Dsp Filters

Virtual Rescanning Features (VRS)

What is Virtual Rescanning

Virtual Rescanning, or VRS for short, is a method for altering an existing image without the necessity of setting hardware changes in your scanning device and/or actually rescanning an image.

The goal to any VRS process is to improve the utility of the images you are scanning or have scanned in the past. All VRS processes in this system utilize what are called filters. Filters are logical algorithms designed to affect a particular change upon an image. These filters can be used separately or in many combinations together to improve their overall effectiveness depending upon your needs. Two different examples are as follows.

• PROBLEM: Your data is all printed on 8”x14” landscape formatted paper, and your scanner only scans in portrait for that size. When you need to view the images, they must all be rotated to view them effectively.

• VRS SOLUTION: A VRS rotation filter can be utilized while scanning to eliminate the need for the user to rotate page by page to review data from the images.

• PROBLEM: You have only hard copy of a series of reports which need to be integrated into the new accounting software. All the data is in dot matrix, and is not very legible when scanned to the Optical Character Recognition engine, thereby making retyping of the complete report highly likely.

• VRS SOLUTION: Apply a series of filters in varying settings, “Deskew”, “Noise Remove”, “Dilate” and “Smoothing” to give the OCR system legible characters for effective use. Only minor verifying is then required.

Optidoc2 has integrated a variety of filters which can be used individually, or in combination. You can apply these filters on every page as it is being scanned to improve the results from the scanner, or it can be applied to specific pages, on the user’s discretion at a later time. In the Image Processing menu, filter groups can be defined to be utilized in either of the above mentioned ways.

How to define a filter in Image Processing

You can define a filter for image processing easily. The Image Processing menu is located in two separate areas. It can be accessed directly through the image, by right clicking the image itself.

As you can see the Image Processing menu option is located near the bottom of the “right click” menu.

This same drag down menu can be accessed through the Input Window, and right clicking on the page you wish to perform the VRS upon.

Select the menu item, “Image Processing..” and you will see the “Define Filter” window. To begin the filter definition, you would type a name into this window

In this example we will reduce the image to the picture portion of the image only.

We will call the Filter Name “Picture only”. Once you have decided upon a name you then must give the name a definition. “Left Click” the button labeled “Definition…”

You then get to see the IP (image processing) Configuration window.

The list on the left is the available filters implemented in Optidoc2. You can then select single filters to add into your “Picture Only” filter.

The first filter I will select is “Deskew”. Once you select the filter in the left window, you then have the buttons in the middle of the window become active and if there is this feature available for the function, the “Configure” button becomes active as well.

Hit the Add button. Then go to the configure button on the far right of the IP Configuration window to make certain that the “Deskew” fills the deskewed area correctly. To do this hit the “Configure” button on the right to get this window. Choose fill color as white if the overscan of the document is white, or black if it is black. (it depends upon the image your scanner generates)

In this instance I have chosen White as the fill color and used Both as the direction to deskew the image. Then hit the OK button to accept your configuration changes.
The next filter to add to this is “Noise Removal” to get rid of any extraneous dots on the image. Then add the filter “Line Removal” and configure to repair and reconstruct lines, both in the vertical and horizontal. Then Add “Invert Image” to invert for the next filter, which will be “Black Overscan Removal” which will crop off the left, top and right, then “Invert Image” again and finally Crop, to crop out the bottom extraneous white field.

The filter hereby defined will provide the following result from the original badly skewed image.

There is great utility, both in file size improvement and in potential image quality from utilizing VRS features. What follows will be a filter-by-filter explanation of the different VRS features available in Optidoc2.

Filters and Undo

There are sixteen different types of VRS implemented in Optidoc2 that can be used individually or in combination with each other. They can be implemented also on single pages at the user’s discretion, or on all pages in a batch if processed while scanning.

Undo Image Processing- when you can and can’t

The Undo feature in this system works in conjunction with the VRS implementation. However, it is a single step feature. Also, Undo image processing only works with the individual page-by-page application of the filters. The reason the Undo feature is not implemented in batch mode is to conserve memory utilization.

