esrXP (
Download Page – warning: don't click on the misleading banner ads on this site)
English and Chinese explanations of the software can be found at this site: (Link)
You can download these settings for reference. You will most probably have to customize the settings to suit your needs. These are registry files; simply extract the archive and double-click on the file, confirm that you would like to add the registry settings, and the settings will show up in the program when you run it.
Settings from Area11 – 
AREA11esrXP-scaningSETUP.zip 2.9K
75 downloads
Settings used for K-pop Ultimate Audition –
esrXP_settings_KPUA.zip 1.55K
56 downloads
One characteristic of this software is that the original color images cannot be saved. Only the filtered black and white image can be saved. The black and white images are smaller than color images, and can be used with OCR, however it may be more difficult to read if manually transcribed.
Main window.
Typical Workflow
File > Open Video... – From the dialog window, select the video file or AviSynth file you wish to work on. The program may stall for a moment while the video is loading.
Subtitle > Filter – A window opens up to with the video image. This window is used to set single–frame detection options.
Use the seek bar at the bottom to find a frame with subtitles.
If you don't see an image in the filter window, your computer probably doesn't have the proper codecs installed. In my case, I installed the K–Lite Mega Codec Pack to fix the issue, although one of the smaller bundles might have been sufficient. (You may already have a codec pack installed on your computer. It is not recommended that you install more than one codec pack, as it may cause conflicts.)
Create a box to specify the area you want analyzed. A box that is tight around the subtitles is less likely to generate false positives.
- Left–click on the image to set the top–left corner of the box.
- Right–click on the image to set the bottom–right corner of the box.
- The Region controls in the left–hand pane can also be used to adjust the size and location of the box.
- Clicking the center of the Region controls toggles a horizontal guide that divides the selection area into two halves. If the subtitles appear on two lines, the horizontal guide can be used to center the selection area. If you intend on using OCR, it is a good idea to center the selection area between the two lines. Otherwise, it's not too important.
The high-lighted area denotes the selected area.
Filter Setting Controls
Filters can be designed specifically for a certain style of subtitle. This methodology will generate fewer false positives, but also means that different filters will need to be created to fit different subtitle styles. The filters can also be designed to be more general so that they work with many different styles, but more false positives will likely be generated.
Enable Filter – display the filtered output. The desired output will have the text in white, and other areas in black.
Additional Color – when Enable Filter is selected, the filter will be displayed with multiple colors to illustrate the 3 filtering stages.
- Red for areas that pass through the Outline filter
- Green for areas that are not red and are not classified as text
- Black for areas that are not red and do not pass through the Pass1 filter
- Blue for areas that are not red and do not pass through the Final filter
- White for areas that are not red, and pass through the Pass1 and Final filters and postprocessing settings
Select Color from the drop-down list if the subtitles does not have an outline.
Select Color + Outline from the drop-down list if the subtitles have an outline.
Subtitle Color – Opens a color selection window. Select the color of the subtitles.
Outline Color – Opens a color selection window. Select the color of the subtitle outline.
Color selection window.
Advance – Opens a dialog for configuring the 3 filtering stages. Have Enable Filter and Additional Color selected to visualize your settings. If you are trying new settings, it is easier to uncheck all the filtering options first, and add each filter one at a time.
Click on the checkbox to enable the desired filtering options.
- Hue Difference and RGB Difference are dependent on the Subtitle Color or Outline Color that was selected. (The Outline filter uses Outline Color, while Pass1 and Final filters use Subtitle Color. If you don't use any of the Difference settings, it doesn't matter what color setting you choose.) All pixels within the difference threshold pass through the filter.
- Enable Min to create a high-pass filter (everything below the minimum is filtered out). If the outline is white, you can set a high-pass filter to retain the bright colors and eliminate the dark colors.
- Enable Max to create a low-pass filter (everything above the maximum is filtered out). If the outline is black, you can set a low-pass filter to eliminate the bright colors and retain the dark colors.
- Enable both Min and Max to create a band-pass filter (everything below the minimum or above the maximum is filtered out).
- Pixel Compensate sets the tolerance level on areas colored in black (from Pass1). Set at zero, if there is even one pixel that is classified as black in an area, that area will not be classified as text.
Allowable values:
Pixel Compensate [0 – 6]
Hue Difference [0 – 181]
RGB Difference [0 – 255]
Lum, Sat [0 – 100] (Min must be less than Max)
Postprocessing – Opens a dialog to select clean–up options, after the image is filtered. (see the guide linked above for more details)
Filtering Example:
Unfiltered selection.
In this example, the outline and subtitle colors have been selected as black and white respectively.
Selecting Enable Filter and Additional Color without any filtering options enabled will result in the selected area to be filled with red.
