Rawnalyze
Table of Content
Note: the term color often occurs in this guide relating to pixels of the raw image. The exact meaning is raw pixel with a specific color filter. It is important to understand, that a raw pixel alone does not represent a specific color. The color of the captured scenery can be evaluated only in conjunction with the surrounding pixels of other colors.
If you have suggestions, critiques regarding the program or this manual, experienced some phenomenon looking
like a program error or want to express your opinion, please email to
If you want to be notified about new versions of this program, drop a line to
Magnifying, reducing the image size
Scrolling, positioning the image
Changing the original raw data
Dialog elements
After starting Rawnalyze, following dialog window appears:
| The capture has been created in the initial state |
The first step after having started the program is usually locating the raw file and opening it.
The file specification can be typed in directly; clicking on Process
starts opening and reading the file. However, more often the Browse subdialog
will be used to select the raw file to process. After a file has been selected, it will be opened and read immediately,
there is no need to click on the Process button; in fact, there is no need to click
on Process any more while analyzing the same raw image. The only exception is,
when switching between analyzing modified and unmodified raw data, see Changing the original raw data later.
Clicking on Process initiates opening and reading the raw file again,
causing a considerable delay in response time.
The shortcut for starting the file browse dialog is Ctrl-O.
Displaying the manual through a Web browser via the Help function may become necessary, but if someone is reading this page, then that goal has been accompished already.
As there is not much else to say about this screen shot, let's skip ahead through opening an actual raw file.
|
|
We go through the components of this window, so that we can refer to the controls in the following chapters.
The initial selection is composite. Any selection remains effective after opening another image.
denotes the normal orientation,
denotes the left turned orientation,
which is how most cameras are held for portrait orientation,
denotes the right turned orientation,
which is how a few people are shooting in portrait mode,
denotes the upside down orientation.
The initial orientation is the one recorded in the raw file. If the orientation is not recorded, then the initial orientation is upright. When clicking on an arrow, the displayed orientation is changed to the one corresponding to that arrow.
The image can be rotated by 90° to the left with the keyboard action Ctrl-L and 90° to the right with the keyboard action Ctrl-R.
The initial position of this slider is usually at null, but in some cases it shows an adjustment, which is indicated in the raw file. Example: Canon 40D images shot with the Highlight Tone Priority (HTP) option start with an adjustment of +1.
Note, that only the color intensity of the displayed image is affected by the selected function. The colors themselves are of the camera's color space and don't get transformed in sRGB, etc.
Notes:
Notes on the pixel statistics:
indicates 3x magnification, while 
shows,
that the reduced size image is displayed.
Resizing the dialog window
Beside being minimized, maximized and restored, the dialog frame can be resized by dragging the right or the bottom edge, or the bottom right corner. The new size can not be narrower, nor shorter than the respective initial dimension.
The size of the image window changes together with the dialog frame size; this way the image can be viewed with less scrolling
in a larger image window.
Here are two examples for different dialog frame sizes:
Magnifying, reducing the image size
In the normal view each pixel of the raw image is displayed as a pixel on the monitor. The image can be magnified by typing in + (plus) character; repeated typing can increase the magnification to up to ten times.Typing in the - (minus) character reduces the magnification. In composite color and greyscale display mode the display size can be effectively reduced, so that each color filter array block is represented by a single pixel.
The normal view (no magnification, no reduction) can be reinstated immediately by entering Alt, Ctrl and S.
The current magnification/reduction state is displayed above the top left corner of the image frame.
Examples:
| This is the image from the selection example below in reduced size |
| and here the same image in six times magnification |
Scrolling, positioning the image
Raw images from newer cameras are quite large, some of them are very large. Looking through the image, even in reduced size view can be tedious. A multitude of scolling/positioning options are available to make this task faster.As long as the mouse pointer is not over the display area (while a scroll bat is active):
This way a precise position can be established.
Working with the sliders
A slider can be adjusted by dragging it: left-clicking on the slider pointer, holding down the left button and moving the mouse. Clicking on the trackbar (the line going in the length of the slider) at the left or right to the slider pointer moves the slider by one "page".While the focus is on a slider, it can be moved just like a scroll bar, using the arrow keys, PageUp, PageDown, Home and End. The magnitude of change the arrow keys or PageUp and PageDown cause depends on the slider.
