless each of the three differences is less than or equal to the respective threasholds. Please note that the range of values for each threshold is different because pixel data size is different -- 10 bits, 14 bits, etc -- in each step. Q. characteristics: In order to control compression ratio (i.e., to control data size of encoded stream), H.264/AVC uses DCT data quantization, in which certain range of values are mapped to a single quantum (discrete) value. The strength of quantization can be controlled for each 16x16 pixels in H.264/AVC. Applying the same strength of quantization to the entire pictures does not bring a visually uniform result. In this field, quantization characteristics are specified that control picture quality. The range of value for this setting is 0 through 200. When this value is small, similar quantization is applied throughout the picture. When this value is large, priority is given to the following parts and more bits are given to them: - spatially simple and having flat patterns (ex. sky) - motion is slow or simple without transformation - dark side In most cases, the optimal value lies between 16 and 64, although it is widely varied by pictures and bitrates. In general, give it a greater value when bitrate is relatively high for the given complexity of the picture (i.e., when distortion is small throughout the picture), and vice versa. Then it gives you the result in which visual destortion is less noticeable. Note: "Q. characteristics" setting is applied at encoding time. Although the setting field is in the "Filter" window, it's just for the sake of convenience, and does not affect filter output files. "Texture" or "Spike" When low-pass filter (abbr LPF, aka high-cut filter), one of the linear filters described later, is used to cut high frequency signals in order to reduce quantization noise, the image gets blurred. It's because the filter also reduces the high frequency signals contained in the edge components. "Texture" or "Spike" have a function that detects granular componets in picture and reduce or remove the edge-blurring effect. By using "Texture" or "Spike", it can effectively reduce quantization noise with keeping sharpness of edges. "Texture" or "Spike": Type "Type" setting is common for both "Texture" and "Spike". The range of value for this setting is 0 through 9. When the value is small, only the fine grains are the target of detection. When the value is large, coarser grains will also be the target of detection. "Texture": others In "Texture", specify thresholds for Y, Cb, and Cr. The filter is applied only if the difference between the original pixel value and internally generated filtered pixel value is less than or equal to the specified thresholds. The thresholds are applied to 10-bit pixel values and the range of values are 0 through 1023. Internally generated filtered picture is blended with the source picture. Hence the purpose of this setting is to reduce (not remove) grain components. Because of this nature, the effect is strengthened if applied multiple times even when the settings are the same. "Texture" can be applied up to three times. "Spike": others A spike-like noise (salt and pepper noise) cannot be removed by the "Texture". "Spike" replaces source pixels with internally generated filtered pixels when hue of the grain components to be removed are nearly similar to each other. As a result, it has a character to remove spike-like noise. "Texture" or "Spike": Technical information This section contains somewhat technical information. You can skip this section if you are not interested in the details. "Texture" or "Spike" are weighted median filters with some restrictions. Grain detection is done through the weighted median filters. Median filter is a filter that sorts the value of target point and its neighbouring values and outputs the median value of them. HDe uses nine pixels -- the target pixel and its neibouring eight pixels (up, down, left, right, obliques) as inputs to the median filters. Here, "Type" defines the weights of the median filter. Type = 9 has the strongest effect, which has the following weights. 1 1 1 1 1 1 1 1 1 All elements have the same weight. Type = 1 has the weakest effect, which has the following weights. 