Stage 1: Outline – Filter out areas not associated with the outline. The outline will show up in red, other areas will show up in green or white. You don't want the red the bleed into the text. The aim is to setup a filter that only has the pixels of the outline passing through. If all the boxes are unchecked, all the pixels pass through the filter, meaning that all the pixels are a part of the outline. Sometimes the color in the background will be similar to the subtitle outline and thus will show up red as well (which is perfectly normal).
For the Outline filter, Lum Max and RGB Difference are enabled and set to 70 and 172 respectively.
Stage 2: Pass1 – Filter out areas not associated with subtitles. Areas that are filtered out will show up in black. This filter does not have to be too accurate.
After stage 2, all areas that are fully surrounded by red, absent of any black are filled in with white. The Pixel Compensate parameter can be increased to provide some tolerance for the black pixels. Any areas within the tolerance will be filled with white as well.
For the Pass1 filter, RGB Difference and Sat Max are enabled and set to 81 and 10 respectively. Pixel Compensate is set to 1.
Stage 3: Final – Filter out any areas that are colored in white. Areas that are filtered out will show up in blue. This filter can be used for fine–tuning which areas are detected.
For the Final filter, Sat Max is enabled and set to 6.
A black and white filtered image can be seen with Additional Color disabled.
Filtered image after postprocessing options are enabled (remove block touch edge, remove single pixel dot, remove block larger than 15x15 pixels).
Unfiltered selection.
Comment
Filtering settings may work well in one frame, but they might not work well with other frames. It's a good idea to verify that the filter works well at many points in the video and to leave some tolerance when filtering.
Subtitle > Rip Option – A small dialog box presents options for frame–to–frame detection.
Skipping more frames will mean faster processing, however the timing will be less accurate.
Pixel Difference and Ignore Change (%) settings modify the sensitivity of when a new line is detected.
- Start – Click the start button at the bottom, and then sit back and wait for the program to finish executing. The time required to finish processing the video will depend on the options that were selected, the video, and your computer's processing power. It usually takes me about 30 minutes to 1 hour to process a 350MB rmvb file (but this is on a 5-year old laptop, running Windows as a virtual machine, so it could be much faster). You may click stop at any time, but the program cannot resume, and will need to start from the beginning again. The captured images will start popping up on the right hand side. If the images are not good, stop the program and modify the settings.
The program will apply the filter and postprocessing settings to all the frames that it analyzes. If text is detected in a frame (areas classified as white), the frame will be compared to adjacent frames to see if it's a new line, or the continuation of a line. If many frames without text are detected as having text, you might want to adjust the filter settings. If the program is not detecting line changes properly (many duplicate lines, or missing lines) you may wish to adjust the rip options.
Subtitle > Manager – The manager displays all the extracted images. You can go through the images and delete any unnecessary ones.
- Left–click an image to highlight it for deletion. You can also click–and–drag to highlight multiple images. Left–click on the image again to deselect it.
- To delete a series of images, right–click on the first one, and left–click on the last one.
- To merge a duplicate line, right–click on the first one, and right–click on the last one. You will see arrows on the right side of the first and last images to indicate the range. The last line will be highlighted in a different shade. This indicates which image will be retained when the duplicates are deleted. Left–click any image within the range to select a different image to be retained. Right–click on any image in the range to cancel the merge. By default, the manager will only allow you to merge lines if the timing is continuous. To override this, you can check the "Force Merge" box in the toolbar.
- After selecting the images for deletion, click on the red X in the toolbar to commit the deletion.
- Use Recover Filtered on areas where text was erroneously filtered out.
- Use Remove Passed to remove falsely detected areas.
Manger Window
Images highlighted for deletion.
Merging a range of images: note the blue arrows on the right. The lighter-colored image will be retained.
File > Save As...
From Save as type, you can select the format you wish to save in.
esrXP (*.esr) – save the state of the current session
SubRip with bitmap (*.srt) – save images and .srt file
Saving Your Settings and Batch ProcessingOn the right end of the toolbar in the main window, there is a combo box, save button and delete button
. By writing a description in the combo box, and clicking save, your current settings will be saved. Use the combo box to switch between your saved settings.
After saving your settings, you can setup batch processing (under the
File menu). This is helpful as you will typically need to process a video once for the dialog, a second time for explanations, and a third time for lyrics.
Select the saved setting you wish to use from the drop-down list, and click add video, to add jobs to the list.
Exporting Images for OCRIf you wish to use OCR, export the images using
Save OCR Image.
If subtitles are captured on two lines, select the checkbox.
See
this post for further details.
Creating an SRT File with text instead of image filenamesOn the left side of the main window, you can type out the subtitles with each line corresponding to one image. If you used OCR, you can paste the OCR text into this area as well. Choose SRT from the Save As... dialog box to create an SRT file with text instead of the image filenames.