The slider becomes focused after having clicked on any part of it, and it remains focused until clicking on another control
(a button, an input field, etc.) If the slider is focused, it is framed by a thin dotted rectangle, like this one:
Each sliders is accompanied by a Reset button, which sets the slider to its initial position.
There are other, very special ways to control the black point and white point values, and thereby the respective sliders; those are described in the Pixel selection.
Changing the original raw data
The raw data remains unchanged in memory in the entire duration of the analysis. However, the raw pixel values may be changed on a particular way immediately after having read the data. That change can not be reversed, the raw file has to be opened and read again in order to work with the original rax pixel values.Following options are available:
Black level correction becomes somewhat inaccurate through this change, although changing only two-three bits from 14 does not make this noticable,
Black level corrections will be applied before this change and then reset to zero, so that it do not get applied later again. Also the white point and the histogram range will be adjusted.
The special key combination has to be kept pressed down when clicking on the Process button or when starting processing of a raw file in the Browse dialog. It's enough to keep these keys pressed one or two seconds long.
Note, that
Pixel selection
An important aspect of the raw pixel analysis is determining, which pixels should be involved in it and how. The selection is based on the color and value of raw pixels.
The effect of pixel color and value selection is described in other chapters, here we deal only with the way the selection occurs.
The tools of pixel selection are:
The current values are shown in the respective input field. The range of possible values for both limits is
The initial value of the black point is either
The initial value of the white point is either
The black point and white point can be set to the same value, thereby limiting the range to a single pixel value. The black point can not be higher than the white point. Selecting a higher value for black point than the current white point is, or a lower value for white point than the current black point is results in changing the other limit as well to the same value.
As the pixel value limits need to be adjusted very often, sometimes with small changes, sometimes faster, a multitude of adjustment options are available. The values can be changed by
The special controls are:
All the above actions change the other limit as well, if required, to maintain that the black point is never higher than the white point.
Following actions always change both limits together from their actual value, no matter if the focus is in the black point or white point input field:
When reaching the lower or upper limit of specifications, it can occur that only one of the values get changed.
For example if the blackpoint is 500 and the whitepoint is 3000 and Outside range is not checked, then pixels with values from 500 to 3000, including 500 and 3000, are selected. However, if Outside range is selected, then pixels with values from 0 to 500 and from 3000 above are selected, including 500 and 3000.
The statistics will be changed when checking or unchecking Outside range, reflecting the changed selection. The sum of the pixel counts in inside and outside mode is not necessarily the total number of the pixels with the respective color, because pixels with values matching a limiting value are counted both inside and outside.
This option can be checked or unchecked only in exposure display mode, therefor it may be necessary to change to exposure display mode in order to check or uncheck Outside range and change back to the other display mode, so that the statistics are adjusted according to this option.
The numbers under the color selection checkboxes in the black point and white point cntrol group reflect the number of pixels in the respective group, determined by the actual black point and white point values and by the inclusion/exclusion selection. The proportion of the selected pixels too is displayed, related to the total number of pixels of the particular color in the related group.
The pixel statistics values are not updated after certain selection adjustments, in order to shorten the response time. In such cases the statistics is not shown; typing Enter causes the statistics recalculated and displayed.
Both the black point and the white point control groups contain a checkbox for each color. In channel related display mode a color is selected if at least one of the respective boxes are selected (i.e. either in the black point control group or in the white point control group).
In exposure display mode the selection is more complicated. If Outside range is unchecked, then
If Outside range is checked, then
The statistics shown in the black point and white point control groups are not affected by the color selections. However, the distribution of the values between the black point and white point control groups reflects the separation of "black" and "white" (or "lower" and "upper") range.
Area selection
When displaying the image in normal size, a rectangular area of the display can be selected by right-clicking at a point of the display, holding down the right mouse button and dragging the selection in the direction and to the size needed. The selected area is marked by an orange rectangle. Another selection can be made at any time; the previous selection will be released. The selection can be removed by the keyboard action Ctrl-D.The unit of selection is the color filter array pattern, i.e. a block of 2x2 pixels, not individual pixels. Thus a single pixel can not be selected.
The selection remains active between different display modes. When changing to histogram mode or to reduced size display, the selection remains defined but not displayed, nor is the related statistics displayed. When changing back to normal size display or to image or exposure display mode from histogram display, the selection will be shown again with the related statistics.