The undo feature is active once a filter has been applied to a document single page of a document. It is applicable to that particular page until the project is closed, after that the Undo feature is no longer active for that page. If you were to close a project, whatever state the page with the Image Processing Filters were applied to will become a static state.

When filters are applied, the undo feature is a single step undo feature, but that is a single step per page to which VRS has been applied. If 4 filters are applied at once, the undo will remove all filters applied. If 4 filters are applied, in the same order, but applied one by one, the Undo Image Processing will only un-apply the last filter. Hence the utilizing the filters is a process to be cautious with. There are many ways to improve an image, but it should be taken slowly and with great caution. Filters that work great on improving one type of image, may not work well with others.

What follow are descriptions of the individual Filters that are implemented in Optidoc2, and how they work individually. There are numerous ones that work better in combination with each other to improve an image than individually, depending upon the particular image with which you are working. These descriptions will come from “definitions” of how these filters work and from experiential trial and error here at ATS. We make no guarantees as to their individual or combined utility. They can help lots, but be careful.

Black Overscan Removal Filter

This filter is particularly useful if you have images with black regular shaped borders around them. The most common situation this filter comes from is from flatbed scanning with the top not closed. Such as when you photocopy a page with the top open, then wish to trim off the excess black.

This is a prime example of such a black rimmed document. As you can see there is a small bit of image in the middle and then since the imaging device was able to get no data back from the open flat bed scanner, you get a page of whatever size you set the scanner for scanning, with the part that is not the image being black.

Open the Image Processing Window and name the filter you wish to define. We’ll use black border remove for this example.

Pick the filter named Black Overscan Removal, and add it to the filter.

The configure key does not do a lot for this particular filter, unless you are processing inverted images.

This filter may need to be applied multiple times to remove all of the border from images. The filter removes borders in parallel lines to the edges of the image. If you have a skewed image, deskew first, then utilize black overscan remove. Here is a before and after for this filter.

Origional image.

This image has one application of the Black Overscan Removal.
Image has a small white strip at the top removed, and border pulled from right side and top compared to the one above.

This image has a second application of the Black Overscan Removal filter.

All border is removed from sides and more from top and bottom of the image.

This image has a third application of the filter and the image is now cropped perfectly top and sides, but the bottom seems to still be there. However, the image view window in this instance is set to display as “fit to width”. To check to see if the border is removed, rotate the doc 90 degrees to see if it still fits the window tightly in that orientation.

With the document rotated you can see the border has been cropped effectively with 3 passes.

Border Removal

This filter is useful to remove borders, as its name implies. There is nothing configurable in this filter. It is useful for document that has odd shapes around the image, generally due to methods of imaging.

This image has an unnecessary white box at the bottom. Apply “Border Remove” one time and you get the picture below.

The extraneous white box at the bottom is gone, and the black border is gone as well. If this were an image that required printing, it would be much improved from a cost perspective due to the removal of the black border. Further filters could be applied so as to crop the image to the picture with no border at all.

Crop

This feature should be called “AutoCrop” as you have no method of defining the area that it should crop.

You should be careful with applying this filter automatically, because can do extreme things to your images.

Here is an image that is needing cropping, due to the overabundance of whiteness around the important part of the image.

Here is the before image.

Here is the after image, once Crop is supplied. White was removed from the right side, bottom and top of the image. To crop the image further, apply the “Noise Removal” filter, then re-apply the Crop feature to make the image picture only.

Deskew

This feature is used to correct the orientation from slight misfeeds during the scanning process. It is especially useful when later Optical Character Recognition will be applied to the images. OCR engines recognize text more effectively when Deskewed.

Here is a skewed image. It is rotated off the 90-degree mark. You may easily re-orient this image to square with the “Deskew” filter in VRS.

Deskew is a filter with a number of configurable settings.

The settings assist in applying the filter correctly to the image for your desired results.

Mode – tells the filter if the image is a Graphic or Text.

Operating Mode – how to proceed with the Deskew (generally Detect angle and deskew is the most effective

Fill Color – what color the rotated image will have on the edges

Direction – defines how the image will be deskewed

Here is the result of Deskew applied once to the above image with Mode – Text, Operating Mode – Detect Angle and Deskew, Fill Color – Black and Direction - Both.