1 1 1 1 7 1 1 1 1 The target pixel has weight 7 times heavier than neibouring pixels. For example, source has the following pixel values: 2 8 5 3 7 6 2 4 4 When Type is 9, the filter outputs the value 4, which is the median of the sorted sequence { 2,2,3,4,4,5,6,7,8 }. When Type is 1, the filter outputs the value 7, which is the median of the sorted sequence { 2,2,3,4,4,5,6,7,7,7,7,7,7,7,8 }. Note that the target pixel 7 is repeated seven times, which is the meaning of weight. In Type = 1, the weight of the target pixel is 7, and the sum of the weights of the neighbouring pixels is 8. Please note that it means the filter modifies the target value only if all of the neighbouring values are greater (or less) than the target value. Also note that with Type = 9, thin lines are easily affected. Especially with "Spike", Type with strong effects is not appropriate. Filters with Type 2 through 8 have weights whose strength is between Type 1 and Type 9. For example, Type = 5 has the following weights. 1 3 1 3 5 3 1 3 1 "Dither", "Error diffusion" "Dither" and "Error diffusion" are applied when the format of the source file is converted to 8-bit 4:2:0. "Dither" setting By adding random noise (dithered signal) to the picture before pixel depth conversion, it diffuses errors accompanied by the conversion. This function is effective for the source in which fading or gradation processing are used and binding (contouring noise) is noticeable. The image before applying "Dither" is internally represented as 14 bits, so the Y, Cb, Cr values for "Dither" should be specified as 14-bit. A value less than or equal to 63 is less than 1 in 8-bit representation, which affects rounding result from 14 bits to 8 bits. A value greater than or equal to 64 is greater than or equal to 1 in 8-bit representation, which directly affects 8-bit 4:2:0 signals that are to be feeded to the encoder core. If 64 or greater value is used for this setting, it has an effect of smoothing the banding resulted from encoding. However, the picture quality will be degraded by the noise where no banding is taken place. Please check the decoded pictures when using such a great value. "Error diffusion" It performs pixel depth conversion and subsampling to 8-bit 4:2:0 by using error diffusion method. The error diffusion method is a kind of adaptive rounding method, also it can be regarded as adaptive dithering method. It has an effect of reducing the mach bands in gradation, useful especially for computer graphics and animations. Since it may increase distortion when applied to natural images, please use it according to the nature of the source. "Dither" and "Error diffusion" have a similar effect. However, unlike "Dither" with greater value specified, "Error diffusion" does not affect 8-bit 4:2:0 that is to be feeded to encoder core but only affect rounding process. You can use both "Dither" and "Error diffusion" together if stronger effect is needed. "Grain" During the encoding process, small amplitude of grain components, which changes the atomosphere of the pictures, may be thrown away. "Grain" adaptively compensates the small amplitude of grain components internally, so as to keep grains after encoding. Note: "Grain" setting is used during encoding time. Although the setting field is in the "Filter" window, it's just for the sake of convenience, and does not affect filter output files. "Grain" process (a) input file (b) process "Texture", linear and other filters (c) trial encode by using encoder-planned QP value minus strength, and decode them (d) subtract the result of (c) from (b) and if the difference is less than or equal to the threshold specified, add the result to (b). (e) encode the result by using encoder-planned QP value. * QP value stands for a degree of quantization. Generally speaking, the purpose of this function is to add grain parts before encoding in order to improve reproduction of the grain parts. The side effect of it is distortion is increased in other than grain parts. "Grain": "strength" and "TH" "Strength" Specifies the difference between the QP value for trial encoding and that for actual encoding. QP (quantization parameter) for H.264/AVC is represented as logarithmic scale, and if 6 is specified, quantization step size will be halved. You can think that it is specified in dB. 0 has the maximum effect. Generally 3 to 12 is to be used. "TH" It is used for grain distinction. It is applied to 8-bit images, so the range of this value is 0 through 255. "Pixel" and "Black level" The upper and lower bounds for output pixel values can be specified. To do this, specify values at "Pixel" fields. The value should be specified as 8-bit representation for each Y, Cb, and Cr, separately. Min=0 and max=255 means there is no restriction. There are some cases in which the part of images the pixel level of which is close to black is rough and harsh. In such cases, Y signal level of the pixels that is below a specified level can be raised to a specified value. To do this, specify a value below which should be raised to "Black level L.T." and a value to which the target value should be raised to "Black level to". Please note here that by default, "Black level L.T." is 16 and "Black level to" is 16. If