It is important to understand, that a part of the display area is selected, not a part of the image. When the image position is changed, the selection remains on the very same place, i.e. another part of the image gets selected, and the statistics is recalculated.
The selection remains unchanged even when opening another image; this is useful when comparing two images of the same scenery.
The use of a selection is:
Following screen capture helps explaining the display associated with a selection:
| The orange colored rectangle frames the selected area;
|
The numbers in the top row, between brackets show the position of the selected area in the entire image; the first number is the horizontal position from 0, the second one is the vertical position from 0. The co-ordinates are always even numbers, because the selection occurs in units of 2x2 pixel blocks.
The bottom row contains the size of the area in pixels, width by height; these too are always even numbers.
Note, that when the image position changes, the co-ordinates of the selected area change accordingly.
The data in the first row relates to the raw pixel values. There are three groups of data, relating to the red, green and blue pixels, respectively. Each group contains four numbers:
The data in the second row reflects the RGB values, as the pixels appear on the display:
The RGB values depend, beside on the raw values, on clipping indication, black point, white point, lightness correction, white balance and the mapping function. Changing any of these settings will be immediately reflected in the selection statistics. Likewise, the effect of changing the image position (scrolling, moving the image around) becomes immediately visible in the statistics.
Display modes
There are three ways the image data can be displayed resembling a "proper" image:
All pixels of a color filter array block appear with the same color. The red component is from the red pixel of that block, the blue component is from the blue pixel and the green component is the average of the two green pixels.
Only one pixel per color filter array block is shown in reduced size display mode, it's color is calculated according the rules described above.
| 
| Left is a part of an image displayed in composite mode, normal size.
|
| 
| Left is a part of an image displayed in composite mode, reduced size.
|
Each pixel is displayed with the color and intensity of the corresponding raw pixel. This display is the closest to how one can imagine the raw pixels. Reduced size display is not supported in this mode.
The color selection checkboxes can be checked/unchecked in this mode. If a color is selected in the black and/or white group, that color is included in the display, otherwise all those pixels remain black.
This display mode allows for a more "close" examination of the image on pixel level
than the composite color mode, as the pixels are displayed on their own, without mixing
them with their neighbors, but the intensity of the displayed pixels allows for good image
recognition, in contrast to the exposure display mode.
The sharpness of an image and fine details can be judged this way better than in other modes;
an example for the usage is the comparison of images created with different lenses.
Identifying the effect of clipping too is easier in channel related display,
when not all pixels of the color filter array blocks are clipped.
As two green pixels, but only one red and one blue are displayed for every color filter array block, such an image is typically greenish, even if correctly white balanced. Furthermore, this display is darker than the composite color display, because two color components of every pixel are always 0.
| 
| Left is a part of an image displayed in channel mode, all colors enabled.
|
| 
| Left is a part of an image displayed in channel mode,
only the red channel is enabled.
|
In normal size display mode each pixel of a color filter array block is displayed as grey with the intensity according to that raw pixel value.
In reduced size display mode one grey pixel per color filter array block is displayed, with the intensity corresponding to the average of the raw pixel values.
| 
| Left is a part of an image displayed in grey scale mode, normal size.
|
| 
| Left is a part of an image displayed in grey scale mode, reduced size.
|
The effects of settings
It is most important to understand the role of the settings in the process of converting the raw pixel value in the RGB value. The effects of the settings/selections are the same in the three image display modes, and the same effects are reflected by the histogram of mapped data.
Some cameras' raw images include several rows and/or columns of pixels outside the captured image as well;
these are called masked pixels. Their values show the effect of the so-called black current,
that part of the pixel value, which is not caused by the captured light.
Other cameras record one or more direct values indicating the black levels.
The pixel values of the image have to be corrected by the calculated values, depending on the pixel's position (typically with Canon cameras), or with the recorded correction value. Other cameras do not record any information for this purpose, rather they record the corrected pixel values.
The process described above is called the black level correction.
The initial value of Black point is the minimum of the black level correction values, if there are any, zero otherwise. However, this is only for orientation; when carrying out the black correction, the program always calculates with the pixel-related value, not with the minimum.
The minimum, maximum and average of the correction values are shown with the histogram display.
