OmniDocHow to identify formulas through screenshots: Convert images to editable formulas
Explain the complete process from formula screenshots to editable formulas: screenshot framing, identification, checking symbols one by one, pasting into the document, and explaining the situations where other output forms need to be used.
Key points of this article
- When taking a screenshot, only frame the formula itself, leaving a small amount of margin. The fraction, superscript, subscript, and square root must be completely included in the frame. Most of the errors in recognition start from here.
- First determine whether the image is a pure image: PDFs that can select text do not require OCR, only image pages need to be recognized.
- The identification results must be checked one by one, focusing on superscripts and subscripts, Greek letters, bracket pairs and the numerator and denominator of fractions.
- Select the output format according to the destination: use plain text to only view the content, use formula objects to put into typesetting documents, and keep the LaTeX source code when editing is needed.
Many people see a formula in a paper, lecture notes or web page and want to put it into their own document to modify it, but they can only get a picture. Take a screenshot and paste it directly. The formula cannot be modified, searched or copied. What is needed at this time is not the name of a tool, but a stable operation sequence: how to intercept, how to identify, how to check after identification, and finally which form to use when putting it into the document.
This article talks about the complete workflow of the formula screenshot recognition method. Readers can turn a picture of a formula into a formula that can be modified by following the steps. Even if you don't have OmniDoc at hand, the screenshot specifications, symbol checks, and output format choices here apply.
One thing needs to be said first: machine recognition of formulas does not ensure that every symbol is correct, especially complex and multi-layered nested formulas. Each step below includes manual verification. Please refer to the original text for important content and confirm again.
How to turn a screenshot of a formula into an editable formula? One sentence conclusion
First determine which category your image belongs to, and then decide whether to identify it directly or convert it to a page suitable for identification. After taking the screenshot, be sure to check the symbols one by one, and then select the output format according to the final use.
| your situation | Recommended practices | reason |
|---|---|---|
| The formula comes from a PDF that can select text | Directly copy the text or source code in the formula area without taking a screenshot | The text layer itself is usable content, but screenshots will reduce the clarity. |
| The formula is a separate screenshot or photo | Crop to the area containing only formulas, and check them one by one after identifying them. | There are many margins and irrelevant text, which can easily interfere with the recognition results. |
| The page where the picture is located is a scanned copy | OCR PDF in OmniDoc desktop client and check | OCR is used to extract text from image pages, suitable for full page content |
| Put the formula into the document and continue editing | Keep the LaTeX source code and then convert or paste it in the editor | The source code can completely preserve superscripts, subscripts, fractions and other structures to facilitate subsequent modifications. |
| Just need to understand the formula content | Just recognize it as ordinary text, no need to pursue typesetting | Correct content is more important than beautiful layout. Avoid unnecessary adjustments. |
Before you start: Confirm these points first
1. Is the picture a clear screenshot or a blurry photo? When the resolution is too low, it is difficult to distinguish small superscripts, subscripts, and parentheses, such as x from χ, 1 from l. When taking a screenshot, try to enlarge the page display ratio as much as possible, and then take the screenshot, which will have a much better effect than enlarging the picture afterwards.
2. Is the formula printed or handwritten? The recognition stability of printed formulas is generally higher than that of handwritten formulas. There are more errors after handwritten formula recognition. It is recommended to manually re-enter key parts first and then compare with the recognition results.
3. Where the formula is finally placed. If it is put into a Word document or PDF report, it only needs to be visually correct; if it is to be continued to be deduced, modified or typeset elsewhere, the editable structure must be retained and not just a string of ordinary characters.
4. Whether the page where the formula is located has other important content. When taking full-page screenshots or full-page OCR, the main text, headers, footers, and legends will be mixed into the results. Deciding ahead of time what you want to identify can save you a lot of cleanup work.
5. Whether your document allows external processing. For formulas containing unpublished research data or confidential information, you should first confirm the data management requirements of your unit or team before deciding whether to upload it to the online tool.
Steps to extract formulas from image pages using OmniDoc
The following steps apply when the page where the formula is located is a scan or picture page. If your formula itself can select text, skip OCR and copy directly.
- Crop the screenshot range. Use the system's built-in screenshot tool to frame only the formula itself, leaving a little blank space at the top, bottom, left, and right. The upper and lower boundaries of fractions, radicals, and summation symbols should be fully included to avoid cutting off superscripts or subscripts.
- Save as image or single page PDF. If you only have an image file, you can use the system print function to save it as a PDF file for subsequent unified processing. The printing functions of most operating systems support this method.
- Determine whether it is a picture page. Open the PDF and try to select a piece of text near the formula with the mouse. If it cannot be selected, it means it is just a picture and requires OCR; if it can be selected, it means there is a text layer and you can just copy it.
