ChatGPT does not do OCR and does not know how to provide the exact settings (in percentages) from Photoscape X to create a readable image


I want the writing in this image to be seen better. I use a program “Photoscape X pro”. On the left are all the settings, from HRD, MAGIC COLOR, FILTERS, etc. I want you to test the image well, pass it through OCR, see which parts of the text are not clearly visible (so which are not properly readable). Then make and show me all the settings that are necessary to do something to get a clear image.

All those settings are in percentages. I want you to tell me the percentages for each of the settings, for absolutely all, so that the good image comes out

This is ChatGPT answer:

"Again, the optical character recognition (OCR) process was automatically terminated due to timeout, indicating difficulties in processing the loaded image. This situation suggests that the image may be too complex or of a quality that requires additional processing time, exceeding the ability of our current system to complete the task in a timely manner.

Due to the inability to OCR the text and identify unclear or illegible sections, detailed and personalized recommendations for adjustments in Photoscape X Pro, including exact percentages for each of the 24 settings, cannot be accurately provided."

This is the text image. As you can see, it is not a very clear picture.

This are the settings from PhotoScape X (each from 1-100) that must be change in order to obtain a better images. ChatGPT did not know how to give me any exact settings, absolutely nothing coherent.

The controls you show have unknown magnitude of effect. It is hard for me to make recommendations, despite having decades of photo editing experience.

Notable is that there is no sharpening control, such as an unsharp mask that could include a radius, threshold, and strength.

The AI is only as smart as what you put into it. If you know what to ask for, it can be a time saver.

And it can give us an initial report…

The statistical analysis of the image reveals that the pixel value for the 2% probability mass cutoff is 217, while the pixel value for the 98% probability mass cutoff is already at the maximum of 256 (pure white). This suggests that the image is already quite bright with very few dark areas.

Let’s go ahead and process the image contrast in Python to set the 2% point to pure black and adjust the contrast accordingly. Since the 98% point is already pure white, we will not need to adjust for the upper end. I will now process the image contrast

And recommendations…

For the settings in your photo editing program, you should adjust the exposure and contrast controls as follows:

  • Exposure: No adjustment should be needed since the bright areas are already at pure white. However, if your image still appears too bright, you may slightly reduce the exposure until you achieve a satisfactory level of brightness without losing detail in the bright areas.
  • Contrast: Increase the contrast until the darkest parts of your image reach pure black. This will likely be a significant increase since we need to move the 2% pixel value from 217 to 0.

A human-processed image is still better:


Super answer, thanks

so, the prompt must be:

Attached is an image of text that I wish to enhance with a photo editing program that has primarily controls for exposure and contrast. Perform a statistical analysis on the image, finding the 2% and 98% probability mass cutoff of pixel values. Then process the image contrast in python to set the 2% and 98% points to pure black and pure white, and return a download link. Then recommend settings of my image software's available controls to set those points to pure black and pure white.

can you please share the Python Code you are using for this kind of image repairing? I see that you using library such as cv2, numpy and matplotlib

In that example, I was within a chat where ChatGPT Plus has access to a code interpreter environment to run Python, and it had automatically written code to fulfill the goals of the task.

You can just ask any AI, even free ChatGPT, what kind of code you would need to perform such image analysis tasks in Python on a regular basis.