The first step of the conversion is substracting the respective black level correction value from the original pixel value. A negative result will be replaced by zero (yes, it can happen that the black level correction is higher than the pixel value).
In case the resulting value is zero, black clipping is said to have occured on raw level. "Clipping" in this sense is not a generally accepted term, but this is analogous to the clipping of highlights in the sense, that the value 0 represents all smaller values, which could not be recorded.
If the Raw clipping checkbox is checked, the black-clipped pixels will be displayed with the RGB value 255, or in case of composite color it contributes to the resulting color with 255. If the raw clipping option is not selected, such pixel values are converted to the RGB value 0, respectively they contribute to the composite color with 0.
| This night shot is obviously far underexposed; but how badly? |
| The raw clipping indication shows, that the entire sky is underexposed; however, that's not bad, as nothing interesting could be found there. The lamp is overexposed into clipping, but that can be accepted |
The result of the black level correction is an absolute pixel value, starting with 0. An adjustment can be applied to either increase the absolute values, or to clip them at a certain level; this adjustment is the black point setting.
If the black point is lower than the minimum of the black levels,
then all absolute pixel values will be increased by the difference between the
black point and the minimum of black levels. This causes the image to appear lighter;
however, the relative differences between the pixels become less, i.,e. the contrast decreases.
Obviously, this can not happen if black level correction does not apply.
If the black point is higher than the minimum of the black levels, then all absolute pixel values will be reduced by the difference between the black point and the minimum of black levels. If black level correction does not apply, then the black point itself is the "clipping point".
This adjustment causes the image to appear darker, and details may be lost, as all pixel values between 0 and the black point will become "equalized", this is induced clipping, as opposed to raw clipping.
If the Show clipping checkbox is checked, then the pixels clipped by the black point but not clipped at the raw level will be displayed with the RGB value 255, or in case of composite color it contributes to the resulting color with 255. If the Show clipping option is not selected, such pixel values are converted to the RGB value 0, respectively they contribute to the composite color with 0.
| Increasing the lightness by 4 EV exaggerates the noisiness, this way it can be judged better |
| Increasing the black point by only ten levels makes the noise invisible |
| Going down with the lightness increase to 2.5 EV while keeping the black point adjustment proves, that this shot is not totally lost |
The pixels of some cameras can reach the maximum value representable with the given bit depth; for example the pixel values of the Canon 20D can reach 4095. Other cameras' pixels may clip at a lower level; the clipping level may depend on the ISO, on other factors, even on the particular copy of the camera model. It often depends on the channel (pixel color); even differences between the two green positions within the color filter array are not seldom. Some cameras' pixels (sometimes only certain channels) do not clip at a single level, but in a level range; this is the range of non-linearity. The pixels have to be regarded as clipped from the lower limit of this range, because the values within the range do not represent meaningful information about the image.
If the original (i.e. not black level corrected) pixel value is equal to or greater than the respective clipping level, then that pixel is said to be raw clipped. If the Raw clipping checkbox is checked, the clipped pixels will be displayed with the RGB value 0, or in case of composite color it contributes to the resulting color with 0. If the Raw clipping option is not selected, such pixel values are converted to the RGB value 255, respectively they contribute to the composite color with 255.
The current version of Rawnalyze contains the clipping levels of most supported cameras hard coded; however, those values may not be accurate with all copies of that particular camera model. This may lead to reporting raw clipping, when it did not occur, or not reporting actual clipping.
| The lamp shades appear too bright; did clipping occur? |
| Turning on the raw clipping indication shows, that in fact some clipping occured. The effect is best visible in channel mode |
Keep in eyes, that the clipping indication shows, which colors did not clip, because the clipped pixels are substituted by 0:
The statistics of the selection in one of the patcher (marked with an orange rectangle) proves, that the blue was in fact far from clipping, but some green pixels came very close to that.
The white point specification, reduced by the current black point value sets an upper limit to the pixel values after black level correction and the application of the white balance and lightness adjustment factors. Thus the white point induces a clipping based on adjustments, as opposed to the raw clipping.
Reducing the specified white point by the black point is necessary, so that the white point value can be regarded relative to the uncorrected raw pixel values. Thus the white point gets "black point adjusted" in certain sense. If a pixel is not clipped on the raw level and the adjusted pixel value is equal to or higher than the white point and the Show clipping checkbox is checked, the clipped pixels will be displayed with the RGB value 0, or in case of composite color it contributes to the resulting color with 0. If the Show clipping option is not selected, such pixel values are converted to the RGB value 255, respectively they contribute to the composite color with 255.