Here is the result of Deskew applied once to the above image with Mode – Text, Operating Mode – Detect Angle and Deskew, Fill Color – White and Direction - Both.

Dilation

This filter is good for improving images which are scanned from faded originals. An Example of this would be scanning a receipt tape from a journal roll that needs a new ribbon.

As you can see from the image there is some fuzziness to the numbers. The legibility of this image can be improved greatly by Dilating the text.

What Dilation does is add pixels onto a bitonal image, increasing the size of whatever group of black pixels it encounters. This dilation will, in effect “darken” the image.

Original image

Dilated image.

As you can see the dilation increases the darkness of the text in the image, but it also increases the noise on the image.

Erosion

This filter is the opposite of Dilation. It reduces pixels on a bi-tonal image, by reducing the thickness of the black area’s edges. In effect it thins down black lines.

Here is an original image.

This is a bit on the dark side, due to the color areas that are gray in the bi-tonal scan we have. If you apply the Erosion filter you then have the following.

This is significantly lighter, but some of the text is a bit spotty.

If after Erosion is applied, you then apply the Dilate filter you will in effect have pulled out most of the gray in the image, as you see from this third image.

Halftone Removal

This particular filter is effective for removing halftones from an image. (halftones are the manner in which a bi-tonal process will define any color that is not black or white, basically patterns of dots)

Here is an original image with some halftone

If you apply the halftone filter all the non-black or non-white areas will be stripped.

Halftone removed

Hole Removal

If you have documents with punch holes in them, you can use this filter to remove the visible aspects of the holes in the image.

Here is a document with holes in it.

Here is the document with the Hole Removal applied.

Invert Image

This filter reverses the pixels of the image. All black pixels are changed to white, and all white pixels are changed to black.

Here is an example

This filter can be used in conjunction with other filters. One example would be to apply Invert, then Overscan Removal, so as to crop an image where the normal Crop filter is not identifying the image correctly.

Line Removal

This filter assists in working with lines on images. It is called line removal but it can be used to reconstruct somewhat damaged lines. This is useful in conjunction with other filters and in Optical Character Recognition processes

The removal assists the OCR module to recognize text more effectively. Here is a before sample of a lined document.

Set the configuration button for Line removal to, Mode - remove line, Horizontal – enable with curvature straight, Vertical – enable with curvature straight.

Here is the same image with one instance of line removal implemented.

In the configuration module of this filter, instead of removing lines, you can set the filter to rebuild lines that are broken, which also assists in other VRS operations. You also have the opportunity to only remove/ rebuild horizontal or vertical lines depending upon your settings in the configure settings.

Noise Removal

This filter will often be used when there are small imperfections in the data that you wish to “clean” off the image. In the sample below, a receipt tape has some detail issues. Many of the specks can be removed as seen below.

Origional

Noise Removed

Rotation

Rotation can be applied a number of formats. It does not require the scanned to have Detect Deskew on to be utilized. Once the filter is defined, you then can customize it’s application with the details tab.

You can set multiple filters, like rotate right, or left, instead of clock-wise or counter-clockwise. The names are self explanatory, though mirror is less common. It works as is detailed below.

Original

Mirror

Scaling

This filter is utilized to increase or decrease the size of the actual tiff from the original image. In this instance, the original was 3.38” x 8.48”. The filter could then be used to effectively fill a single page. In this instance the original tiff when the Scaling filter is applied at 8.5” x 11” increased the image size to 4.35” x 10.94”. The aspect ratio is held constant, but the image is scaled larger.

Though increasing the image size will have no detrimental effect on the image quality, reduction in image size could. If you scale an 8.5”x14” image and to fit on 8.5”x11” paper there will be a loss of detail in the image, thought the dpi will remain the same.

Skeleton

This takes the lines and color vaiations and minimizes the colored area into a variation of a line drawing. It does some dramatic things on Pictures as seen below

Original

Skeleton

Here is how it functions in bold text.

With Skeleton applied.

Smoothing