OmniDocHow to choose a picture formula recognition tool: taking photos and screenshots
Mathematical formulas are recognized differently in photos, screenshots and scans. This article explains the applicable scenarios of several types of methods according to the input source, and gives several inspection criteria to check when selecting tools.
Key points of this article
- The clarity, tilt and background of photos, screenshots and scans are very different. You must first select the input source and then look at the tool.
- For a single formula screenshot, it is usually enough to use a formula recognition tool. Full-page scanned papers or handouts are more suitable for OCR first and then sorting.
- When choosing a tool, check whether the output format can be used directly, such as LaTeX, Word formulas or plain text, and whether the layout structure of the original image can be retained.
- No matter which tool is used, the identified formulas must be compared with the original picture item by item, especially superscripts and subscripts, fractions and Greek letters.
- Privacy-sensitive materials (unpublished papers, internal materials) should first confirm the destination of uploading, or use local processing instead.
In class notes, paper PDF illustrations, or handwritten assignments, we often encounter situations where a formula can only be viewed but not modified. If you want to put it into Word, LaTeX or Markdown, you need to first turn the formulas in the picture into editable content. This type of demand is usually searched as "formula photo recognition", "formula screenshot recognition software" or "picture recognition formula editor". The problem behind it is actually very specific: whether you have a photo, a screenshot or a scan, what method should be used, and what should you look for when choosing a tool.
This article only discusses the situation of image input, and does not expand on document translation or entire typesetting. After reading, you can determine which route to take based on the input source, know which scenarios are suitable for online websites, desktop software, and OmniDoc, and also know who to hand over the recognition results for verification and how to verify them.
Whether to take photos, screenshots or scans, how to choose the recognition method? One sentence conclusion
The conclusion is: for screenshots or photos of a single formula, use formula recognition tools directly; for scanned papers and handouts of entire pages, it is more reliable to do OCR first to turn the page into selectable text, and then process the formulas in it.
| your situation | Recommended practices | reason |
|---|---|---|
| One or several formulas taken with a mobile phone, with a clean background | Formula recognition tool, output LaTeX or MathML | The input range is small and the output can be directly pasted into the editor. |
| A small piece of formula cut from a web page or PDF | Formula recognition tools or general chatbots for image viewing | The results are more direct when the screenshots are clear, but you still need to check them one by one. |
| Scanned paper or handout with full page text and formulas | Do OCR first, then export to Word or UniDoc to organize the formulas | The entire page content needs to be identified first. Cutting the formula alone can easily miss the context. |
| Contains unpublished data or internal information | Desktop tool for local processing, or use online service after confirming upload | The content in the picture may be more sensitive than the formula itself |
| Handwritten formulas, scrawled handwriting | First manually sort it into clear print, and then identify | The error rate of handwriting recognition is usually higher than that of print. The blurrier the original image is, the more manual review is needed. |
What are the different methods?
Online formula identification website
Such sites typically open a page, upload an image or paste a screenshot, and return LaTeX or MathML code within seconds. Its advantage is that it does not require installation and is suitable for occasionally processing one or two formulas. The shortcomings are: it is sensitive to resolution and tilt, and complex matrices and multi-line derivation are prone to errors; in addition, pictures will be uploaded to the server, so please read the privacy statement first when it comes to undisclosed content.
Mobile phone camera recognition application
With the recognition function of the mobile phone camera, you can directly select the formula after taking the picture. It is suitable for quick recording in the classroom or library, provided the lighting is even, the paper is flat, and the lens is aligned. If the page is curved or has shadows, the numerator, denominator, and superscript and subscript in the formula may easily be misplaced. It is recommended that you zoom in and check after taking the photo, and take another photo if necessary.
OCR function in desktop office software
Many desktop office software have text recognition functions that can convert scanned documents into editable text. It usually works well for ordinary paragraphs, but formula parts are often treated as ordinary characters, and superscripts, subscripts, and symbols are often lost. If your requirement is to retain the formula structure, this type of function can usually only be used as a preparatory step, and the formula still needs to be re-formatted in the editor after recognition.
General chatbot to view pictures
Send the screenshot to the universal chatbot and ask it to output LaTeX. This method is flexible and can explain the meaning of the formula. But its output is generative, and similar symbols may be misread, or parts not in the picture may be completed. When using it, it is recommended to let it mark the uncertain areas and compare the results with the original picture one by one.
