This article provides hints to generating pure white backgrounds or alpha transparent backgrounds with Shutter Stream version 2.1 and higher.
|Product Photography Background Removal: Magic Wand Tool||Using magic wand to remove background in version 6.1|
|Practical Algorithms for Product Photography Background Removal||Overview of background removal tools in version 6.1|
|iPhone Product Photography & Removing Backgrounds||Using iPhones to do background removal in version 6.1|
This Knowledge Base article focuses on the background removal tool. For readers that are more video inclined, here are a few video links related to background removal:
- Automated background removal (BR) tool: here
- BR Tool used with lasso to remove additional frame (ex. shadow): here
- BR Tool used with lasso to add frame back in: here
- Green screen tutorial for BR: here
- Eliminating shadows with Iconasys lumipads: here
- Automated background removal guide: here
Step-by-Step Guide to Efficient Use of Shutter Stream's Background Removal Tools
Shutter Stream 2.1 supports three types of background removal algorithms:
Background Image Removal
Use the "background image" algorithm when you can capture a foreground and a background image, without changing any of the camera settings or camera position. You can capture multiple foreground images for a single background image.
Solid Color (i.e. Green Background Removal)
Solid Color Removal
Use the "solid color" algorithm when the background color is red, green or blue and the background is distinctly different than any of the colors in the foreground object. The algorithm works best when the background color is green, red or blue. It is NOT recommended to use this algorithm if the background is not red, green or blue! Remember this: due to demosaicing patterns of most modern CCD cameras, green pixels are twice as dense as the red and blue pixels and hence the optimal background color is green. For this reason we sometimes refer to this algorithm as the Green Background Removal.
The "solid color" algorithm can be used to remove colored objects from the image. This can be used efficiently to create "floating objects." For example, you can use a green mannequin to put clothing articles and then remove the mannequin. A second example would be to provide a green support to a piece of jewelry and then remove the green support. In all cases we strongly recommend that the color of the objects to be removed should be red, green or blue and the foreground objects should have no background color in them.
Manual Removal and Addition Via the Lasso Tool
Manual Removal and Addition of RGB Pixels
The lasso tool (and all background removal tools) automatically converts all images to an RGBA image. The fourth channel is the Alpha channel. When a region is cut, the lasso tool sets the corresponding alpha values to zero, indicating full transparency. However, the RGB values of the original image remain untouched. When using the lasso tool, users have the option of removing (transparent alpha) or adding (opaque alpha) the RGB pixels from/into view. The lasso tool is now a powerful tool that can be used for manual background removal after one of the other background removal tools has been used:
a) If the automated background removal was too aggressive in removing pixels, you can add them back in by selecting the region with the lasso tool and then choosing Opaque Alpha as the lasso option.
b) If the automated background removal was not very aggressive and there are additional pixels that you'd like to remove, then select Transparent Alpha when you apply the lasso tool.
The lasso tool can also be applied in a batch process now, making it that much more useful for editing multiple images at once.
Background Image Removal
When using the "background image" removal algorithm keep in mind the following:
- First, the main challenge of this algorithm are shadows. When strong shadows are present in the foreground image (i.e. in the image with the object) the shadows can be miss-interpreted as object.
- Second, shadows can be removed in several different ways, one of which is bottom light panels. The following page has a pretty good example on using the algorithm in conjunction with light panels:
- Third, for larger objects, such as mannequins, make sure the background is flooded with lights that are set between the mannequin and the foreground. This tends to minimize shadows on the background.
- When applying the algorithm, start off with all the control variables very small, close to zero
then increase the threshold slowly until you're satisfied with your results or you start seeing holes creep into the object.
- Once you start seeing holes appear in the object, you can start increasing the "Hole Fill Radius" slightly. Play with a combination of the Threshold and Hole Fill Radius until you achieve your desired results.
- If optimal results cannot be achieved automatically then use the lasso tools from the edit tool box to manually remove the background. There are two types of lasso tools
- The first lasso tool cuts away from the object to reveal a solid color. Use this when you want to cut away background.
- The second lasso tool cuts away from the object to reveal a background object. When setting the background object to be the foreground image, this is equivalent to adding back parts of the original object.
- Use Edge Sensitivity and Edge Blending to achieve slightly artistic edge artifacts of your final image. Experimentation with these parameters is recommended.
Solid Color Removal (Green Background Removal)
The solid color removal algorithm has only two parameters to adjust (in this example, the image is a green image with the green color being removed):
- Smoothing. This value should be kept very low. Ideally a 1 or 2, unless the image is of an extremely high resolution.
- Soft Transparency. This is the main value you need to adjust. Start with a small value and then increase it until you achieve the desired effect.
Questions and Answers
|1||Why can't I preview the entire background removed image, like I was able to preview in previous versions of Shutter Stream?||The previous version of Shutter Stream applied the BR tool to a down-sampled image. However, that preview was not always 100% accurate. Applying the BR tool to a down-sampled image is not always equivalent to applying the BR tool to a full size image. In version 2.1 we've added "hole filling" which significantly differentiates the performance of the BR on the best fit image versus the full size image. Because of these non-correspondence issues between the full size image and the best fit image, we had to remove the "preview entire image" feature. Instead, we've added a view 1-to-1, where a window can be moved in different locations on the image and we show you a true preview of the actual 1-to-1 image.|
If you want to see the preview of the entire image, you will have to hit the Apply button and view the final result that way.
|2||When doing BR, why does the screen flash in the background?|
The flashing screen highlights the regions of the image where the background is removed. From our internal experiments, the flashing screen helps users pinpoint the areas of the screen where the background is removed. Unfortunately, the flashing screen can cause problems for some users. This feature can be disabled by unchecking Blink transparent area.
|3||We are currently having problems using the background removal tool (Version 2.1 or higher). Instead of removing the background to a solid white, it seems to be making it transparent.|
Starting with version 2.1 and higher, background removed images are now saved internally as RGBA images. The original RGB pixels are never edited and instead the background visibility is controlled via the alpha channel. You can now set the background color via the Alpha blending color option from the settings menu.
When you export your images to formats that support only RGB (i.e. JPG) the alpha blending color will be applied. If you export your images to an RGBA capable image format, such as a PNG or TIFF, then the output will be an RGBA image.
In Version 2.1, if you require a background removed RGB PNG/TIFF image, you can import the RGBA TIFF/PNG image and then export as desired. Alternatively, you could export the images as JPG, then import them back into Shutter Stream and then export them again as TIFF/PNG images. When you export them as JPG the background color will be automatically applied. After you import them again, the images are only RGB so when you export them the second time as TIFF/PNG, they will be exported as RGB TIFF/PNG with the background color applied.
Starting with Version 3.0 you can force RGB only images for TIFF and PNG file format exports. When forcing an RGB format (versus an RGBA format) the background color is automatically applied to the output.
For printing purposes, the lossless TIFF or PNG RGBA image can be read into Photoshop and then printed with any desired background color.
|4||When the “solid color” option is selected it throws off the color balance of the product.||Please make sure that you are using the solid color option only when the background color is red, green or blue. The algorithm works best when the background color is green, red or blue. It is NOT recommended to use this algorithm if the background is not red, green or blue! Remember this: due to demosaicing patterns of most modern CCD cameras, green pixels are twice as dense as the red and blue pixels and hence the optimal background color is green. For this reason, this algorithm is sometimes referred as the Green Background Removal algorithm.|