you want to keep the value below 16 as is, specify zero to both fields. Linear filters HDe has two linear filters, low-pass filter (abbr LPF, aka high-cut filter) and high-gain filter (abbr HGF). Low-pass filter passes low frequency signals but attenuates signals with frequencies higher than the specified frequency. Excessive use of this filter makes images blurred. High-gain filter does not affect frequencies lower than specified frequency but enhances high frequency signals. Excessive use of this filter enhances edges and makes images glaring. Linear filter setting: There are six linear filters for Y, Cb, Cr, horizontal, and vertical, but all filter settings are the same. "Freq." Specifies cutoff frequency. The range is 1 through 100. Small value indicates low frequency. When the "Gain" is 100, output value deviates -6dB (x 1/2) on low-pass filters and +6dB (x 2) on high pass filters at specified frequency. "Att." Low-pass filter has passband, transition band, and stopband. The specified value in dB guarantees the attenuated value in stopband. For example, value 40 stands for 2^(40/6) (2 to the 40/6 th power) = 101.6 dB, which indicates that the level of frequency signals in stopband is about 1/100. Such level of value is necessary to cut off high frequencies sufficiently. "Gain" Generally, low-pass filter cuts off high frequencies. But by reducing "Gain", it can be changed so that it mildly reduces the high frequency signals. When both low-pass and high-gain filter are used together, they negate each other at high frequencies. But if "Gain" for the low-pass filter is raised then, the low-pass filter overcomes the effect of high-gain filter at high frequencies, and thus they make the setting that emphasizes middle frequency range and reduces high frequencies. "Taps" Specifies coefficient length of the filter. The range of this value is 0 through 15, which is the length of one side of a filter. As for Y, filter with odd numbers of taps is used, and the value 8 indicates 17-tap filter. As for Cb and Cr, filter with even numbers of taps is used, and the value 8 indicates 16-tap filter. The effect of "Taps" is a similar to that of "Att." The greater the "Taps", the steeper the slope in transition band. And the less the "Taps", the more guradual the slope in transition band. Tips for the filter settings - The most popular setting at "Linear" is to use low-pass filter to reduce distortion caused by compression. Although low-pass filter, as well as compression, also creates distortion in the sense that produced picture is different from the original one. However, distortion by compression has strong temporal and spatial locality and they are visually noticeable. On the other hand, usage of low-pass filter transforms the distortion so that they seem to be natural blurrings, which is visually not very noticeable. - In "Linear", the ideal setting is to cut high frequencies efficiently without affecting low frequencies. In this sense, the ideal filter should have steep graph in transition band since gradual graph in transition band causes image blurred. However, character parts in telop and computer graphics have step-wise edges with strong contrast, at which region steep filter causes strong ringing. Because of this, steep filter setting is not suitable for those kinds of pictures. The settings that have an effect to filter slope in transition band are "Att." and "Taps", both of which have similar effect. Especially, "Taps" is more instinctive setting for ringing. Since ringing is caused by a mixture of left and right pixels or pixels above and below across the edge, the ringing band size is always less than or equal to "Taps". - "Texture" reduces only a fine amplitude of shallow bumps. By reducing only visually unnoticeable components and simplifying images, it decreases distortion caused by compression. "Texture" does not affect edges, also, it does not produce ringing. - "LPF" and "Texture" have an effect of reducing grain components. "HGF" and "Grain" have an effect of enhancing grain components. Although these four filters correlates with each other and seemingly complicated, please take them as follows: It is not appropreate to use low-pass filter or high-gain filter to control grain components. Because they do not have selectance, they produce side effect of blurring or enhancing edges, bring about ringing, and so on. Unnecessary fine grain components should be reduced beforehand by using "Texture", which simplifies images and reduces distortion caused by compression. The purpose of "Grain" is to keep grain components by changing encoding characteristics. Although there are contradictive characters in "Texture" and "Grain", it is never unnatural to use both at the same time.