Thus a pixel can show up as clipped only in response to one of the clipping indication options, not both.
The initial value of the white point is either
Decreasing the white point can induce clipping and it generally increases the contrast and lightness of the displayed image. Increasing the white point causes the image appear darker and less contrasty, but it may be necessary to avoid induced clipping, which can be caused by whute balancing and lightness adjustment. Turning on-off-on-off the clipping indication (by checking-unchecking the respective box) makes the affected pixels appear alternatingly very bright and very dark; this way it is easy to spot the affected areas.
Thus a pixel can show up as clipped only in response to one of the clipping indication options, not both.
| Decreasing the white point down to 2100 from 3692 lightens the image a lot and increases the contrast - but the clipped area too is larger now |
| By turning off the raw clipping indication but keeping the induced indication on shows the effect of the clipping caused by reducing the white point becomes visible |
Keep in eyes, that the clipping indication shows, which colors did not clip, because the clipped pixels are substituted by 0:
The color appearance of light reflecting objects depends on the light source. Sunlight midday, sunlight at the evening, cloudy day, indor lighting with halogen or non-halogen incandescent bulbs, fluorescent light, flashlight, etc. make objects appear in different colors. Additionally, the environment influences the light as well: the reflection from water surfaces, from the side of a forested mountain, from the colored walls of a room, etc. change the light falling on the objects.
In order to display colors like those of the actual scenery, the raw pixel values have to be adjusted relatively to each other, depending on the pixel filter color. This occurs via the multiplication of all black level adjusted pixel values of a channel by a coefficient depending on the sensor and the light source; this is called white balancing.
Some raw processors input and display the white balancing parameters in terms of temperature and tone adjustment. Rawnalyze keeps the parameters of white balancing in form of coefficients: one for the red channel, one for the blue channel and two coefficients for the green channel, seperate values for the two positions of the green pixels of the color filter array blocks. The coefficients will be applied when constructing the image display, and when displaying the mapped histogram and the Apply WB checkbox is checked.
The exposure display is affected only by a zero coefficients: that color is excluded from the display. This is interesting only as a way to exclude one of the green channels.
The current coefficients are displayed with the histograms:
The first value, in this case 2.4878 is the coefficient for the red pixels, the second and third values, here 1.0000 are the coefficients for the green pixels, and the fourth value, in this example 1.1423 is the the coefficient for the blue pixels. The coefficient of the channel in which the highest raw pixel values occur will be 1.0000 (that is the green in most cases), and the other coefficients are selected such way, that they "equalize" the respective pixels in the selected area with the unchanged pixels.
The raw file usually contains a suggestion recorded by the camera for these coefficients; however, the current version of Rawnalyze does not take such specifications into account. The initial white balancing coefficients are all 1.0000, i.e. white balancing does not take place. Rawnalyze offers two methods to carry out white balancing after having read and interpreted the image data:
If there is a spot in the image, which should appear white or grey, then a selection has to be made in that spot. The pixels in the selected area are used for the calculation of the coefficients by the keyboard action Ctrl-W. The evaluated values become immediately active.
Note, that the two green coefficients are always set to the same value by this method.
There is a more sophisticated way to define the coefficients directly, i.e. without selecting a white/grey area in the image, namely in the white balancing dialog. That dialog can be started by the keyboard action Alt-Ctrl-W; it appears in its own small window. The currently active coefficients are displayed initially and then they can be changed. The changes made in that dialog become immediately effective, the result can be viewed in the original dialog, which can be activated by clicking on its window (the white balancing dialog window can be moved out of the way).
| The two green channels are initially linked; the value of the first green channel applies to the second one as well. By clicking on the Green 2 button the slider and input field for the second green channel becomes enabled and its value becomes independent from the first green channel |
The coefficients can be changed by the sliders in steps of 0.05; the change becomes effective immediately. Alternatively, the values can be entered in the input field (then it can be more accurate than 0.05). The entered value becomes effective when the Enter/Return key is pressed (the control remains in that input field), or by the Tab key, which then passes the control to the next input field or to the Accept button.