OmniDoc’s OCR and document processing
OmniDoc's desktop client provides a PDF editor that includes OCR functionality for scanned PDFs and can also export PDFs to Word, Excel, or PowerPoint. Document translation on the web supports scanning PDFs and images. It is positioned to process the entire document containing formulas: first turn the scanned page into editable content, and then organize the formulas in Word or UniDoc editor. This is not a direct path if you just need the LaTeX code for a single formula; it's closer to the actual workflow if you're dealing with a full page paper or handout.
📊 Core comparison
| practice | Can the page structure be preserved? | Suitable input amount | privacy considerations | Check costs |
|---|---|---|---|---|
| Online formula recognition website | The page structure is not retained, only the formula code is output. | single or small number of formulas | Pictures will be uploaded, please refer to the terms of service | If there are many formulas, you need to check them one by one. |
| Mobile phone photo recognition application | Does not preserve page structure | Class or field temporary notes | Depends on whether the application is connected to the Internet | When the light and angle are not good, the verification cost will increase significantly. |
| OCR for desktop office software | Paragraphs are retained, formula structures are often lost | Ordinary document | Mainly local processing, depends on the specific software | Need to reformat the formula |
| General chatbot picture viewing | The page structure is not retained, and explanations can be included | few formulas | The image will be uploaded to the corresponding service | Need to check whether you read the wrong symbol |
| OmniDoc's OCR and document processing | Preserves document structure and can be exported to Word | Full page or entire scan | Documents are processed in OmniDoc, and the privacy policy shall prevail on the official website | Need to check the exported formula |
🎯 Which method is suitable for you?
If you only have one or two screenshots of the formula
Just use an online formula to identify the website or a general chatbot to look at the picture. Once you have the code, paste it into the editor and check each superscript and subscript against the original image.
If you need a quick note in class or the library
Mobile phone photo recognition applications are more convenient. When taking pictures, try to keep the paper flat and evenly lit. After taking the picture, zoom in and look at it again before deciding whether to retake the picture.
If you are working with a full page scanned paper or handout
First do OCR, turn the page into editable content, and then organize the formulas in Word or UniDoc. This allows the text, charts, and formulas on the page to remain in context together.
If your material has not been made public yet
Prioritize local processing tools, or first confirm the data destination of online services. Do not upload unpublished data directly to unclear websites for the sake of convenience.
If you need the entire document that can be edited
Working with OmniDoc is easier. It can convert scanned PDFs into a format that can be exported for further editing in UniDoc.
How to process scanned documents in OmniDoc
- Open the PDF or image you want to process in omnidoc.top or a desktop client.
- If it is a scanned PDF, use the OCR function in the PDF editor to make the text on the page selectable and editable.
- Check the OCR results to confirm that the paragraph where the formula is located is not misaligned and the text is in the correct order.
- Export the PDF to Word, or modify the paragraph text directly in the PDF editor.
- Open the exported Word file and check the superscripts, subscripts, fractions and symbols of the formula one by one.
- When LaTeX or online collaboration is required, you can put the organized content into the UniDoc web editor (app.unidoc.top) and continue editing.
❓ Frequently Asked Questions (FAQ)
1. How to choose a free image formula identification tool?
When choosing, look at a few things: the size limit of a single file or page, whether the output results have watermarks, whether the images are uploaded to the server, and whether the formulas can be output into a directly usable LaTeX or Word format. Being able to meet your material volume and privacy requirements is more important than simply looking at price.
2. The formulas taken with the mobile phone always recognize the wrong formula, what should I do?
First check the shooting conditions: whether the light is even, whether the paper is flat, and whether the lens is parallel to the paper. If the picture itself is blurry, it will be difficult for the recognition tool to remedy it, and it is recommended to retake it. After identification, compare it with the original picture, focusing on fraction lines, superscripts and subscripts, and Greek letters.
3. How to use the recognized LaTeX code?
Paste the code into an editor that supports LaTeX and compare it with the original image after compilation. If the rendering result is inconsistent with the original formula, it is usually because a certain symbol or bracket is not recognized accurately, and you can modify it item by item.
4. Are the formula recognition results accurate?
Recognition results require manual verification, and tools cannot ensure that every symbol is correct. Especially for formulas involving papers, examinations or formal reports, checking is a necessary step and cannot be directly regarded as the final result.
📝 Summary
The key to formula photo recognition is to first determine the source of the input: use formula recognition tools for single formula screenshots, scan the entire page and then do OCR before sorting it out, and pay extra attention to clarity and privacy for handwritten or sensitive materials. When choosing a tool, focus on the output format, whether the page structure is preserved, and whether the privacy terms are clear. No matter which method is used, the recognition results must be compared item by item with the original image.
