Understanding HDRMerge

HDRMerge’s functionality is very simple in its core. You take a set of raw images of the same subject but with different levels of exposure. You feed these raw images into HDRMerge. It stacks them into layers, with the most-exposed shot at the top. Then, it removes the clipped areas (where the photosites were saturated) of each layer, revealing details of the layers beneath it. As a result, you obtain a new raw image, with the most-exposed pixels of each original image that are not saturated. This raw image contains as much detail and as little noise as possible. HDRMerge provides tools to control the details of this process.

Usage

HDRMerge has both a graphical (GUI) and a command-line (CLI) interface.

The GUI contains a set of editing tools which allow you to decide which parts of each image get transferred into the final output. This is mainly useful for removing ghosts created by objects which moved between images. If there are no such moving objects, the command-line interface is intended to be used in batch processing, e.g. from a script. You can start the batch process either by providing an output filename with the -o switch, or by generating an automatic one with the -a switch. Other switches control the input and output parameters. If neither -o nor -a is given, the program will show the GUI and load the files specified in the command line, if any, but ignoring all other options.

Loading Raw Images

When the GUI is invoked, the program will directly present you the Open dialog:

The Open dialog allows you to select multiple raw images. This dialog can also be shown later with the Open command in the File menu.

Options:

• Align source images. If your images do not align perfectly, check this option to apply the MBT alignment algorithm. The MBT alignment algorithm can correct small misalignments by “translation”. Given the nature of the color organization in a raw image (only Bayer and X-Trans type raw files are supported by HDRMerge) rotations cannot be compensated and translations must be performed in steps of 2 pixels. However, this should be enough, even for steady handheld shots in burst mode, though a tripod is highly recommended to obtain the best results. The alignment algorithm can sometimes fail to detect perfectly aligned images, mostly with highly exposed shots.
• Crop result image to optimal size. If you enabled “align source images” and if your images were actually aligned, check this option to have HDRMerge retain only the common area shared by all images, and crop off the rest. Not doing so could result in artifacts around the periphery, depending on the mask. You can select not crop the image, in which case the output image will cover the same area as the least exposed input, to avoid burnt areas of more exposed shots. One reason for not cropping is if you want to apply a flat-field correction later, as RawTherapee only applies this correction to images that have the same dimensions as the flat-field image.
• Use custom white level. What is a white level? A sensor is made of millions of tiny photo-sensitive elements called photosites or sensels. Each one measures the intensity of the light which falls upon it, and records that intensity as a number - the more light, the higher the number. The bulk of the raw file consists of these recorded measurements. Each photosite has a level beyond which the photosite will not register a change in light intensity even if the light does keep getting brighter - this is the white level. A photosite which cannot record any brighter light is said to be fully saturated, and in post-processing this state is called clipping. For most camera models you do not need to provide a custom white level as one is automatically detected, but for some models the automatic white level is detected incorrectly and so you must provide one manually. You will know when the automatic white level is incorrect when parts of the resulting HDR DNG image which should be clipped white appear as not so - they usually have a strong magenta color cast. Finding the raw white level requires some knowledge - read more about it in RawTherapee’s “Adding support for new raw formats” documentation – or just start at 16000, create a HDR DNG, and if the areas which should be clipped white are not white then subtract 500 and try again - repeat until success.

When you accept a list of input images, the application will show a progress dialog with information of each load step. The process will only succeed if images are compatible. The size, orientation, color filter array pattern and black and white levels must be the same, which generally means that they have been made with the same camera.

Editing

Once the input images are loaded, the interface presents you with the main window:

Most of its space is occupied by a 100% preview of the result. There, input images are stacked on top of each other, and you can see the selected pixels from each layer painted with a different color. You can then pan the result to inspect it.

The tools to modify which pixels of each layer end in the final output are located in the toolbars. They appear at the top by default, but you can move them to any edge of the main window. From left to right, the three main tools allows you to pan the preview, add pixels to the current layer, and remove pixels from it. You can quickly change to the add and remove tools by holding down the Shift and Control keys, respectively. When you release the key, the last tool in use will be selected again. Then, a slider controls the radius of the editing brush. A big radius is useful to wipe large areas, and a small radius can be used for the details. You can also change the size of the brush using the mouse wheel and pressing the Alt key at the same time. A second slider controls the brightness of the preview, for better visualization of the shadow areas. This brightness does not affect the output, just the preview in the GUI.

In a second toolbar, there is a button associated with each layer, starting from the top layer (most exposed shot). You can select the layer that you want to edit with these buttons. The last layer cannot be modified, there is no sense in removing or adding pixels to it. Beware that, since the pixels of the most exposed image are used in the shadow areas, they appear the darkest. To avoid this confusion, if you hover the mouse pointer over each button, a tooltip will show you the exposures steps of that image relative to the least exposed shot. Each button contains the layer number and the color associated with it. You can also select a layer by pressing its number in the keyboard.

When some objects were moving while you took the shots, there will appear “ghosts”. You can use the edit tools to add or remove pixels from each image but the last one, until all the pixels that belong to a moving object only come from one of the input images. To make your job more precise, you will notice that you can only remove pixels that are visible (not covered by a higher layer) and add pixels over the next layer in the stack. Usually, you will want to only remove pixels, starting at the first layer and then going down. Adding too many pixels to a layer may result in burned areas appearing in the result image, so be careful. On the other hand, the pixels of the first layers contain less noise in the shadows. These operations can be undone and redone with the actions of the Edit menu, or by pressing Ctrl+Z and Ctrl+Shift+Z, respectively.

Saving the Result

Once you are satisfied with the preview, the Save HDR command of the File menu generates the output DNG file. It will first ask you for a file name, and then it will present the Save dialog:

You can select the number of bits per sample (16, 24 or 32), the size of the embedded preview (full, half or no preview) and whether to save an image with the mask that was used to merge the input files. The number of bits per sample has an important impact in the output file size. As a rule of thumb, the default value of 16 bits will be enough most of the time. Empirical tests (thanks to DrSlony) show no apparent difference between 16- and 32-bit images, after merging 5 exposures with 2EV steps, despite strong manipulation of shadows/mid-tones/highlights. Nevertheless, if you see some unexpected noise in the shadows of the output image, you can try a 24-bit output. 32 bits will almost never be necessary, but it can be selected anyway.

The Command Line Interface

You can also run HDRMerge without GUI, in batch mode. With the command line switches, you can control all the process described before, except for the edition tools. In this way, HDR images can be created with very little interaction.

The main arguments of the batch mode is a list of raw files to be loaded, and an output switch. The “-o file” switch specifies that “file” will be the name of the output file. The “-a” switch generates this name from the names of the input files.

The rest of the switches control the rest of the process:

• The MTB alignment and crop algorithms are enabled by default, and can be disabled with the switches “–no-align” and “–no-crop”, respectively.
• The “-b bps” switch selects the output bits per sample. The allowed values are 16, 24 and 32.
• The “-p size” switch selects the size of the embedded preview. The allowed values are full, half and none.
• The “-m mask” switch selects the name of the mask file, if you want to save it.

You can also control the level of verbosity of the program with the “-v” and “-vv” switches. The “-v” switch shows progress information, and the “-vv” switch also shows debug information. Finally, the “–help” switch presents a short guide to the command line interface.