The entered value must be between 0.0000 and 4.0000. Entering zero "disables" the respective color; this is particularly interesting in exposure display mode: the two green channels can be separated this way.
The Reset button changes the respective value to 1.0000.
The Save button records the currently displayed coefficients in one of the numbered registers. The Recall button reinstates the coefficients of the respective register. These registers preserve coefficients past the lifetime of this dialog. Before registering a setting, all coefficients are 1.0000; thus the white balancing can be inactivated by using a register, which has not been loaded yet.
The Accept button saves the currently active coefficients and closes the dialog.
The Cancel button reinstates the coefficients, which were active when the dialog has been started, and closes the dialog. The coefficients saved in the registers remain preserved.
The five sets of coefficients are accessible outside the white balancing dialog as well: the currently active coefficients can be saved via the keyboard action Ctrl-Shift-1 or Ctrl-Shift-2 or Ctrl-Shift-3 or Ctrl-Shift-4 or Ctrl-Shift-5 in the respective register. The coefficients of a register can be activated any time via the keyboard action Ctrl-1, Ctrl-2, Ctrl-3, Ctrl-4 or Ctrl-5.
The coefficients are reset to 1.0000 when an image is read; however, the registers keep the recorded coefficients. Thus the same coefficients can be used in the analysis of several images.
| The small flowers at the left edge are supposed to be white. The selection statistics show that the green and blue are virtually on the same level, but the red is much lower. The raw clipping indication is turned on to show, where white balance picked should not be done, for clipped pixels do not show the true proportions between the scenery's colors |
| Setting the white balance based on the selected area changes the appearance
of the image; particularly, the small flowers became white.
|
The effect of white balancing on the histogram is demonstrated in the Histogram section.
The unlabelled slider is for lightness adjustment. The adjustment can be made at 1/3 EV, 1/2 EV or 2/3 EV points.
The initial position of this slider is usually at null, but in some cases it shows an adjustment, which is indicated in the raw file. Example: Canon 40D images shot with the Highlight Tone Priority (HTP) option start with an adjustment of +1.
The lightness adjustment is a multiplication of the black level adjusted raw pixel values by a factor according to the adjustment. Its effect is comparable to the "Exposure" adjustment in Adobe Camera Raw. The adjustment can be reset to the initial value by clicking on the Reset button under the slider.
The examples at Black point demonstrate the effect of lightness adjustment.
Die transformation of the linear raw values from a large numerical range, after black level, white balance and lightness adjustment in RGB values occurs based on some mathematical formulas, often called the "gamma function". Every mapping function results in a different distriobution of the contrast, thereby showing more details in some lightness ranges while less details in other ranges.
The default is the formula defined in sRGB. The best way to learn the differences between these is by trying them out.
Note, that
| 
| The left side image was created by sRGB mapping,
while the right side with linear mapping.
|
The exposure display mode is another way to analyze the raw image: selected pixels are displayed with red, green or blue, according to the pixel's position in the color filter array block; however, the displayed intensity is fixed, that does not represent the raw pixels' intensity (i.e. the raw pixel value). In other words: the selection does depend on the raw pixel values, but the pixels appearance does not.
Such a display resembles a "proper" image less than the image display modes, but there is no problem with recognizing a pixel of low value.
The way of pixel selection is the same as that in image display mode, except
Keep in eye, that the limits of the value range are always included in the selection, no matter if the inside or outside range is selected.
An entire color channel (even one of the green channels separated from the other) can be excluded from the display by setting the corresponding white balance coefficient to zero; see the detailed description of the white balance feature in the Image Display section.
The exposure display offers an additional choice: if the BL corrected checkbox is checked, then the black level corrected pixel values are compared against the selection range limits (this is independent on the black point setting); otherwise the original raw pixel values are used in the selection.
The exposure analysis is done best in maximized window: here the small images are presented.
