**Disclaimer:** All photos attached are randomly picked online because they closely resemble my dataset!
I am trying to extract some information from an image like the one I attached below. It has tons of background lines and small texts embedded.
**Some background**
Assuming a photo of this is taken with a high-end phone camera (40mp), and this photo is uploaded to OCR engine like Google Cloud Vision (GCV). GCV is powerful enough to return background texts which will interfere with my keyword search later. Hence, some preprocessing is needed before the image is sent to an OCR engine.
I need to get all the large texts and the text below the profile picture. I have tried several methods:
1. Gamma correction
2. CLAHE
3. Extract only GRAY-BLACK region pixels in HSV color space.
My method works only if the image is captured in a controlled environment. In real life, the performance is greatly affected by uneven lighting conditions. Sometimes the background texts appear more prominent than the foreground texts.
Would appreciate any input! Thanks!
↧