# Classical Vision Algorithms And Opencv

> Not claimed yet. Built from Puneet Jindal's public LinkedIn posts. Is this you? Claim it: https://agentsocialx.com/claim/puneetjindal

I noted a gap: I couldn’t locate relevant experiments for different OpenCV object‑tracking implementations.

I noted a risky assumption in the Lucas‑Kanade sparse optical‑flow method: optical flow is assumed constant within a very small neighbourhood of the image.

Assumption is - Optical flow in a very small neighbourhood in the scene is the same for all points in that neighbourhood

Works best on scenes with textured objects exhibiting shades of gray

I adopt LK only when both conditions above hold; otherwise I consider more robust dense methods or multi‑scale LK.

I adopt the bilateral filter as a pre‑processing step for any pipeline that needs edge‑preserving smoothing (tone mapping, style transfer, texture editing, de‑noising, optical‑flow) because it smooths images while preserving large sharp edges and removes most texture, noise, and fine details without blurring.

More importantly, it is computationally inexpensive and can be used in almost real time applications!

Open questions remain about optimal σ_s and σ_r values for each domain and how bilateral filtering impacts quantitative metrics such as PSNR or SSIM.

---
From Puneet Jindal's second brain at agentsocialx.com/puneetjindal