Example 1:
Some cameras offer ISO settings at 1/3 stops as well, but are those "real ISOs"? Let's take an image, shot with ISO 125 and look for a pixel value range, which contains many pixels (the histogram helps finding such ranges). Type in a pixel value in the black point group (don't forget to Enter it) and enter the same value in the white point group as well; after pressing Enter, the control remains in that field. The level 1901 has been selected in both groups. Only red pixels will be displayed at the moment; the color needs to be selected only in the black point group, as that governs the values up to 2047.
|
|
| As the control is still in the level input field (no matter if in the black point
or white point group), pressing the right cursor increases both limits by one.
|
| This shows the red pixels with the value 1904; there are 712 such red pixels in the image |
| There are no red pixels with the value 1905 in the entire image |
Example 2: the saturation level of a certain camera.
| The histogram shows a strange behaviour of the green pixels when clipping: the clipping does not occur at once, it occupies a range of pixel values |
| There are no pixels with 4026 and above |
| but there are some with 4025, so that is that ultimate clipping point here |
| by lowering the upper limit, the outline of a block appears |
| one level lower the block starts to fill out |
| only half of the green pixels appear in the upper part, and a quarter of them in the lower part |
| until the entire block is filled, but only with one of the two green pixels in each color filter array block |
| then from 4009 the other green pixels are appearing |
| and at 4001 the block is almost completely filled |
The importance of knowing the saturation level lies in the clipping indication in image display mode.
There are several ways the distibution of pixel values can be presented. Common between these histograms is, that they are always based on non-demosaiced pixel values, representing the three "channels" of the sensor, and all pixels are involved in the calculation (as opposed to sampling), therefor the histograms are always accurate.
The basic raw histogram
The basic raw histogram will be shown if neither the Mapped
nor the Fine checkbox are checked.
In basic mode the histogram represents the entire range of unmodified raw pixel values. The histogram consists of 512 columns; as the pixel value range is alway larger than 512, one column of the histogram represents a number of pixel values. For example if the numerical range of the pixel values is 0-3900, then each each column represents alternatingly seven or eight consecutive pixel values. If the value range is 4096 (12 bit depth), then each subrange is 8 levels wide; the first column represents the levels 0 to 7, the second one the values 8 to 15 and the 512th column represents the pixel values 4088 to 4095.
Note, that the numerical pixel value range is not always the same as the numerical range of the bit depth. Some cameras create raw data, which fully occupies the range of the bit depth, while the numerical pixel value range with some other cameras (for example Canon DSLRs) can be substantially smaller than the range given by the bit depth. In some cases the range depends on the ISO and/or on the channel. The actual range used in the histogram is declared in the camera's description, which is hard coded in Rawnalyze; it is equal to or slightly greater that the initial white point, and it is always the same for all three channels, even if the saturation levels of the channels are different.
The height of the histogram is 128 pixels; the height of a column represents the number of pixels with the value falling into the subrange of that column in proportion of all pixels of that color. If the sensor contains 8 million pixels, 2 million of those are capturing the red range. If there are 100,000 pixels in the value range 1000 to 1007, then the 126th pixel column of the histogram represents 5% of all red pixels, assumed the range of pixel values is 0-4095.
However, the height of the pixel column is not in direct proportion to the percentage of pixels it is representing, because that way small pixel amounts would not be visible - or the histogram would have to be several thousand pixels tall. Instead, the hight represents the logarythm of the percentage.
A two pixels tall column represents 0.01%; a column as tall as the first white marker from the bottom upwards represents 0.05%, the hight of the next marker is at 0.1%, etc. It is not easy to determine exactly how many pixels are in a column, but it is not important either.
The meaning of a single-pixel column is, that there is at least one pixel in that range, but less than 0.01%. Note, that this can be triggered even by a "hot" or "stuck" pixel.
The short white lines under the histogram mark the exposure value oriented distribution in the levels. If one imagines the right end as EV 0, then the first marker from the right side represents the 1/3 EV lower point, the second marker is at -2/3 EV, the third, somewhat larger marker is at -1 EV, etc. Obviously the range of the highest stop occupies half on the entire histogram.
The yellow dotted pixel column close to the left edge marks the black level (the minimum of the black correction values), the one close to the righ end marks the saturation level. The black level, if visible, is at the same location in all three histrograms; the location of the saturation marker may be different for the colors, depending on the sensor.
The left turquoise dotted pixel column (this does not have to be at the left side) marks the black point, the value is shown in the input field of the black point control group; the right turquoise pixel column represents the white point, the value is displayed in the input field of the white point control group. These two can be at the same location.
If an indicator pixel column is not visible, then it can be imagined at the left respectively right edge of the histogram.
The fine raw histogram
The basic raw histogram conveys an overview of the exposure. This is suitable in many cases;
however, it is somewhat coarse, because each column represents a range of pixel values.