- Open the file in the OmniDoc desktop client. The client can be downloaded at download.omnidoc.top and supports Windows, macOS and Linux. Open the PDF file you prepared.
- Perform OCR on the image page. Use OCR in the PDF editing function to recognize scanned pages and extract text and symbols in images into selectable content. Note that the goal of OCR is to extract page text, and the structure of complex formulas may be lost, so the results of this step require manual inspection.
- Locate and copy the formula result. Find the text area corresponding to the formula on the recognized page and copy it paragraph by paragraph. If there are garbled characters or breaks in the results, go back to the screenshot, crop it again, and identify it again. It is usually faster than manual repair.
- Paste it into the target document and organize the format. Paste the results into a Word or UniDoc document. If you need a standardized formula object, you can re-enter it using the formula function in the UniDoc web editor (app.unidoc.top), or keep the LaTeX source code for subsequent modification.
- Save and export. Save the document after confirming that the formula is correct. If you want to hand it over to others for editing later, it is recommended to save a version with source code at the same time so that others can continue to modify it.
Check each item after completion
The areas that are prone to errors after formula identification focus on several types of symbols. When checking, don't just look at whether the whole thing is smooth, but check the original drawings item by item.
| Check items | How to see | What to do if there is a problem |
|---|---|---|
| superscript and subscript | Compare the original image to confirm whether the position and content of each superscript and subscript are consistent. | Manually adjust in the editor, or re-screenshot the corresponding area |
| Greek letters and similar characters | Check whether α, a, λ, l, μ, u, etc. are confused | Make judgments based on the context of the original text and change to standard writing if necessary. |
| bracket pairing | Count the number of left and right brackets and make sure they are not misplaced. | After completing the parentheses, check whether the range of the fraction and radical is correct. |
| Fractional structure | Make sure that the numerator and denominator are not split into two lines of ordinary text | Use the formula editing function to reconstruct fractions |
| Numbers and symbols | Check for confusion between 0 and O, 1 and l, minus sign and dash, multiplication sign and letter x | Correct them one by one according to the original picture, especially the numerical formulas |
| overall meaning | Substitute a simple numerical value or compare the context to see if the conclusion is reasonable. | If the meaning does not make sense, go back to the original text and confirm it. Do not modify based on your feelings. |
📊 Core comparison
The recognition results have three common output forms, which are suitable for different scenarios. You can compare and select according to the table below.
| Output form | Suitable for the scene | Main limitations |
|---|---|---|
| plain text symbols | You only need to understand the formula content, or quote it into chats and notes | The superscript and subscript and fraction structures are incomplete and not suitable for direct use in typesetting. |
| formula object | Put it into a formal report, paper draft or handout, and it will look visually close to the printed effect. | It needs to be restructured in the document editor, and you must be familiar with the formula editing function when modifying it. |
| LaTeX source code | Requires long-term maintenance, repeated modifications or interoperability with LaTeX documents | The readability of the source code is low, and direct reading is not as intuitive as the typesetting results. |
When choosing, you can ask yourself a question: Who is using this formula in my workflow and what it is used for. Just look at the conclusion and choose pure text. If you want to deliver formatting, choose formula objects. If you want to continuously modify it, choose source code.
❓ Frequently Asked Questions (FAQ)
1. What are the free ways to identify image formulas, and what should you look for when choosing?
When choosing a free method, you can focus on several things: file size and page limits, whether the output results are watermarked, whether the files will be uploaded and how they are saved, and whether the formatting and formula structure are retained. These are more decisive for the actual experience than the simple word "free". If you are dealing with unpublished research content, you must first confirm the privacy policy.
2. The identified formula is always wrong, what should I do?
First go back to the original image and check whether the screenshot is clear and whether the superscript and subscript have been cut off by cropping. Most errors come from improper image quality or scope. Re-capturing a clearer area and narrowing it down before identifying it can often significantly improve the results. Key formulas still need to be manually checked.
3. Can this method be used for handwritten formulas?
You can try, but the effect is usually not as stable as print. Handwritten formulas are prone to character confusion and structural misalignment. It is recommended to identify them first as a reference, then manually re-enter the key parts, and finally compare them item by item with the original manuscript.
4. Can the formula recognition results be used directly in Word?
Ordinary text results can be pasted directly, but structures such as superscripts, subscripts, and fractions may need to be readjusted. If you want to get a standardized formula object, you can use the formula function to reconstruct it in the document, or keep the LaTeX source code and then convert it. After pasting, be sure to check whether the format is consistent with the original text.
📝 Summary
The key to identifying formula screenshots is not to find which tool, but to do a solid job in the four steps of screenshots, identification, verification and output. First frame the correct formula and determine whether it is a picture page, then use OCR to extract the content, check the superscripts, subscripts and symbols one by one, and finally select the output format according to the purpose. If every step is done well, the formula screenshot recognition method can stably turn pictures into editable formulas.