The fine raw histogram shows the detailed distibution of pixel levels, as each column of the histogram
represents a single pixel level. The fine raw histogram will be shown if the Fine checkbox is checked.
The histogram consists of 512 columns, each maximum 128 pixels tall. These 512 columns show a "crop" of the complete fine histogram, starting with the pixel level indicated in the black point field. As the black point value is changed, the histogram's range moves in the total pixel value range, i.e. it represents the crop starting with the new black point value.
The height of the a column represents the number of pixels with the respective value in proportion of all pixels of that color, relative to the total pixel value range, on the same logarythmic scale as in the basic raw histogram.
The white markers at the bottom of the histograms indicate the locations of 1/10th of the 512 wide range. In other words, the markers are in 51 pixels distances. The position of a marker added to the current black point yields the pixel value represented by the column over the marker. For example if the current black point value is 128, then the second marker indicates the position correspoding to the pixel value 230.
Gaps in the histogram indicate, that there is no pixel with the respective value. The gaps in the above example are due to the lossy compression of the image file.
The mapped histogram
If the Mapped checkbox is checked, then the histogram represents the pixel values after following adjustments:
This histogram too is based on non-demosaiced data, i.e. the red, green and blue histograms still represent the pixels with the corresponding color filters. Thus this this histogram reflects the image displayed in composite, channel or greyscale mode, except for the clipping indication.
The appearance of the mapped histogram differs from the basic raw histogram in following aspects:
A particular combination of setting worth of mentioning is, when the black point and white point are at their initial values, the lightness adjustment is null, the mapping method is "Linear" and the white balance is set. The mapped histogram with white balance applied appears like a white balanced raw histogram.
Examples
| This is the initial state of the raw histogram of an image created by a Canon 40D.
The saturation level is 12740 at ISO 160, and the histogram range ends at 12800.
|
| The histogram shows, how well (or not) exposed the shot was. In this case only slightly more than one third of a stop could have been added to the exposure without causing clipping, even though the shot is strongly underexposed (the dynamic range exceeded the camera's capturing capacity) |
| This is the typical sign of clipping caused by specular spots or direct light; in this case the sun was the cuplrit, but it caused less than 0.1% of all pixels clipping |
| The saturation levels of the three colors are
very different with this camera.
|
| This is the effect of the lossy raw compression by some cameras, for example Nikon: groups of levels are simpy left out. The range of these group is increasing at higher pixel values. Therefor there are no gaps, only spikes at the dark end, but the ranges of "gaps" are becoming larger towards the bright end, so far, that there is no pixel in the entire subrange associated with some columns of the histogram, causing the gaps in the histogram |
| The effect of the lossy raw compression up-close, pixel level by pixel level |
| The mapping of the six hundred some different raw values to the 256 RGB values eliminates the gaps, but some strange spikes are still appearing |
| The application of white balance blends the levels more; these spikes are not perceivable on the resulting image when displaying or printing |
The Stouffer wedge is a colorless transparent consisting of strips with transparency 1/3 stop apart. Despite looking boring as a photographic object, it is useful in the explanation of some subjects.
| This is a crop of the image before white balancing. As the strips are colorless, they ought to appear grey; as such they are suitable for white balance picking |
| This is the raw histogram; the spikes/bumps represent the strips. Had they been captured as grey, i.e. red, green and blue with equal intensity, the spikes/bumps would line up nicely |
| The histogram of the mapped data without white balance too shows a shift between the intensities of the colors of strips |
| White balancing multiplied the red pixel values by 1.09 and the green ones by 0.6094, causing the three spikes from the strips lining up |
| As the result, the strips now appear grey |
| Linear mapping of the white balanced image shows the effect of white balancing on the raw data |
| Setting the black point to 500 and the white point to 2000 increases the contrast, while pushing the lower values in black and the higher values in white |
| the histogram shows how the contast is increased: a range of the raw values is mapped on the entire RGB range |
| With black point 1000 and white point 2500, only 1500 levels are mapped on the entire RGB range, which increases the contrast even more |
| the histogram shows, that less strips are mapped to the same RGB range, which in turn increases the spacing between them even more |
Did you know, that JPEG images can be used to hide data? Here is how:
Steganography by CryptoBola